diff --git a/.gitignore b/.gitignore index 66d1beb2..ae3261f0 100644 --- a/.gitignore +++ b/.gitignore @@ -25,6 +25,8 @@ examples/ time.log celerybeat-schedule.db search_results.json +redbear-mem-metrics/ +pitch-deck/ api/migrations/versions tmp diff --git a/api/LICENSE b/LICENSE similarity index 100% rename from api/LICENSE rename to LICENSE diff --git a/api/app/aioRedis.py b/api/app/aioRedis.py index aac2aa84..dfb63dad 100644 --- a/api/app/aioRedis.py +++ b/api/app/aioRedis.py @@ -1,6 +1,8 @@ import asyncio import json import logging +import os +import threading from typing import Dict, Any, Optional import redis.asyncio as redis @@ -21,6 +23,50 @@ pool = ConnectionPool.from_url( ) aio_redis = redis.StrictRedis(connection_pool=pool) +_REDIS_URL = f"redis://{settings.REDIS_HOST}:{settings.REDIS_PORT}" + +# Thread-local storage for connection pools. +# Each thread (and each forked process) gets its own pool to avoid +# "Future attached to a different loop" errors in Celery --pool=threads +# and stale connections after fork in --pool=prefork. +_thread_local = threading.local() + + +def get_thread_safe_redis() -> redis.StrictRedis: + """Return a Redis client whose connection pool is bound to the current + thread, process **and** event loop. + + The pool is recreated when: + - The PID changes (fork, Celery --pool=prefork) + - The thread has no pool yet (Celery --pool=threads) + - The previously-cached event loop has been closed (Celery tasks call + ``_shutdown_loop_gracefully`` which closes the loop after each run) + """ + current_pid = os.getpid() + cached_loop = getattr(_thread_local, "loop", None) + loop_stale = cached_loop is not None and cached_loop.is_closed() + + if not hasattr(_thread_local, "pool") \ + or getattr(_thread_local, "pid", None) != current_pid \ + or loop_stale: + _thread_local.pid = current_pid + # Python 3.10+: get_event_loop() raises RuntimeError in threads + # where no loop has been set yet (e.g. Celery --pool=threads). + try: + _thread_local.loop = asyncio.get_event_loop() + except RuntimeError: + _thread_local.loop = None + _thread_local.pool = ConnectionPool.from_url( + _REDIS_URL, + db=settings.REDIS_DB, + password=settings.REDIS_PASSWORD, + decode_responses=True, + max_connections=5, + health_check_interval=30, + ) + + return redis.StrictRedis(connection_pool=_thread_local.pool) + async def get_redis_connection(): """获取Redis连接""" @@ -44,10 +90,8 @@ async def aio_redis_set(key: str, val: str | dict, expire: int = None): val = json.dumps(val, ensure_ascii=False) if expire is not None: - # 设置带过期时间的键值 await aio_redis.set(key, val, ex=expire) else: - # 设置永久键值 await aio_redis.set(key, val) except Exception as e: logger.error(f"Redis set错误: {str(e)}") diff --git a/api/app/cache/memory/activity_stats_cache.py b/api/app/cache/memory/activity_stats_cache.py index 6b162cdd..e0008353 100644 --- a/api/app/cache/memory/activity_stats_cache.py +++ b/api/app/cache/memory/activity_stats_cache.py @@ -10,7 +10,7 @@ import logging from typing import Optional, Dict, Any from datetime import datetime -from app.aioRedis import aio_redis +from app.aioRedis import get_thread_safe_redis logger = logging.getLogger(__name__) @@ -68,7 +68,7 @@ class ActivityStatsCache: "cached": True, } value = json.dumps(payload, ensure_ascii=False) - await aio_redis.set(key, value, ex=expire) + await get_thread_safe_redis().set(key, value, ex=expire) logger.info(f"设置活动统计缓存成功: {key}, 过期时间: {expire}秒") return True except Exception as e: @@ -90,7 +90,7 @@ class ActivityStatsCache: """ try: key = cls._get_key(workspace_id) - value = await aio_redis.get(key) + value = await get_thread_safe_redis().get(key) if value: payload = json.loads(value) logger.info(f"命中活动统计缓存: {key}") @@ -116,7 +116,7 @@ class ActivityStatsCache: """ try: key = cls._get_key(workspace_id) - result = await aio_redis.delete(key) + result = await get_thread_safe_redis().delete(key) logger.info(f"删除活动统计缓存: {key}, 结果: {result}") return result > 0 except Exception as e: diff --git a/api/app/cache/memory/interest_memory.py b/api/app/cache/memory/interest_memory.py index 108e2a37..2881f06c 100644 --- a/api/app/cache/memory/interest_memory.py +++ b/api/app/cache/memory/interest_memory.py @@ -9,7 +9,7 @@ import logging from typing import Optional, List, Dict, Any from datetime import datetime -from app.aioRedis import aio_redis +from app.aioRedis import get_thread_safe_redis logger = logging.getLogger(__name__) @@ -62,7 +62,7 @@ class InterestMemoryCache: "cached": True, } value = json.dumps(payload, ensure_ascii=False) - await aio_redis.set(key, value, ex=expire) + await get_thread_safe_redis().set(key, value, ex=expire) logger.info(f"设置兴趣分布缓存成功: {key}, 过期时间: {expire}秒") return True except Exception as e: @@ -86,7 +86,7 @@ class InterestMemoryCache: """ try: key = cls._get_key(end_user_id, language) - value = await aio_redis.get(key) + value = await get_thread_safe_redis().get(key) if value: payload = json.loads(value) logger.info(f"命中兴趣分布缓存: {key}") @@ -114,7 +114,7 @@ class InterestMemoryCache: """ try: key = cls._get_key(end_user_id, language) - result = await aio_redis.delete(key) + result = await get_thread_safe_redis().delete(key) logger.info(f"删除兴趣分布缓存: {key}, 结果: {result}") return result > 0 except Exception as e: diff --git a/api/app/celery_app.py b/api/app/celery_app.py index 807c59f4..23fd82ed 100644 --- a/api/app/celery_app.py +++ b/api/app/celery_app.py @@ -1,5 +1,6 @@ import os import platform +import re from datetime import timedelta from urllib.parse import quote @@ -11,21 +12,24 @@ from app.core.logging_config import get_logger logger = get_logger(__name__) + +def _mask_url(url: str) -> str: + """隐藏 URL 中的密码部分,适用于 redis:// 和 amqp:// 等协议""" + return re.sub(r'(://[^:]*:)[^@]+(@)', r'\1***\2', url) + # macOS fork() safety - must be set before any Celery initialization if platform.system() == 'Darwin': os.environ.setdefault('OBJC_DISABLE_INITIALIZE_FORK_SAFETY', 'YES') # 创建 Celery 应用实例 -# broker: 任务队列(使用 Redis DB,由 CELERY_BROKER_DB 指定) -# backend: 结果存储(使用 Redis DB,由 CELERY_BACKEND_DB 指定) +# broker: 优先使用环境变量 CELERY_BROKER_URL(支持 amqp:// 等任意协议), +# 未配置则回退到 Redis 方案 +# backend: 结果存储(使用 Redis) # NOTE: 不要在 .env 中设置 BROKER_URL / RESULT_BACKEND / CELERY_BROKER / CELERY_BACKEND, # 这些名称会被 Celery CLI 的 Click 框架劫持,详见 docs/celery-env-bug-report.md -# Build canonical broker/backend URLs and force them into os.environ so that -# Celery's Settings.broker_url property (which checks CELERY_BROKER_URL first) -# cannot be overridden by stray env vars. -# See: https://github.com/celery/celery/issues/4284 -_broker_url = f"redis://:{quote(settings.REDIS_PASSWORD)}@{settings.REDIS_HOST}:{settings.REDIS_PORT}/{settings.REDIS_DB_CELERY_BROKER}" +_broker_url = os.getenv("CELERY_BROKER_URL") or \ + f"redis://:{quote(settings.REDIS_PASSWORD)}@{settings.REDIS_HOST}:{settings.REDIS_PORT}/{settings.REDIS_DB_CELERY_BROKER}" _backend_url = f"redis://:{quote(settings.REDIS_PASSWORD)}@{settings.REDIS_HOST}:{settings.REDIS_PORT}/{settings.REDIS_DB_CELERY_BACKEND}" os.environ["CELERY_BROKER_URL"] = _broker_url os.environ["CELERY_RESULT_BACKEND"] = _backend_url @@ -45,8 +49,8 @@ celery_app = Celery( logger.info( "Celery app initialized", extra={ - "broker": _broker_url.replace(quote(settings.REDIS_PASSWORD), "***"), - "backend": _backend_url.replace(quote(settings.REDIS_PASSWORD), "***"), + "broker": _mask_url(_broker_url), + "backend": _mask_url(_backend_url), }, ) # Default queue for unrouted tasks @@ -77,6 +81,7 @@ celery_app.conf.update( # Worker 设置 (per-worker settings are in docker-compose command line) worker_prefetch_multiplier=1, # Don't hoard tasks, fairer distribution + worker_redirect_stdouts_level='INFO', # stdout/print → INFO instead of WARNING # 结果过期时间 result_expires=3600, # 结果保存1小时 @@ -103,6 +108,9 @@ celery_app.conf.update( 'app.core.memory.agent.long_term_storage.time': {'queue': 'memory_tasks'}, 'app.core.memory.agent.long_term_storage.aggregate': {'queue': 'memory_tasks'}, + # Clustering tasks → memory_tasks queue (使用相同的 worker,避免 macOS fork 问题) + 'app.tasks.run_incremental_clustering': {'queue': 'memory_tasks'}, + # Document tasks → document_tasks queue (prefork worker) 'app.core.rag.tasks.parse_document': {'queue': 'document_tasks'}, 'app.core.rag.tasks.build_graphrag_for_kb': {'queue': 'document_tasks'}, diff --git a/api/app/controllers/__init__.py b/api/app/controllers/__init__.py index 585de2ed..869eb039 100644 --- a/api/app/controllers/__init__.py +++ b/api/app/controllers/__init__.py @@ -8,11 +8,13 @@ from fastapi import APIRouter from . import ( api_key_controller, app_controller, + app_log_controller, auth_controller, chunk_controller, document_controller, emotion_config_controller, emotion_controller, + end_user_controller, file_controller, file_storage_controller, home_page_controller, @@ -69,6 +71,7 @@ manager_router.include_router(chunk_controller.router) manager_router.include_router(test_controller.router) manager_router.include_router(knowledgeshare_controller.router) manager_router.include_router(app_controller.router) +manager_router.include_router(app_log_controller.router) manager_router.include_router(upload_controller.router) manager_router.include_router(memory_agent_controller.router) manager_router.include_router(memory_dashboard_controller.router) @@ -96,5 +99,6 @@ manager_router.include_router(file_storage_controller.router) manager_router.include_router(ontology_controller.router) manager_router.include_router(skill_controller.router) manager_router.include_router(i18n_controller.router) +manager_router.include_router(end_user_controller.router) __all__ = ["manager_router"] diff --git a/api/app/controllers/app_controller.py b/api/app/controllers/app_controller.py index e9b539df..74991bcf 100644 --- a/api/app/controllers/app_controller.py +++ b/api/app/controllers/app_controller.py @@ -65,16 +65,42 @@ def list_apps( - 默认包含本工作空间的应用和分享给本工作空间的应用 - 设置 include_shared=false 可以只查看本工作空间的应用 - 当提供 ids 参数时,按逗号分割获取指定应用,不分页 + - search 参数支持:应用名称模糊搜索、API Key 精确搜索 """ + from sqlalchemy import select as sa_select + from app.models.api_key_model import ApiKey + workspace_id = current_user.current_workspace_id service = app_service.AppService(db) - # 当 ids 存在且不为 None 时,根据 ids 获取应用 + # 通过 search 参数搜索:支持应用名称模糊搜索和 API Key 精确搜索 + if search: + search = search.strip() + # 尝试作为 API Key 精确匹配(API Key 通常较长) + if len(search) >= 10: + matched_id = db.execute( + sa_select(ApiKey.resource_id).where( + ApiKey.workspace_id == workspace_id, + ApiKey.api_key == search, + ApiKey.resource_id.isnot(None), + ) + ).scalar_one_or_none() + if matched_id: + # 找到 API Key,直接返回关联的应用 + ids = str(matched_id) + + # 当 ids 存在时,根据 ids 获取应用(不分页) if ids is not None: app_ids = [app_id.strip() for app_id in ids.split(',') if app_id.strip()] - items_orm = app_service.get_apps_by_ids(db, app_ids, workspace_id) - items = [service._convert_to_schema(app, workspace_id) for app in items_orm] - return success(data=items) + if app_ids: + items_orm = app_service.get_apps_by_ids(db, app_ids, workspace_id) + items = [service._convert_to_schema(app, workspace_id) for app in items_orm] + # 返回标准分页格式 + meta = PageMeta(page=1, pagesize=len(items), total=len(items), hasnext=False) + return success(data=PageData(page=meta, items=items)) + # ids 为空时,返回空列表 + meta = PageMeta(page=1, pagesize=0, total=0, hasnext=False) + return success(data=PageData(page=meta, items=[])) # 正常分页查询 items_orm, total = app_service.list_apps( diff --git a/api/app/controllers/app_log_controller.py b/api/app/controllers/app_log_controller.py new file mode 100644 index 00000000..92b5becd --- /dev/null +++ b/api/app/controllers/app_log_controller.py @@ -0,0 +1,89 @@ +"""应用日志(消息记录)接口""" +import uuid +from typing import Optional + +from fastapi import APIRouter, Depends, Query +from sqlalchemy.orm import Session + +from app.core.logging_config import get_business_logger +from app.core.response_utils import success +from app.db import get_db +from app.dependencies import get_current_user, cur_workspace_access_guard +from app.schemas.app_log_schema import AppLogConversation, AppLogConversationDetail +from app.schemas.response_schema import PageData, PageMeta +from app.services.app_service import AppService +from app.services.app_log_service import AppLogService + +router = APIRouter(prefix="/apps", tags=["App Logs"]) +logger = get_business_logger() + + +@router.get("/{app_id}/logs", summary="应用日志 - 会话列表") +@cur_workspace_access_guard() +def list_app_logs( + app_id: uuid.UUID, + page: int = Query(1, ge=1), + pagesize: int = Query(20, ge=1, le=100), + is_draft: Optional[bool] = None, + db: Session = Depends(get_db), + current_user=Depends(get_current_user), +): + """查看应用下所有会话记录(分页) + + - 支持按 is_draft 筛选(草稿会话 / 发布会话) + - 按最新更新时间倒序排列 + - 所有人(包括共享者和被共享者)都只能查看自己的会话记录 + """ + workspace_id = current_user.current_workspace_id + + # 验证应用访问权限 + app_service = AppService(db) + app_service.get_app(app_id, workspace_id) + + # 使用 Service 层查询 + log_service = AppLogService(db) + conversations, total = log_service.list_conversations( + app_id=app_id, + workspace_id=workspace_id, + page=page, + pagesize=pagesize, + is_draft=is_draft + ) + + items = [AppLogConversation.model_validate(c) for c in conversations] + meta = PageMeta(page=page, pagesize=pagesize, total=total, hasnext=(page * pagesize) < total) + + return success(data=PageData(page=meta, items=items)) + + +@router.get("/{app_id}/logs/{conversation_id}", summary="应用日志 - 会话消息详情") +@cur_workspace_access_guard() +def get_app_log_detail( + app_id: uuid.UUID, + conversation_id: uuid.UUID, + db: Session = Depends(get_db), + current_user=Depends(get_current_user), +): + """查看某会话的完整消息记录 + + - 返回会话基本信息 + 所有消息(按时间正序) + - 消息 meta_data 包含模型名、token 用量等信息 + - 所有人(包括共享者和被共享者)都只能查看自己的会话详情 + """ + workspace_id = current_user.current_workspace_id + + # 验证应用访问权限 + app_service = AppService(db) + app_service.get_app(app_id, workspace_id) + + # 使用 Service 层查询 + log_service = AppLogService(db) + conversation = log_service.get_conversation_detail( + app_id=app_id, + conversation_id=conversation_id, + workspace_id=workspace_id + ) + + detail = AppLogConversationDetail.model_validate(conversation) + + return success(data=detail) diff --git a/api/app/controllers/end_user_controller.py b/api/app/controllers/end_user_controller.py new file mode 100644 index 00000000..b9d54fea --- /dev/null +++ b/api/app/controllers/end_user_controller.py @@ -0,0 +1,48 @@ +"""End User 管理接口 - 无需认证""" + +from app.core.logging_config import get_business_logger +from app.core.response_utils import success +from app.db import get_db +from app.repositories.end_user_repository import EndUserRepository +from app.schemas.memory_api_schema import ( + CreateEndUserRequest, + CreateEndUserResponse, +) +from fastapi import APIRouter, Depends +from sqlalchemy.orm import Session + +router = APIRouter(prefix="/end_users", tags=["End Users"]) +logger = get_business_logger() + + +@router.post("") +async def create_end_user( + data: CreateEndUserRequest, + db: Session = Depends(get_db), +): + """ + Create an end user. + + Creates a new end user for the given workspace. + If an end user with the same other_id already exists in the workspace, + returns the existing one. + """ + logger.info(f"Create end user request - other_id: {data.other_id}, workspace_id: {data.workspace_id}") + + end_user_repo = EndUserRepository(db) + end_user = end_user_repo.get_or_create_end_user( + app_id=None, + workspace_id=data.workspace_id, + other_id=data.other_id, + ) + + logger.info(f"End user ready: {end_user.id}") + + result = { + "id": str(end_user.id), + "other_id": end_user.other_id or "", + "other_name": end_user.other_name or "", + "workspace_id": str(end_user.workspace_id), + } + + return success(data=CreateEndUserResponse(**result).model_dump(), msg="End user created successfully") diff --git a/api/app/controllers/file_storage_controller.py b/api/app/controllers/file_storage_controller.py index 55149cce..4e1ba74c 100644 --- a/api/app/controllers/file_storage_controller.py +++ b/api/app/controllers/file_storage_controller.py @@ -14,6 +14,9 @@ Routes: import os import uuid from typing import Any +import httpx +import mimetypes +from urllib.parse import urlparse, unquote from fastapi import APIRouter, Depends, File, HTTPException, Request, UploadFile, status from fastapi.responses import FileResponse, RedirectResponse @@ -290,6 +293,101 @@ async def upload_file_with_share_token( ) +@router.get("/files/info-by-url", response_model=ApiResponse) +async def get_file_info_by_url( + url: str, +): + """ + Get file information by network URL (no authentication required). + + Fetches file metadata from a remote URL via HTTP HEAD request. + Falls back to GET request if HEAD is not supported. + Returns file type, name, and size. + + Args: + url: The network URL of the file. + + Returns: + ApiResponse with file information. + """ + api_logger.info(f"File info by URL request: url={url}") + + try: + async with httpx.AsyncClient(timeout=10.0) as client: + # Try HEAD request first + response = await client.head(url, follow_redirects=True) + + # If HEAD fails, try GET request (some servers don't support HEAD) + if response.status_code != 200: + api_logger.info(f"HEAD request failed with {response.status_code}, trying GET request") + response = await client.get(url, follow_redirects=True) + + if response.status_code != 200: + api_logger.error(f"Failed to fetch file info: HTTP {response.status_code}") + raise HTTPException( + status_code=status.HTTP_400_BAD_REQUEST, + detail=f"Unable to access file: HTTP {response.status_code}" + ) + + # Get file size from Content-Length header or actual content + file_size = response.headers.get("Content-Length") + if file_size: + file_size = int(file_size) + elif hasattr(response, 'content'): + file_size = len(response.content) + else: + file_size = None + + # Get content type from Content-Type header + content_type = response.headers.get("Content-Type", "application/octet-stream") + # Remove charset and other parameters from content type + content_type = content_type.split(';')[0].strip() + + # Extract filename from Content-Disposition or URL + file_name = None + content_disposition = response.headers.get("Content-Disposition") + if content_disposition and "filename=" in content_disposition: + parts = content_disposition.split("filename=") + if len(parts) > 1: + file_name = parts[1].strip('"').strip("'") + + if not file_name: + parsed_url = urlparse(url) + file_name = unquote(os.path.basename(parsed_url.path)) or "unknown" + + # Extract file extension from filename + _, file_ext = os.path.splitext(file_name) + + # If no extension found, infer from content type + if not file_ext: + ext = mimetypes.guess_extension(content_type) + if ext: + file_ext = ext + file_name = f"{file_name}{file_ext}" + + api_logger.info(f"File info retrieved: name={file_name}, size={file_size}, type={content_type}") + + return success( + data={ + "url": url, + "file_name": file_name, + "file_ext": file_ext.lower() if file_ext else "", + "file_size": file_size, + "content_type": content_type, + }, + msg="File information retrieved successfully" + ) + + except HTTPException: + raise + except Exception as e: + api_logger.error(f"Unexpected error: {e}") + raise HTTPException( + status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, + detail=f"Failed to retrieve file information: {str(e)}" + ) + + @router.get("/files/{file_id}", response_model=Any) async def download_file( request: Request, @@ -476,8 +574,12 @@ async def get_file_url( # For local storage, generate signed URL with expiration url = generate_signed_url(str(file_id), expires) else: - # For remote storage (OSS/S3), get presigned URL - url = await storage_service.get_file_url(file_key, expires=expires) + # For remote storage (OSS/S3), get presigned URL with forced download + url = await storage_service.get_file_url( + file_key, + expires=expires, + file_name=file_metadata.file_name, + ) url = _match_scheme(request, url) api_logger.info(f"Generated file URL: file_id={file_id}") @@ -688,7 +790,7 @@ async def permanent_download_file( # For remote storage, redirect to presigned URL with long expiration try: # Use a very long expiration (7 days max for most cloud providers) - presigned_url = await storage_service.get_file_url(file_key, expires=604800) + presigned_url = await storage_service.get_file_url(file_key, expires=604800, file_name=file_metadata.file_name) presigned_url = _match_scheme(request, presigned_url) return RedirectResponse(url=presigned_url, status_code=status.HTTP_302_FOUND) except Exception as e: @@ -697,3 +799,44 @@ async def permanent_download_file( status_code=status.HTTP_500_INTERNAL_SERVER_ERROR, detail=f"Failed to retrieve file: {str(e)}" ) + + +@router.get("/files/{file_id}/status", response_model=ApiResponse) +async def get_file_status( + file_id: uuid.UUID, + db: Session = Depends(get_db), +): + """ + Get file upload/processing status (no authentication required). + + This endpoint is used to check if a file (e.g., TTS audio) is ready. + Returns status: pending, completed, or failed. + + Args: + file_id: The UUID of the file. + db: Database session. + + Returns: + ApiResponse with file status and metadata. + """ + api_logger.info(f"File status request: file_id={file_id}") + + # Query file metadata from database + file_metadata = db.query(FileMetadata).filter(FileMetadata.id == file_id).first() + if not file_metadata: + api_logger.warning(f"File not found in database: file_id={file_id}") + raise HTTPException( + status_code=status.HTTP_404_NOT_FOUND, + detail="The file does not exist" + ) + + return success( + data={ + "file_id": str(file_id), + "status": file_metadata.status, + "file_name": file_metadata.file_name, + "file_size": file_metadata.file_size, + "content_type": file_metadata.content_type, + }, + msg="File status retrieved successfully" + ) diff --git a/api/app/controllers/mcp_market_config_controller.py b/api/app/controllers/mcp_market_config_controller.py index 0f2da3b0..6f27d87a 100644 --- a/api/app/controllers/mcp_market_config_controller.py +++ b/api/app/controllers/mcp_market_config_controller.py @@ -91,9 +91,11 @@ async def get_mcp_servers( try: cookies = api.get_cookies(token) + headers=api.builder_headers(api.headers) + headers['Authorization'] = f'Bearer {token}' r = api.session.put( url=api.mcp_base_url, - headers=api.builder_headers(api.headers), + headers=headers, json=body, cookies=cookies) raise_for_http_status(r) @@ -173,6 +175,7 @@ async def get_operational_mcp_servers( url = f'{api.mcp_base_url}/operational' headers = api.builder_headers(api.headers) + headers['Authorization'] = f'Bearer {token}' try: cookies = api.get_cookies(access_token=token, cookies_required=True) @@ -260,7 +263,9 @@ async def create_mcp_market_config( api.login(create_data.token) body = {'filter': {}, 'page_number': 1, 'page_size': 1, 'search': None} cookies = api.get_cookies(create_data.token) - r = api.session.put(url=api.mcp_base_url, headers=api.builder_headers(api.headers), json=body, cookies=cookies) + headers = api.builder_headers(api.headers) + headers['Authorization'] = f'Bearer {create_data.token}' + r = api.session.put(url=api.mcp_base_url, headers=headers, json=body, cookies=cookies) raise_for_http_status(r) except Exception as e: api_logger.warning(f"Token validation failed for ModelScope MCP market: {str(e)}") @@ -290,9 +295,11 @@ async def create_mcp_market_config( 'search': "" } cookies = api.get_cookies(token) + headers = api.builder_headers(api.headers) + headers['Authorization'] = f'Bearer {token}' r = api.session.put( url=api.mcp_base_url, - headers=api.builder_headers(api.headers), + headers=headers, json=body, cookies=cookies) raise_for_http_status(r) @@ -393,7 +400,9 @@ async def update_mcp_market_config( api.login(update_data.token) body = {'filter': {}, 'page_number': 1, 'page_size': 1, 'search': None} cookies = api.get_cookies(update_data.token) - r = api.session.put(url=api.mcp_base_url, headers=api.builder_headers(api.headers), json=body, cookies=cookies) + headers = api.builder_headers(api.headers) + headers['Authorization'] = f'Bearer {update_data.token}' + r = api.session.put(url=api.mcp_base_url, headers=headers, json=body, cookies=cookies) raise_for_http_status(r) except Exception as e: api_logger.warning(f"Token validation failed for ModelScope MCP market: {str(e)}") diff --git a/api/app/controllers/memory_agent_controller.py b/api/app/controllers/memory_agent_controller.py index e3d2bf92..aa4d48e3 100644 --- a/api/app/controllers/memory_agent_controller.py +++ b/api/app/controllers/memory_agent_controller.py @@ -118,142 +118,142 @@ async def download_log( return fail(BizCode.INTERNAL_ERROR, "启动日志流式传输失败", str(e)) -@router.post("/writer_service", response_model=ApiResponse) -@cur_workspace_access_guard() -async def write_server( - user_input: Write_UserInput, - language_type: str = Header(default=None, alias="X-Language-Type"), - db: Session = Depends(get_db), - current_user: User = Depends(get_current_user) -): - """ - Write service endpoint - processes write operations synchronously - - Args: - user_input: Write request containing message and end_user_id - language_type: 语言类型 ("zh" 中文, "en" 英文),通过 X-Language-Type Header 传递 - - Returns: - Response with write operation status - """ - # 使用集中化的语言校验 - language = get_language_from_header(language_type) - - config_id = user_input.config_id - workspace_id = current_user.current_workspace_id - api_logger.info(f"Write service: workspace_id={workspace_id}, config_id={config_id}, language_type={language}") - - # 获取 storage_type,如果为 None 则使用默认值 - storage_type = workspace_service.get_workspace_storage_type( - db=db, - workspace_id=workspace_id, - user=current_user - ) - if storage_type is None: storage_type = 'neo4j' - user_rag_memory_id = '' - - # 如果 storage_type 是 rag,必须确保有有效的 user_rag_memory_id - if storage_type == 'rag': - if workspace_id: - knowledge = knowledge_repository.get_knowledge_by_name( - db=db, - name="USER_RAG_MERORY", - workspace_id=workspace_id - ) - if knowledge: - user_rag_memory_id = str(knowledge.id) - else: - api_logger.warning( - f"未找到名为 'USER_RAG_MERORY' 的知识库,workspace_id: {workspace_id},将使用 neo4j 存储") - storage_type = 'neo4j' - else: - api_logger.warning("workspace_id 为空,无法使用 rag 存储,将使用 neo4j 存储") - storage_type = 'neo4j' - - api_logger.info( - f"Write service requested for group {user_input.end_user_id}, storage_type: {storage_type}, user_rag_memory_id: {user_rag_memory_id}") - try: - messages_list = memory_agent_service.get_messages_list(user_input) - result = await memory_agent_service.write_memory( - user_input.end_user_id, - messages_list, - config_id, - db, - storage_type, - user_rag_memory_id, - language - ) - - return success(data=result, msg="写入成功") - except BaseException as e: - # Handle ExceptionGroup from TaskGroup (Python 3.11+) or BaseExceptionGroup - if hasattr(e, 'exceptions'): - error_messages = [f"{type(sub_e).__name__}: {str(sub_e)}" for sub_e in e.exceptions] - detailed_error = "; ".join(error_messages) - api_logger.error(f"Write operation error (TaskGroup): {detailed_error}", exc_info=True) - return fail(BizCode.INTERNAL_ERROR, "写入失败", detailed_error) - api_logger.error(f"Write operation error: {str(e)}", exc_info=True) - return fail(BizCode.INTERNAL_ERROR, "写入失败", str(e)) - - -@router.post("/writer_service_async", response_model=ApiResponse) -@cur_workspace_access_guard() -async def write_server_async( - user_input: Write_UserInput, - language_type: str = Header(default=None, alias="X-Language-Type"), - db: Session = Depends(get_db), - current_user: User = Depends(get_current_user) -): - """ - Async write service endpoint - enqueues write processing to Celery - - Args: - user_input: Write request containing message and end_user_id - language_type: 语言类型 ("zh" 中文, "en" 英文),通过 X-Language-Type Header 传递 - - Returns: - Task ID for tracking async operation - Use GET /memory/write_result/{task_id} to check task status and get result - """ - # 使用集中化的语言校验 - language = get_language_from_header(language_type) - - config_id = user_input.config_id - workspace_id = current_user.current_workspace_id - api_logger.info( - f"Async write service: workspace_id={workspace_id}, config_id={config_id}, language_type={language}") - - # 获取 storage_type,如果为 None 则使用默认值 - storage_type = workspace_service.get_workspace_storage_type( - db=db, - workspace_id=workspace_id, - user=current_user - ) - if storage_type is None: storage_type = 'neo4j' - user_rag_memory_id = '' - if workspace_id: - - knowledge = knowledge_repository.get_knowledge_by_name( - db=db, - name="USER_RAG_MERORY", - workspace_id=workspace_id - ) - if knowledge: user_rag_memory_id = str(knowledge.id) - api_logger.info(f"Async write: storage_type={storage_type}, user_rag_memory_id={user_rag_memory_id}") - try: - # 获取标准化的消息列表 - messages_list = memory_agent_service.get_messages_list(user_input) - - task = celery_app.send_task( - "app.core.memory.agent.write_message", - args=[user_input.end_user_id, messages_list, config_id, storage_type, user_rag_memory_id, language] - ) - api_logger.info(f"Write task queued: {task.id}") - - return success(data={"task_id": task.id}, msg="写入任务已提交") - except Exception as e: - api_logger.error(f"Async write operation failed: {str(e)}") - return fail(BizCode.INTERNAL_ERROR, "写入失败", str(e)) +# @router.post("/writer_service", response_model=ApiResponse) +# @cur_workspace_access_guard() +# async def write_server( +# user_input: Write_UserInput, +# language_type: str = Header(default=None, alias="X-Language-Type"), +# db: Session = Depends(get_db), +# current_user: User = Depends(get_current_user) +# ): +# """ +# Write service endpoint - processes write operations synchronously +# +# Args: +# user_input: Write request containing message and end_user_id +# language_type: 语言类型 ("zh" 中文, "en" 英文),通过 X-Language-Type Header 传递 +# +# Returns: +# Response with write operation status +# """ +# # 使用集中化的语言校验 +# language = get_language_from_header(language_type) +# +# config_id = user_input.config_id +# workspace_id = current_user.current_workspace_id +# api_logger.info(f"Write service: workspace_id={workspace_id}, config_id={config_id}, language_type={language}") +# +# # 获取 storage_type,如果为 None 则使用默认值 +# storage_type = workspace_service.get_workspace_storage_type( +# db=db, +# workspace_id=workspace_id, +# user=current_user +# ) +# if storage_type is None: storage_type = 'neo4j' +# user_rag_memory_id = '' +# +# # 如果 storage_type 是 rag,必须确保有有效的 user_rag_memory_id +# if storage_type == 'rag': +# if workspace_id: +# knowledge = knowledge_repository.get_knowledge_by_name( +# db=db, +# name="USER_RAG_MERORY", +# workspace_id=workspace_id +# ) +# if knowledge: +# user_rag_memory_id = str(knowledge.id) +# else: +# api_logger.warning( +# f"未找到名为 'USER_RAG_MERORY' 的知识库,workspace_id: {workspace_id},将使用 neo4j 存储") +# storage_type = 'neo4j' +# else: +# api_logger.warning("workspace_id 为空,无法使用 rag 存储,将使用 neo4j 存储") +# storage_type = 'neo4j' +# +# api_logger.info( +# f"Write service requested for group {user_input.end_user_id}, storage_type: {storage_type}, user_rag_memory_id: {user_rag_memory_id}") +# try: +# messages_list = memory_agent_service.get_messages_list(user_input) +# result = await memory_agent_service.write_memory( +# user_input.end_user_id, +# messages_list, +# config_id, +# db, +# storage_type, +# user_rag_memory_id, +# language +# ) +# +# return success(data=result, msg="写入成功") +# except BaseException as e: +# # Handle ExceptionGroup from TaskGroup (Python 3.11+) or BaseExceptionGroup +# if hasattr(e, 'exceptions'): +# error_messages = [f"{type(sub_e).__name__}: {str(sub_e)}" for sub_e in e.exceptions] +# detailed_error = "; ".join(error_messages) +# api_logger.error(f"Write operation error (TaskGroup): {detailed_error}", exc_info=True) +# return fail(BizCode.INTERNAL_ERROR, "写入失败", detailed_error) +# api_logger.error(f"Write operation error: {str(e)}", exc_info=True) +# return fail(BizCode.INTERNAL_ERROR, "写入失败", str(e)) +# +# +# @router.post("/writer_service_async", response_model=ApiResponse) +# @cur_workspace_access_guard() +# async def write_server_async( +# user_input: Write_UserInput, +# language_type: str = Header(default=None, alias="X-Language-Type"), +# db: Session = Depends(get_db), +# current_user: User = Depends(get_current_user) +# ): +# """ +# Async write service endpoint - enqueues write processing to Celery +# +# Args: +# user_input: Write request containing message and end_user_id +# language_type: 语言类型 ("zh" 中文, "en" 英文),通过 X-Language-Type Header 传递 +# +# Returns: +# Task ID for tracking async operation +# Use GET /memory/write_result/{task_id} to check task status and get result +# """ +# # 使用集中化的语言校验 +# language = get_language_from_header(language_type) +# +# config_id = user_input.config_id +# workspace_id = current_user.current_workspace_id +# api_logger.info( +# f"Async write service: workspace_id={workspace_id}, config_id={config_id}, language_type={language}") +# +# # 获取 storage_type,如果为 None 则使用默认值 +# storage_type = workspace_service.get_workspace_storage_type( +# db=db, +# workspace_id=workspace_id, +# user=current_user +# ) +# if storage_type is None: storage_type = 'neo4j' +# user_rag_memory_id = '' +# if workspace_id: +# +# knowledge = knowledge_repository.get_knowledge_by_name( +# db=db, +# name="USER_RAG_MERORY", +# workspace_id=workspace_id +# ) +# if knowledge: user_rag_memory_id = str(knowledge.id) +# api_logger.info(f"Async write: storage_type={storage_type}, user_rag_memory_id={user_rag_memory_id}") +# try: +# # 获取标准化的消息列表 +# messages_list = memory_agent_service.get_messages_list(user_input) +# +# task = celery_app.send_task( +# "app.core.memory.agent.write_message", +# args=[user_input.end_user_id, messages_list, config_id, storage_type, user_rag_memory_id, language] +# ) +# api_logger.info(f"Write task queued: {task.id}") +# +# return success(data={"task_id": task.id}, msg="写入任务已提交") +# except Exception as e: +# api_logger.error(f"Async write operation failed: {str(e)}") +# return fail(BizCode.INTERNAL_ERROR, "写入失败", str(e)) @router.post("/read_service", response_model=ApiResponse) diff --git a/api/app/controllers/memory_dashboard_controller.py b/api/app/controllers/memory_dashboard_controller.py index cc0efab3..fe4337d1 100644 --- a/api/app/controllers/memory_dashboard_controller.py +++ b/api/app/controllers/memory_dashboard_controller.py @@ -663,9 +663,12 @@ async def dashboard_data( rag_data["total_memory"] = total_chunk # total_app: 统计当前空间下的所有app数量 - from app.repositories import app_repository - apps_orm = app_repository.get_apps_by_workspace_id(db, workspace_id) - rag_data["total_app"] = len(apps_orm) + # 包含自有app + 被分享给本工作空间的app + from app.services import app_service as _app_svc + _, total_app = _app_svc.AppService(db).list_apps( + workspace_id=workspace_id, include_shared=True, pagesize=1 + ) + rag_data["total_app"] = total_app # total_knowledge: 使用 total_kb(总知识库数) total_kb = memory_dashboard_service.get_rag_total_kb(db, current_user) @@ -687,7 +690,7 @@ async def dashboard_data( api_logger.warning(f"获取RAG模式API调用统计失败,使用默认值: {str(e)}") rag_data["total_api_call"] = 0 - api_logger.info(f"成功获取RAG相关数据: memory={total_chunk}, app={len(apps_orm)}, knowledge={total_kb}, api_calls={rag_data['total_api_call']}") + api_logger.info(f"成功获取RAG相关数据: memory={total_chunk}, app={total_app}, knowledge={total_kb}, api_calls={rag_data['total_api_call']}") except Exception as e: api_logger.warning(f"获取RAG相关数据失败: {str(e)}") diff --git a/api/app/controllers/memory_forget_controller.py b/api/app/controllers/memory_forget_controller.py index 2b5ef72f..51ce92b3 100644 --- a/api/app/controllers/memory_forget_controller.py +++ b/api/app/controllers/memory_forget_controller.py @@ -31,6 +31,7 @@ from app.schemas.memory_storage_schema import ( ForgettingCurveRequest, ForgettingCurveResponse, ForgettingCurvePoint, + PendingNodesResponse, ) from app.schemas.response_schema import ApiResponse from app.services.memory_forget_service import MemoryForgetService @@ -308,6 +309,100 @@ async def get_forgetting_stats( return fail(BizCode.INTERNAL_ERROR, "获取遗忘引擎统计失败", str(e)) +@router.get("/pending-nodes", response_model=ApiResponse) +async def get_pending_nodes( + end_user_id: str, + page: int = 1, + pagesize: int = 10, + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db) +): + """ + 获取待遗忘节点列表(独立分页接口) + + 查询满足遗忘条件的节点(激活值低于阈值且最后访问时间超过最小天数)。 + 此接口独立分页,与 /stats 接口分离。 + + Args: + end_user_id: 组ID(即 end_user_id,必填) + page: 页码(从1开始,默认1) + pagesize: 每页数量(默认10) + current_user: 当前用户 + db: 数据库会话 + + Returns: + ApiResponse: 包含待遗忘节点列表和分页信息的响应 + + Examples: + - 第1页,每页10条:GET /memory/forget-memory/pending-nodes?end_user_id=xxx&page=1&pagesize=10 + - 第2页,每页20条:GET /memory/forget-memory/pending-nodes?end_user_id=xxx&page=2&pagesize=20 + + Notes: + - page 从1开始,pagesize 必须大于0 + - 返回格式:{"items": [...], "page": {"page": 1, "pagesize": 10, "total": 100, "hasnext": true}} + """ + workspace_id = current_user.current_workspace_id + # 检查用户是否已选择工作空间 + if workspace_id is None: + api_logger.warning(f"用户 {current_user.username} 尝试获取待遗忘节点但未选择工作空间") + return fail(BizCode.INVALID_PARAMETER, "请先切换到一个工作空间", "current_workspace_id is None") + + # 验证 end_user_id 必填 + if not end_user_id: + api_logger.warning(f"用户 {current_user.username} 尝试获取待遗忘节点但未提供 end_user_id") + return fail(BizCode.INVALID_PARAMETER, "end_user_id 不能为空", "end_user_id is required") + + # 通过 end_user_id 获取关联的 config_id + try: + from app.services.memory_agent_service import get_end_user_connected_config + + connected_config = get_end_user_connected_config(end_user_id, db) + config_id = connected_config.get("memory_config_id") + config_id = resolve_config_id(config_id, db) + + if config_id is None: + api_logger.warning(f"终端用户 {end_user_id} 未关联记忆配置") + return fail(BizCode.INVALID_PARAMETER, f"终端用户 {end_user_id} 未关联记忆配置", "memory_config_id is None") + + api_logger.debug(f"通过 end_user_id={end_user_id} 获取到 config_id={config_id}") + except ValueError as e: + api_logger.warning(f"获取终端用户配置失败: {str(e)}") + return fail(BizCode.INVALID_PARAMETER, str(e), "ValueError") + except Exception as e: + api_logger.error(f"获取终端用户配置时发生错误: {str(e)}") + return fail(BizCode.INTERNAL_ERROR, "获取终端用户配置失败", str(e)) + + # 验证分页参数 + if page < 1: + return fail(BizCode.INVALID_PARAMETER, "page 必须大于等于1", "page < 1") + if pagesize < 1: + return fail(BizCode.INVALID_PARAMETER, "pagesize 必须大于等于1", "pagesize < 1") + + api_logger.info( + f"用户 {current_user.username} 在工作空间 {workspace_id} 请求获取待遗忘节点: " + f"end_user_id={end_user_id}, page={page}, pagesize={pagesize}" + ) + + try: + # 调用服务层获取待遗忘节点列表 + result = await forget_service.get_pending_nodes( + db=db, + end_user_id=end_user_id, + config_id=config_id, + page=page, + pagesize=pagesize + ) + + # 构建响应 + response_data = PendingNodesResponse(**result) + + return success(data=response_data.model_dump(), msg="查询成功") + + except Exception as e: + api_logger.error(f"获取待遗忘节点列表失败: {str(e)}") + return fail(BizCode.INTERNAL_ERROR, "获取待遗忘节点列表失败", str(e)) + + @router.post("/forgetting_curve", response_model=ApiResponse) async def get_forgetting_curve( request: ForgettingCurveRequest, diff --git a/api/app/controllers/memory_storage_controller.py b/api/app/controllers/memory_storage_controller.py index d91dfc36..d8b39325 100644 --- a/api/app/controllers/memory_storage_controller.py +++ b/api/app/controllers/memory_storage_controller.py @@ -54,8 +54,8 @@ router = APIRouter( @router.get("/info", response_model=ApiResponse) async def get_storage_info( - storage_id: str, - current_user: User = Depends(get_current_user) + storage_id: str, + current_user: User = Depends(get_current_user) ): """ Example wrapper endpoint - retrieves storage information @@ -75,24 +75,19 @@ async def get_storage_info( return fail(BizCode.INTERNAL_ERROR, "存储信息获取失败", str(e)) - - - - - -@router.post("/create_config", response_model=ApiResponse) # 创建配置文件,其他参数默认 +@router.post("/create_config", response_model=ApiResponse) # 创建配置文件,其他参数默认 def create_config( - payload: ConfigParamsCreate, - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), - x_language_type: Optional[str] = Header(None, alias="X-Language-Type"), + payload: ConfigParamsCreate, + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), + x_language_type: Optional[str] = Header(None, alias="X-Language-Type"), ) -> dict: workspace_id = current_user.current_workspace_id # 检查用户是否已选择工作空间 if workspace_id is None: api_logger.warning(f"用户 {current_user.username} 尝试创建配置但未选择工作空间") return fail(BizCode.INVALID_PARAMETER, "请先切换到一个工作空间", "current_workspace_id is None") - + api_logger.info(f"用户 {current_user.username} 在工作空间 {workspace_id} 请求创建配置: {payload.config_name}") try: # 将 workspace_id 注入到 payload 中(保持为 UUID 类型) @@ -107,9 +102,11 @@ def create_config( api_logger.warning(f"重复的配置名称 '{config_name}' 在工作空间 {workspace_id}") lang = get_language_from_header(x_language_type) if lang == "en": - msg = fail(BizCode.BAD_REQUEST, "Config name already exists", f"A config named \"{config_name}\" already exists in the current workspace. Please use a different name.") + msg = fail(BizCode.BAD_REQUEST, "Config name already exists", + f"A config named \"{config_name}\" already exists in the current workspace. Please use a different name.") else: - msg = fail(BizCode.BAD_REQUEST, "配置名称已存在", f"当前工作空间下已存在名为「{config_name}」的记忆配置,请使用其他名称") + msg = fail(BizCode.BAD_REQUEST, "配置名称已存在", + f"当前工作空间下已存在名为「{config_name}」的记忆配置,请使用其他名称") return JSONResponse(status_code=400, content=msg) api_logger.error(f"Create config failed: {err_str}") return fail(BizCode.INTERNAL_ERROR, "创建配置失败", err_str) @@ -119,9 +116,11 @@ def create_config( api_logger.warning(f"重复的配置名称 '{payload.config_name}' 在工作空间 {workspace_id}") lang = get_language_from_header(x_language_type) if lang == "en": - msg = fail(BizCode.BAD_REQUEST, "Config name already exists", f"A config named \"{payload.config_name}\" already exists in the current workspace. Please use a different name.") + msg = fail(BizCode.BAD_REQUEST, "Config name already exists", + f"A config named \"{payload.config_name}\" already exists in the current workspace. Please use a different name.") else: - msg = fail(BizCode.BAD_REQUEST, "配置名称已存在", f"当前工作空间下已存在名为「{payload.config_name}」的记忆配置,请使用其他名称") + msg = fail(BizCode.BAD_REQUEST, "配置名称已存在", + f"当前工作空间下已存在名为「{payload.config_name}」的记忆配置,请使用其他名称") return JSONResponse(status_code=400, content=msg) api_logger.error(f"Create config failed: {str(e)}") return fail(BizCode.INTERNAL_ERROR, "创建配置失败", str(e)) @@ -129,10 +128,10 @@ def create_config( @router.delete("/delete_config", response_model=ApiResponse) # 删除数据库中的内容(按配置名称) def delete_config( - config_id: UUID|int, - force: bool = Query(False, description="是否强制删除(即使有终端用户正在使用)"), - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + config_id: UUID | int, + force: bool = Query(False, description="是否强制删除(即使有终端用户正在使用)"), + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> dict: """删除记忆配置(带终端用户保护) @@ -145,24 +144,24 @@ def delete_config( force: 设置为 true 可强制删除(即使有终端用户正在使用) """ workspace_id = current_user.current_workspace_id - config_id=resolve_config_id(config_id, db) + config_id = resolve_config_id(config_id, db) # 检查用户是否已选择工作空间 if workspace_id is None: api_logger.warning(f"用户 {current_user.username} 尝试删除配置但未选择工作空间") return fail(BizCode.INVALID_PARAMETER, "请先切换到一个工作空间", "current_workspace_id is None") - + api_logger.info( f"用户 {current_user.username} 在工作空间 {workspace_id} 请求删除配置: " f"config_id={config_id}, force={force}" ) - + try: # 使用带保护的删除服务 from app.services.memory_config_service import MemoryConfigService - + config_service = MemoryConfigService(db) result = config_service.delete_config(config_id=config_id, force=force) - + if result["status"] == "error": api_logger.warning( f"记忆配置删除被拒绝: config_id={config_id}, reason={result['message']}" @@ -172,7 +171,7 @@ def delete_config( msg=result["message"], data={"config_id": str(config_id), "is_default": result.get("is_default", False)} ) - + if result["status"] == "warning": api_logger.warning( f"记忆配置正在使用,无法删除: config_id={config_id}, " @@ -186,7 +185,7 @@ def delete_config( "force_required": result["force_required"] } ) - + api_logger.info( f"记忆配置删除成功: config_id={config_id}, " f"affected_users={result['affected_users']}" @@ -195,7 +194,7 @@ def delete_config( msg=result["message"], data={"affected_users": result["affected_users"]} ) - + except Exception as e: api_logger.error(f"Delete config failed: {str(e)}", exc_info=True) return fail(BizCode.INTERNAL_ERROR, "删除配置失败", str(e)) @@ -203,9 +202,9 @@ def delete_config( @router.post("/update_config", response_model=ApiResponse) # 更新配置文件中name和desc def update_config( - payload: ConfigUpdate, - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + payload: ConfigUpdate, + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> dict: workspace_id = current_user.current_workspace_id payload.config_id = resolve_config_id(payload.config_id, db) @@ -213,12 +212,13 @@ def update_config( if workspace_id is None: api_logger.warning(f"用户 {current_user.username} 尝试更新配置但未选择工作空间") return fail(BizCode.INVALID_PARAMETER, "请先切换到一个工作空间", "current_workspace_id is None") - + # 校验至少有一个字段需要更新 if payload.config_name is None and payload.config_desc is None and payload.scene_id is None: api_logger.warning(f"用户 {current_user.username} 尝试更新配置但未提供任何更新字段") - return fail(BizCode.INVALID_PARAMETER, "请至少提供一个需要更新的字段", "config_name, config_desc, scene_id 均为空") - + return fail(BizCode.INVALID_PARAMETER, "请至少提供一个需要更新的字段", + "config_name, config_desc, scene_id 均为空") + api_logger.info(f"用户 {current_user.username} 在工作空间 {workspace_id} 请求更新配置: {payload.config_id}") try: svc = DataConfigService(db) @@ -231,9 +231,9 @@ def update_config( @router.post("/update_config_extracted", response_model=ApiResponse) # 更新数据库中的部分内容 所有业务字段均可选 def update_config_extracted( - payload: ConfigUpdateExtracted, - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + payload: ConfigUpdateExtracted, + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> dict: workspace_id = current_user.current_workspace_id payload.config_id = resolve_config_id(payload.config_id, db) @@ -241,7 +241,7 @@ def update_config_extracted( if workspace_id is None: api_logger.warning(f"用户 {current_user.username} 尝试更新提取配置但未选择工作空间") return fail(BizCode.INVALID_PARAMETER, "请先切换到一个工作空间", "current_workspace_id is None") - + api_logger.info(f"用户 {current_user.username} 在工作空间 {workspace_id} 请求更新提取配置: {payload.config_id}") try: svc = DataConfigService(db) @@ -256,11 +256,11 @@ def update_config_extracted( # 遗忘引擎配置接口已迁移到 memory_forget_controller.py # 使用新接口: /api/memory/forget/read_config 和 /api/memory/forget/update_config -@router.get("/read_config_extracted", response_model=ApiResponse) # 通过查询参数读取某条配置(固定路径) 没有意义的话就删除 +@router.get("/read_config_extracted", response_model=ApiResponse) # 通过查询参数读取某条配置(固定路径) 没有意义的话就删除 def read_config_extracted( - config_id: UUID | int, - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + config_id: UUID | int, + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> dict: workspace_id = current_user.current_workspace_id config_id = resolve_config_id(config_id, db) @@ -268,7 +268,7 @@ def read_config_extracted( if workspace_id is None: api_logger.warning(f"用户 {current_user.username} 尝试读取提取配置但未选择工作空间") return fail(BizCode.INVALID_PARAMETER, "请先切换到一个工作空间", "current_workspace_id is None") - + api_logger.info(f"用户 {current_user.username} 在工作空间 {workspace_id} 请求读取提取配置: {config_id}") try: svc = DataConfigService(db) @@ -278,18 +278,19 @@ def read_config_extracted( api_logger.error(f"Read config extracted failed: {str(e)}") return fail(BizCode.INTERNAL_ERROR, "查询配置失败", str(e)) -@router.get("/read_all_config", response_model=ApiResponse) # 读取所有配置文件列表 + +@router.get("/read_all_config", response_model=ApiResponse) # 读取所有配置文件列表 def read_all_config( - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> dict: workspace_id = current_user.current_workspace_id - + # 检查用户是否已选择工作空间 if workspace_id is None: api_logger.warning(f"用户 {current_user.username} 尝试查询配置但未选择工作空间") return fail(BizCode.INVALID_PARAMETER, "请先切换到一个工作空间", "current_workspace_id is None") - + api_logger.info(f"用户 {current_user.username} 在工作空间 {workspace_id} 请求读取所有配置") try: svc = DataConfigService(db) @@ -303,14 +304,14 @@ def read_all_config( @router.post("/pilot_run", response_model=None) async def pilot_run( - payload: ConfigPilotRun, - language_type: str = Header(default=None, alias="X-Language-Type"), - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + payload: ConfigPilotRun, + language_type: str = Header(default=None, alias="X-Language-Type"), + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> StreamingResponse: # 使用集中化的语言校验 language = get_language_from_header(language_type) - + api_logger.info( f"Pilot run requested: config_id={payload.config_id}, " f"dialogue_text_length={len(payload.dialogue_text)}, " @@ -333,9 +334,9 @@ async def pilot_run( @router.get("/search/kb_type_distribution", response_model=ApiResponse) async def get_kb_type_distribution( - end_user_id: Optional[str] = None, - current_user: User = Depends(get_current_user), - ) -> dict: + end_user_id: Optional[str] = None, + current_user: User = Depends(get_current_user), +) -> dict: api_logger.info(f"KB type distribution requested for end_user_id: {end_user_id}") try: result = await kb_type_distribution(end_user_id) @@ -344,12 +345,12 @@ async def get_kb_type_distribution( api_logger.error(f"KB type distribution failed: {str(e)}") return fail(BizCode.INTERNAL_ERROR, "知识库类型分布查询失败", str(e)) - + @router.get("/search/dialogue", response_model=ApiResponse) async def search_dialogues_num( - end_user_id: Optional[str] = None, - current_user: User = Depends(get_current_user), - ) -> dict: + end_user_id: Optional[str] = None, + current_user: User = Depends(get_current_user), +) -> dict: api_logger.info(f"Search dialogue requested for end_user_id: {end_user_id}") try: result = await search_dialogue(end_user_id) @@ -361,9 +362,9 @@ async def search_dialogues_num( @router.get("/search/chunk", response_model=ApiResponse) async def search_chunks_num( - end_user_id: Optional[str] = None, - current_user: User = Depends(get_current_user), - ) -> dict: + end_user_id: Optional[str] = None, + current_user: User = Depends(get_current_user), +) -> dict: api_logger.info(f"Search chunk requested for end_user_id: {end_user_id}") try: result = await search_chunk(end_user_id) @@ -375,9 +376,9 @@ async def search_chunks_num( @router.get("/search/statement", response_model=ApiResponse) async def search_statements_num( - end_user_id: Optional[str] = None, - current_user: User = Depends(get_current_user), - ) -> dict: + end_user_id: Optional[str] = None, + current_user: User = Depends(get_current_user), +) -> dict: api_logger.info(f"Search statement requested for end_user_id: {end_user_id}") try: result = await search_statement(end_user_id) @@ -389,9 +390,9 @@ async def search_statements_num( @router.get("/search/entity", response_model=ApiResponse) async def search_entities_num( - end_user_id: Optional[str] = None, - current_user: User = Depends(get_current_user), - ) -> dict: + end_user_id: Optional[str] = None, + current_user: User = Depends(get_current_user), +) -> dict: api_logger.info(f"Search entity requested for end_user_id: {end_user_id}") try: result = await search_entity(end_user_id) @@ -403,9 +404,9 @@ async def search_entities_num( @router.get("/search", response_model=ApiResponse) async def search_all_num( - end_user_id: Optional[str] = None, - current_user: User = Depends(get_current_user), - ) -> dict: + end_user_id: Optional[str] = None, + current_user: User = Depends(get_current_user), +) -> dict: api_logger.info(f"Search all requested for end_user_id: {end_user_id}") try: result = await search_all(end_user_id) @@ -417,9 +418,9 @@ async def search_all_num( @router.get("/search/detials", response_model=ApiResponse) async def search_entities_detials( - end_user_id: Optional[str] = None, - current_user: User = Depends(get_current_user), - ) -> dict: + end_user_id: Optional[str] = None, + current_user: User = Depends(get_current_user), +) -> dict: api_logger.info(f"Search details requested for end_user_id: {end_user_id}") try: result = await search_detials(end_user_id) @@ -431,9 +432,9 @@ async def search_entities_detials( @router.get("/search/edges", response_model=ApiResponse) async def search_entity_edges( - end_user_id: Optional[str] = None, - current_user: User = Depends(get_current_user), - ) -> dict: + end_user_id: Optional[str] = None, + current_user: User = Depends(get_current_user), +) -> dict: api_logger.info(f"Search edges requested for end_user_id: {end_user_id}") try: result = await search_edges(end_user_id) @@ -443,14 +444,12 @@ async def search_entity_edges( return fail(BizCode.INTERNAL_ERROR, "边查询失败", str(e)) - - @router.get("/analytics/hot_memory_tags", response_model=ApiResponse) async def get_hot_memory_tags_api( - limit: int = 10, - db: Session = Depends(get_db), - current_user: User = Depends(get_current_user), - ) -> dict: + limit: int = 10, + db: Session = Depends(get_db), + current_user: User = Depends(get_current_user), +) -> dict: """ 获取热门记忆标签(带Redis缓存) @@ -461,18 +460,18 @@ async def get_hot_memory_tags_api( - 缓存未命中:~600-800ms(取决于LLM速度) """ workspace_id = current_user.current_workspace_id - + # 构建缓存键 cache_key = f"hot_memory_tags:{workspace_id}:{limit}" - + api_logger.info(f"Hot memory tags requested for workspace: {workspace_id}, limit: {limit}") - + try: # 尝试从Redis缓存获取 import json from app.aioRedis import aio_redis_get, aio_redis_set - + cached_result = await aio_redis_get(cache_key) if cached_result: api_logger.info(f"Cache hit for key: {cache_key}") @@ -481,11 +480,11 @@ async def get_hot_memory_tags_api( return success(data=data, msg="查询成功(缓存)") except json.JSONDecodeError: api_logger.warning(f"Failed to parse cached data, will refresh") - + # 缓存未命中,执行查询 api_logger.info(f"Cache miss for key: {cache_key}, executing query") result = await analytics_hot_memory_tags(db, current_user, limit) - + # 写入缓存(过期时间:5分钟) # 注意:result是列表,需要转换为JSON字符串 try: @@ -495,9 +494,9 @@ async def get_hot_memory_tags_api( except Exception as cache_error: # 缓存写入失败不影响主流程 api_logger.warning(f"Failed to cache result: {str(cache_error)}") - + return success(data=result, msg="查询成功") - + except Exception as e: api_logger.error(f"Hot memory tags failed: {str(e)}") return fail(BizCode.INTERNAL_ERROR, "热门标签查询失败", str(e)) @@ -505,8 +504,8 @@ async def get_hot_memory_tags_api( @router.delete("/analytics/hot_memory_tags/cache", response_model=ApiResponse) async def clear_hot_memory_tags_cache( - current_user: User = Depends(get_current_user), - ) -> dict: + current_user: User = Depends(get_current_user), +) -> dict: """ 清除热门标签缓存 @@ -516,12 +515,12 @@ async def clear_hot_memory_tags_cache( - 数据更新后立即生效 """ workspace_id = current_user.current_workspace_id - + api_logger.info(f"Clear hot memory tags cache requested for workspace: {workspace_id}") - + try: from app.aioRedis import aio_redis_delete - + # 清除所有limit的缓存(常见的limit值) cleared_count = 0 for limit in [5, 10, 15, 20, 30, 50]: @@ -530,12 +529,12 @@ async def clear_hot_memory_tags_cache( if result: cleared_count += 1 api_logger.info(f"Cleared cache for key: {cache_key}") - + return success( - data={"cleared_count": cleared_count}, + data={"cleared_count": cleared_count}, msg=f"成功清除 {cleared_count} 个缓存" ) - + except Exception as e: api_logger.error(f"Clear cache failed: {str(e)}") return fail(BizCode.INTERNAL_ERROR, "清除缓存失败", str(e)) @@ -543,7 +542,7 @@ async def clear_hot_memory_tags_cache( @router.get("/analytics/recent_activity_stats", response_model=ApiResponse) async def get_recent_activity_stats_api( - current_user: User = Depends(get_current_user), + current_user: User = Depends(get_current_user), ) -> dict: workspace_id = str(current_user.current_workspace_id) if current_user.current_workspace_id else None api_logger.info(f"Recent activity stats requested: workspace_id={workspace_id}") @@ -553,4 +552,3 @@ async def get_recent_activity_stats_api( except Exception as e: api_logger.error(f"Recent activity stats failed: {str(e)}") return fail(BizCode.INTERNAL_ERROR, "最近活动统计失败", str(e)) - diff --git a/api/app/controllers/model_controller.py b/api/app/controllers/model_controller.py index 6204a745..71fd41ad 100644 --- a/api/app/controllers/model_controller.py +++ b/api/app/controllers/model_controller.py @@ -42,6 +42,7 @@ def get_model_strategies(): @router.get("", response_model=ApiResponse) def get_model_list( type: Optional[list[str]] = Query(None, description="模型类型筛选(支持多个,如 ?type=LLM 或 ?type=LLM,EMBEDDING)"), + capability: Optional[list[str]] = Query(None, description="能力筛选(支持多个,如 ?capability=chat 或 ?capability=chat, embedding)"), provider: Optional[model_schema.ModelProvider] = Query(None, description="提供商筛选(基于API Key)"), is_active: Optional[bool] = Query(None, description="激活状态筛选"), is_public: Optional[bool] = Query(None, description="公开状态筛选"), @@ -74,10 +75,21 @@ def get_model_list( unique_flat_type = list(dict.fromkeys(flat_type)) type_list = [ModelType(t.lower()) for t in unique_flat_type] + capability_list = [] + if capability is not None: + flat_capability = [] + for item in capability: + split_items = [c.strip() for c in item.split(', ') if c.strip()] + flat_capability.extend(split_items) + + unique_flat_capability = list(dict.fromkeys(flat_capability)) + capability_list = unique_flat_capability + api_logger.error(f"获取模型type_list: {type_list}") query = model_schema.ModelConfigQuery( type=type_list, provider=provider, + capability=capability_list, is_active=is_active, is_public=is_public, search=search, diff --git a/api/app/controllers/public_share_controller.py b/api/app/controllers/public_share_controller.py index 33d7b60c..fc2916ed 100644 --- a/api/app/controllers/public_share_controller.py +++ b/api/app/controllers/public_share_controller.py @@ -27,6 +27,7 @@ from app.services.conversation_service import ConversationService from app.services.release_share_service import ReleaseShareService from app.services.shared_chat_service import SharedChatService from app.services.workflow_service import WorkflowService +from app.models.file_metadata_model import FileMetadata from app.utils.app_config_utils import workflow_config_4_app_release, \ agent_config_4_app_release, multi_agent_config_4_app_release @@ -259,8 +260,41 @@ def get_conversation( conv_service = ConversationService(db) messages = conv_service.get_messages(conversation_id) - # 构建响应 - conv_dict = conversation_schema.Conversation.model_validate(conversation).model_dump() + file_ids = [] + message_file_id_map = {} + + # 第一次遍历:解析 audio_url,收集所有有效的 file_id + for idx, m in enumerate(messages): + if m.role == "assistant" and m.meta_data: + audio_url = m.meta_data.get("audio_url") + if not audio_url: + continue + try: + file_id = uuid.UUID(audio_url.rstrip("/").split("/")[-1]) + except (ValueError, IndexError): + # audio_url 无法解析为 UUID,标记为 unknown + m.meta_data["audio_status"] = "unknown" + continue + + file_ids.append(file_id) + message_file_id_map[idx] = file_id + + # 批量查询所有相关的 FileMetadata + file_status_map = {} + if file_ids: + file_metas = ( + db.query(FileMetadata) + .filter(FileMetadata.id.in_(set(file_ids))) + .all() + ) + file_status_map = {fm.id: fm.status for fm in file_metas} + + # 第二次遍历:将查询结果映射回消息 + for idx, file_id in message_file_id_map.items(): + m = messages[idx] + m.meta_data["audio_status"] = file_status_map.get(file_id, "unknown") + + conv_dict = conversation_schema.Conversation.model_validate(conversation).model_dump(mode="json") conv_dict["messages"] = [ conversation_schema.Message.model_validate(m) for m in messages ] @@ -320,6 +354,16 @@ async def chat( other_id=other_id, original_user_id=user_id ) + + # Only extract and set memory_config_id when the end user doesn't have one yet + if not new_end_user.memory_config_id: + from app.services.memory_config_service import MemoryConfigService + memory_config_service = MemoryConfigService(db) + memory_config_id, _ = memory_config_service.extract_memory_config_id(release.type, release.config or {}) + if memory_config_id: + new_end_user.memory_config_id = memory_config_id + db.commit() + db.refresh(new_end_user) end_user_id = str(new_end_user.id) # appid = share.app_id @@ -669,6 +713,7 @@ async def config_query( content = { "app_type": release.app.type, "variables": release.config.get("variables"), + "memory": release.config.get("memory", {}).get("enabled"), "features": release.config.get("features") } elif release.app.type == AppType.MULTI_AGENT: diff --git a/api/app/controllers/service/app_api_controller.py b/api/app/controllers/service/app_api_controller.py index 32a911f9..d4573464 100644 --- a/api/app/controllers/service/app_api_controller.py +++ b/api/app/controllers/service/app_api_controller.py @@ -91,7 +91,7 @@ async def chat( app = app_service.get_app(api_key_auth.resource_id, api_key_auth.workspace_id) other_id = payload.user_id - workspace_id = app.workspace_id + workspace_id = api_key_auth.workspace_id end_user_repo = EndUserRepository(db) new_end_user = end_user_repo.get_or_create_end_user( app_id=app.id, diff --git a/api/app/controllers/service/memory_api_controller.py b/api/app/controllers/service/memory_api_controller.py index 34489e8a..08a94a89 100644 --- a/api/app/controllers/service/memory_api_controller.py +++ b/api/app/controllers/service/memory_api_controller.py @@ -6,6 +6,7 @@ from app.core.response_utils import success from app.db import get_db from app.schemas.api_key_schema import ApiKeyAuth from app.schemas.memory_api_schema import ( + ListConfigsResponse, MemoryReadRequest, MemoryReadResponse, MemoryWriteRequest, @@ -31,14 +32,15 @@ async def write_memory_api_service( request: Request, api_key_auth: ApiKeyAuth = None, db: Session = Depends(get_db), - payload: MemoryWriteRequest = Body(..., embed=False), - + message: str = Body(..., description="Message content"), ): """ Write memory to storage. Stores memory content for the specified end user using the Memory API Service. """ + body = await request.json() + payload = MemoryWriteRequest(**body) logger.info(f"Memory write request - end_user_id: {payload.end_user_id}, workspace_id: {api_key_auth.workspace_id}") memory_api_service = MemoryAPIService(db) @@ -62,13 +64,15 @@ async def read_memory_api_service( request: Request, api_key_auth: ApiKeyAuth = None, db: Session = Depends(get_db), - payload: MemoryReadRequest = Body(..., embed=False), + message: str = Body(..., description="Query message"), ): """ Read memory from storage. Queries and retrieves memories for the specified end user with context-aware responses. """ + body = await request.json() + payload = MemoryReadRequest(**body) logger.info(f"Memory read request - end_user_id: {payload.end_user_id}") memory_api_service = MemoryAPIService(db) @@ -85,3 +89,27 @@ async def read_memory_api_service( logger.info(f"Memory read successful for end_user: {payload.end_user_id}") return success(data=MemoryReadResponse(**result).model_dump(), msg="Memory read successfully") + + +@router.get("/configs") +@require_api_key(scopes=["memory"]) +async def list_memory_configs( + request: Request, + api_key_auth: ApiKeyAuth = None, + db: Session = Depends(get_db), +): + """ + List all memory configs for the workspace. + + Returns all available memory configurations associated with the authorized workspace. + """ + logger.info(f"List configs request - workspace_id: {api_key_auth.workspace_id}") + + memory_api_service = MemoryAPIService(db) + + result = memory_api_service.list_memory_configs( + workspace_id=api_key_auth.workspace_id, + ) + + logger.info(f"Listed {result['total']} configs for workspace: {api_key_auth.workspace_id}") + return success(data=ListConfigsResponse(**result).model_dump(), msg="Configs listed successfully") diff --git a/api/app/controllers/user_controller.py b/api/app/controllers/user_controller.py index 16213690..cc16a6b4 100644 --- a/api/app/controllers/user_controller.py +++ b/api/app/controllers/user_controller.py @@ -111,6 +111,18 @@ def get_current_user_info( break api_logger.info(f"当前用户信息获取成功: {result.username}, 角色: {result_schema.role}, 工作空间: {result_schema.current_workspace_name}") + + # 设置权限:如果用户来自 SSO Source,则使用该 Source 的 permissions;否则返回 "all" 表示拥有所有权限 + if current_user.external_source: + from premium.sso.models import SSOSource + source = db.query(SSOSource).filter(SSOSource.source_code == current_user.external_source).first() + if source and source.permissions: + result_schema.permissions = source.permissions + else: + result_schema.permissions = [] + else: + result_schema.permissions = ["all"] + return success(data=result_schema, msg=t("users.info.get_success")) @@ -135,7 +147,6 @@ def get_tenant_superusers( return success(data=superusers_schema, msg=t("users.list.superusers_success")) - @router.get("/{user_id}", response_model=ApiResponse) def get_user_info_by_id( user_id: uuid.UUID, diff --git a/api/app/controllers/user_memory_controllers.py b/api/app/controllers/user_memory_controllers.py index be796ff9..10b396a7 100644 --- a/api/app/controllers/user_memory_controllers.py +++ b/api/app/controllers/user_memory_controllers.py @@ -5,7 +5,7 @@ from typing import Optional import datetime from sqlalchemy.orm import Session -from fastapi import APIRouter, Depends,Header +from fastapi import APIRouter, Depends, Header from app.db import get_db from app.core.language_utils import get_language_from_header @@ -19,13 +19,15 @@ from app.services.user_memory_service import ( analytics_graph_data, analytics_community_graph_data, ) -from app.services.memory_entity_relationship_service import MemoryEntityService,MemoryEmotion,MemoryInteraction +from app.services.memory_entity_relationship_service import MemoryEntityService, MemoryEmotion, MemoryInteraction from app.schemas.response_schema import ApiResponse from app.schemas.memory_storage_schema import GenerateCacheRequest from app.repositories.workspace_repository import WorkspaceRepository -from app.schemas.end_user_schema import ( - EndUserProfileResponse, - EndUserProfileUpdate, +from app.repositories.end_user_repository import EndUserRepository +from app.schemas.end_user_info_schema import ( + EndUserInfoResponse, + EndUserInfoCreate, + EndUserInfoUpdate, ) from app.models.end_user_model import EndUser from app.dependencies import get_current_user @@ -45,9 +47,9 @@ router = APIRouter( @router.get("/analytics/memory_insight/report", response_model=ApiResponse) async def get_memory_insight_report_api( - end_user_id: str, - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + end_user_id: str, + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> dict: """ 获取缓存的记忆洞察报告 @@ -73,10 +75,10 @@ async def get_memory_insight_report_api( @router.get("/analytics/user_summary", response_model=ApiResponse) async def get_user_summary_api( - end_user_id: str, - language_type: str = Header(default=None, alias="X-Language-Type"), - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + end_user_id: str, + language_type: str = Header(default=None, alias="X-Language-Type"), + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> dict: """ 获取缓存的用户摘要 @@ -90,7 +92,7 @@ async def get_user_summary_api( """ # 使用集中化的语言校验 language = get_language_from_header(language_type) - + workspace_id = current_user.current_workspace_id workspace_repo = WorkspaceRepository(db) workspace_models = workspace_repo.get_workspace_models_configs(workspace_id) @@ -102,7 +104,7 @@ async def get_user_summary_api( api_logger.info(f"用户摘要查询请求: end_user_id={end_user_id}, user={current_user.username}") try: # 调用服务层获取缓存数据 - result = await user_memory_service.get_cached_user_summary(db, end_user_id,model_id,language) + result = await user_memory_service.get_cached_user_summary(db, end_user_id, model_id, language) if result["is_cached"]: api_logger.info(f"成功返回缓存的用户摘要: end_user_id={end_user_id}") @@ -117,10 +119,10 @@ async def get_user_summary_api( @router.post("/analytics/generate_cache", response_model=ApiResponse) async def generate_cache_api( - request: GenerateCacheRequest, - language_type: str = Header(default=None, alias="X-Language-Type"), - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + request: GenerateCacheRequest, + language_type: str = Header(default=None, alias="X-Language-Type"), + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> dict: """ 手动触发缓存生成 @@ -134,7 +136,7 @@ async def generate_cache_api( """ # 使用集中化的语言校验 language = get_language_from_header(language_type) - + workspace_id = current_user.current_workspace_id # 检查用户是否已选择工作空间 @@ -155,10 +157,12 @@ async def generate_cache_api( api_logger.info(f"开始为单个用户生成缓存: end_user_id={end_user_id}") # 生成记忆洞察 - insight_result = await user_memory_service.generate_and_cache_insight(db, end_user_id, workspace_id, language=language) + insight_result = await user_memory_service.generate_and_cache_insight(db, end_user_id, workspace_id, + language=language) # 生成用户摘要 - summary_result = await user_memory_service.generate_and_cache_summary(db, end_user_id, workspace_id, language=language) + summary_result = await user_memory_service.generate_and_cache_summary(db, end_user_id, workspace_id, + language=language) # 构建响应 result = { @@ -209,9 +213,9 @@ async def generate_cache_api( @router.get("/analytics/node_statistics", response_model=ApiResponse) async def get_node_statistics_api( - end_user_id: str, - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + end_user_id: str, + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> dict: workspace_id = current_user.current_workspace_id @@ -220,7 +224,8 @@ async def get_node_statistics_api( api_logger.warning(f"用户 {current_user.username} 尝试查询节点统计但未选择工作空间") return fail(BizCode.INVALID_PARAMETER, "请先切换到一个工作空间", "current_workspace_id is None") - api_logger.info(f"记忆类型统计请求: end_user_id={end_user_id}, user={current_user.username}, workspace={workspace_id}") + api_logger.info( + f"记忆类型统计请求: end_user_id={end_user_id}, user={current_user.username}, workspace={workspace_id}") try: # 调用新的记忆类型统计函数 @@ -228,21 +233,23 @@ async def get_node_statistics_api( # 计算总数用于日志 total_count = sum(item["count"] for item in result) - api_logger.info(f"成功获取记忆类型统计: end_user_id={end_user_id}, 总记忆数={total_count}, 类型数={len(result)}") + api_logger.info( + f"成功获取记忆类型统计: end_user_id={end_user_id}, 总记忆数={total_count}, 类型数={len(result)}") return success(data=result, msg="查询成功") except Exception as e: api_logger.error(f"记忆类型查询失败: end_user_id={end_user_id}, error={str(e)}") return fail(BizCode.INTERNAL_ERROR, "记忆类型查询失败", str(e)) + @router.get("/analytics/graph_data", response_model=ApiResponse) async def get_graph_data_api( - end_user_id: str, - node_types: Optional[str] = None, - limit: int = 100, - depth: int = 1, - center_node_id: Optional[str] = None, - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + end_user_id: str, + node_types: Optional[str] = None, + limit: int = 100, + depth: int = 1, + center_node_id: Optional[str] = None, + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> dict: workspace_id = current_user.current_workspace_id @@ -298,9 +305,9 @@ async def get_graph_data_api( @router.get("/analytics/community_graph", response_model=ApiResponse) async def get_community_graph_data_api( - end_user_id: str, - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), + end_user_id: str, + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), ) -> dict: workspace_id = current_user.current_workspace_id @@ -331,111 +338,130 @@ async def get_community_graph_data_api( api_logger.error(f"社区图谱查询失败: end_user_id={end_user_id}, error={str(e)}") return fail(BizCode.INTERNAL_ERROR, "社区图谱查询失败", str(e)) +#=======================终端用户信息接口======================= -@router.get("/read_end_user/profile", response_model=ApiResponse) -async def get_end_user_profile( +@router.get("/end_user_info", response_model=ApiResponse) +async def get_end_user_info( end_user_id: str, current_user: User = Depends(get_current_user), db: Session = Depends(get_db), ) -> dict: - workspace_id = current_user.current_workspace_id - workspace_repo = WorkspaceRepository(db) - workspace_models = workspace_repo.get_workspace_models_configs(workspace_id) + """ + 查询终端用户信息记录 + + 根据 end_user_id 查询单条终端用户信息记录。 + """ + workspace_id = current_user.current_workspace_id - if workspace_models: - model_id = workspace_models.get("llm", None) - else: - model_id = None - # 检查用户是否已选择工作空间 if workspace_id is None: - api_logger.warning(f"用户 {current_user.username} 尝试查询用户信息但未选择工作空间") + api_logger.warning(f"用户 {current_user.username} 尝试查询终端用户信息但未选择工作空间") return fail(BizCode.INVALID_PARAMETER, "请先切换到一个工作空间", "current_workspace_id is None") api_logger.info( - f"用户信息查询请求: end_user_id={end_user_id}, user={current_user.username}, " + f"查询终端用户信息请求: end_user_id={end_user_id}, user={current_user.username}, " f"workspace={workspace_id}" ) - try: - # 查询终端用户 - end_user = db.query(EndUser).filter(EndUser.id == end_user_id).first() - - if not end_user: - api_logger.warning(f"终端用户不存在: end_user_id={end_user_id}") - return fail(BizCode.INVALID_PARAMETER, "终端用户不存在", f"end_user_id={end_user_id}") - # 构建响应数据 - profile_data = EndUserProfileResponse( - id=end_user.id, - other_name=end_user.other_name, - position=end_user.position, - department=end_user.department, - contact=end_user.contact, - phone=end_user.phone, - hire_date=end_user.hire_date, - updatetime_profile=end_user.updatetime_profile + # 校验 end_user 是否属于当前工作空间 + end_user_repo = EndUserRepository(db) + end_user = end_user_repo.get_end_user_by_id(end_user_id) + if end_user is None: + return fail(BizCode.USER_NOT_FOUND, "终端用户不存在", "end_user not found") + if str(end_user.workspace_id) != str(workspace_id): + api_logger.warning( + f"用户 {current_user.username} 尝试查询不属于工作空间 {workspace_id} 的终端用户 {end_user_id}" ) + return fail(BizCode.PERMISSION_DENIED, "该终端用户不属于当前工作空间", "end_user workspace mismatch") - api_logger.info(f"成功获取用户信息: end_user_id={end_user_id}") - return success(data=UserMemoryService.convert_profile_to_dict_with_timestamp(profile_data), msg="查询成功") + result = user_memory_service.get_end_user_info(db, end_user_id) - except Exception as e: - api_logger.error(f"用户信息查询失败: end_user_id={end_user_id}, error={str(e)}") - return fail(BizCode.INTERNAL_ERROR, "用户信息查询失败", str(e)) + if result["success"]: + api_logger.info(f"成功查询终端用户信息: end_user_id={end_user_id}") + return success(data=result["data"], msg="查询成功") + else: + error_msg = result["error"] + api_logger.error(f"查询终端用户信息失败: end_user_id={end_user_id}, error={error_msg}") + + if error_msg == "终端用户信息记录不存在": + return fail(BizCode.USER_NOT_FOUND, "终端用户信息记录不存在", error_msg) + elif error_msg == "无效的终端用户ID格式": + return fail(BizCode.INVALID_USER_ID, "无效的终端用户ID格式", error_msg) + else: + return fail(BizCode.INTERNAL_ERROR, "查询终端用户信息失败", error_msg) -@router.post("/updated_end_user/profile", response_model=ApiResponse) -async def update_end_user_profile( - profile_update: EndUserProfileUpdate, +@router.post("/end_user_info/updated", response_model=ApiResponse) +async def update_end_user_info( + info_update: EndUserInfoUpdate, current_user: User = Depends(get_current_user), db: Session = Depends(get_db), ) -> dict: """ - 更新终端用户的基本信息 + 更新终端用户信息记录 - 该接口可以更新用户的姓名、职位、部门、联系方式、电话和入职日期等信息。 - 所有字段都是可选的,只更新提供的字段。 + 根据 end_user_id 更新终端用户信息记录,支持批量更新多个别名。 + + 示例请求体: + { + "end_user_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890", + "other_name": "张三1", + "aliases": ["小张", "张工"], + "meta_data": {"position": "工程师", "department": "技术部"} + } """ workspace_id = current_user.current_workspace_id - end_user_id = profile_update.end_user_id + end_user_id = info_update.end_user_id - # 验证工作空间 if workspace_id is None: - api_logger.warning(f"用户 {current_user.username} 尝试更新用户信息但未选择工作空间") + api_logger.warning(f"用户 {current_user.username} 尝试更新终端用户信息但未选择工作空间") return fail(BizCode.INVALID_PARAMETER, "请先切换到一个工作空间", "current_workspace_id is None") api_logger.info( - f"用户信息更新请求: end_user_id={end_user_id}, user={current_user.username}, " + f"更新终端用户信息请求: end_user_id={end_user_id}, user={current_user.username}, " f"workspace={workspace_id}" ) - # 调用 Service 层处理业务逻辑 - result = user_memory_service.update_end_user_profile(db, end_user_id, profile_update) + # 校验 end_user 是否属于当前工作空间 + end_user_repo = EndUserRepository(db) + end_user = end_user_repo.get_end_user_by_id(end_user_id) + if end_user is None: + return fail(BizCode.USER_NOT_FOUND, "终端用户不存在", "end_user not found") + if str(end_user.workspace_id) != str(workspace_id): + api_logger.warning( + f"用户 {current_user.username} 尝试更新不属于工作空间 {workspace_id} 的终端用户 {end_user_id}" + ) + return fail(BizCode.PERMISSION_DENIED, "该终端用户不属于当前工作空间", "end_user workspace mismatch") + + # 获取更新数据(排除 end_user_id) + update_data = info_update.model_dump(exclude_unset=True, exclude={'end_user_id'}) + + result = user_memory_service.update_end_user_info(db, end_user_id, update_data) if result["success"]: - api_logger.info(f"成功更新用户信息: end_user_id={end_user_id}") + api_logger.info(f"成功更新终端用户信息: end_user_id={end_user_id}") return success(data=result["data"], msg="更新成功") else: error_msg = result["error"] - api_logger.error(f"用户信息更新失败: end_user_id={end_user_id}, error={error_msg}") + api_logger.error(f"终端用户信息更新失败: end_user_id={end_user_id}, error={error_msg}") - # 根据错误类型映射到合适的业务错误码 - if error_msg == "终端用户不存在": - return fail(BizCode.USER_NOT_FOUND, "终端用户不存在", error_msg) - elif error_msg == "无效的用户ID格式": - return fail(BizCode.INVALID_USER_ID, "无效的用户ID格式", error_msg) + if error_msg == "终端用户信息记录不存在": + return fail(BizCode.USER_NOT_FOUND, "终端用户信息记录不存在", error_msg) + elif error_msg == "无效的终端用户ID格式": + return fail(BizCode.INVALID_USER_ID, "无效的终端用户ID格式", error_msg) else: - # 只有未预期的错误才使用 INTERNAL_ERROR - return fail(BizCode.INTERNAL_ERROR, "用户信息更新失败", error_msg) + return fail(BizCode.INTERNAL_ERROR, "终端用户信息更新失败", error_msg) @router.get("/memory_space/timeline_memories", response_model=ApiResponse) -async def memory_space_timeline_of_shared_memories(id: str, label: str,language_type: str = Header(default=None, alias="X-Language-Type"), - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), - ): +async def memory_space_timeline_of_shared_memories( + id: str, label: str, + language_type: str = Header(default=None, alias="X-Language-Type"), + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), +): # 使用集中化的语言校验 language = get_language_from_header(language_type) - - workspace_id=current_user.current_workspace_id + + workspace_id = current_user.current_workspace_id workspace_repo = WorkspaceRepository(db) workspace_models = workspace_repo.get_workspace_models_configs(workspace_id) @@ -447,11 +473,13 @@ async def memory_space_timeline_of_shared_memories(id: str, label: str,language_ timeline_memories_result = await MemoryEntity.get_timeline_memories_server(model_id, language) return success(data=timeline_memories_result, msg="共同记忆时间线") + + @router.get("/memory_space/relationship_evolution", response_model=ApiResponse) async def memory_space_relationship_evolution(id: str, label: str, - current_user: User = Depends(get_current_user), - db: Session = Depends(get_db), - ): + current_user: User = Depends(get_current_user), + db: Session = Depends(get_db), + ): try: api_logger.info(f"关系演变查询请求: id={id}, table={label}, user={current_user.username}") diff --git a/api/app/core/agent/langchain_agent.py b/api/app/core/agent/langchain_agent.py index 88b6371c..7314ab5f 100644 --- a/api/app/core/agent/langchain_agent.py +++ b/api/app/core/agent/langchain_agent.py @@ -329,7 +329,6 @@ class LangChainAgent: db.close() except Exception as e: logger.warning(f"Failed to get db session: {e}") - actual_end_user_id = end_user_id if end_user_id is not None else "unknown" logger.info(f'写入类型{storage_type, str(end_user_id), message, str(user_rag_memory_id)}') print(f'写入类型{storage_type, str(end_user_id), message, str(user_rag_memory_id)}') try: @@ -598,8 +597,10 @@ class LangChainAgent: for msg in reversed(output_messages): if isinstance(msg, AIMessage): response_meta = msg.response_metadata if hasattr(msg, 'response_metadata') else None - total_tokens = response_meta.get("token_usage", {}).get("total_tokens", - 0) if response_meta else 0 + total_tokens = response_meta.get("token_usage", {}).get( + "total_tokens", + 0 + ) if response_meta else 0 yield total_tokens break if memory_flag: diff --git a/api/app/core/config.py b/api/app/core/config.py index 4a944557..64c5520e 100644 --- a/api/app/core/config.py +++ b/api/app/core/config.py @@ -231,8 +231,8 @@ class Settings: # Celery configuration (internal) # NOTE: 变量名不以 CELERY_ 开头,避免被 Celery CLI 的前缀匹配机制劫持 # 详见 docs/celery-env-bug-report.md - # 默认使用 Redis DB 3 (broker) 和 DB 4 (backend),与业务缓存 (DB 1/2) 隔离 - # 多人共用同一 Redis 时,每位开发者应在 .env 中配置不同的 DB 编号避免任务互相干扰 + # 默认使用 Redis 作为 broker 和 backend,与业务缓存隔离 + # 如需使用 RabbitMQ,在 .env 中设置 CELERY_BROKER_URL=amqp://user:pass@host:5672/vhost REDIS_DB_CELERY_BROKER: int = int(os.getenv("REDIS_DB_CELERY_BROKER", "3")) REDIS_DB_CELERY_BACKEND: int = int(os.getenv("REDIS_DB_CELERY_BACKEND", "4")) diff --git a/api/app/core/logging_config.py b/api/app/core/logging_config.py index 28a98a46..d0dda84b 100644 --- a/api/app/core/logging_config.py +++ b/api/app/core/logging_config.py @@ -529,8 +529,9 @@ def log_time(step_name: str, duration: float, log_file: str = "logs/time.log") - # Fallback to console only if file write fails print(f"Warning: Could not write to timing log: {e}") - # Always print to console (backward compatible behavior) - print(f"✓ {step_name}: {duration:.2f}s") + # Always log at INFO level (avoids Celery treating stdout as WARNING) + _timing_logger = logging.getLogger(__name__) + _timing_logger.info(f"✓ {step_name}: {duration:.2f}s") def get_agent_logger(name: str = "agent_service", diff --git a/api/app/core/memory/agent/langgraph_graph/nodes/retrieve_nodes.py b/api/app/core/memory/agent/langgraph_graph/nodes/retrieve_nodes.py index f2cd0d3d..68260f26 100644 --- a/api/app/core/memory/agent/langgraph_graph/nodes/retrieve_nodes.py +++ b/api/app/core/memory/agent/langgraph_graph/nodes/retrieve_nodes.py @@ -155,7 +155,7 @@ async def clean_databases(data) -> str: # Process reranked results reranked = results.get('reranked_results', {}) if reranked: - for category in ['summaries', 'statements', 'chunks', 'entities']: + for category in ['summaries', 'communities', 'statements', 'chunks', 'entities']: items = reranked.get(category, []) if isinstance(items, list): content_list.extend(items) @@ -169,11 +169,18 @@ async def clean_databases(data) -> str: elif isinstance(time_search, list): content_list.extend(time_search) - # Extract text content + # Extract text content,对 community 按 name 去重(多次 tool 调用会产生重复) text_parts = [] + seen_community_names = set() for item in content_list: if isinstance(item, dict): - text = item.get('statement') or item.get('content', '') + # community 节点用 name 去重 + if 'member_count' in item or 'core_entities' in item: + community_name = item.get('name') or item.get('id', '') + if community_name in seen_community_names: + continue + seen_community_names.add(community_name) + text = item.get('statement') or item.get('content') or item.get('summary', '') if text: text_parts.append(text) elif isinstance(item, str): @@ -354,7 +361,11 @@ async def retrieve(state: ReadState) -> ReadState: ) time_retrieval_tool = create_time_retrieval_tool(end_user_id) - search_params = {"end_user_id": end_user_id, "return_raw_results": True} + search_params = { + "end_user_id": end_user_id, + "return_raw_results": True, + "include": ["summaries", "statements", "chunks", "entities", "communities"], + } hybrid_retrieval = create_hybrid_retrieval_tool_sync(memory_config, **search_params) agent = create_agent( llm, @@ -390,8 +401,32 @@ async def retrieve(state: ReadState) -> ReadState: raw_results = tool_results['content'] clean_content = await clean_databases(raw_results) + # 社区展开:从 tool 返回结果中提取命中的 community, + # 沿 BELONGS_TO_COMMUNITY 关系拉取关联 Statement 追加到 clean_content + _expanded_stmts_to_write = [] + try: + results_dict = raw_results.get('results', {}) if isinstance(raw_results, dict) else {} + reranked = results_dict.get('reranked_results', {}) + community_hits = reranked.get('communities', []) + if not community_hits: + community_hits = results_dict.get('communities', []) + if community_hits: + from app.core.memory.agent.services.search_service import expand_communities_to_statements + _expanded_stmts_to_write, new_texts = await expand_communities_to_statements( + community_results=community_hits, + end_user_id=end_user_id, + existing_content=clean_content, + ) + if new_texts: + clean_content = clean_content + '\n' + '\n'.join(new_texts) + except Exception as parse_err: + logger.warning(f"[Retrieve] 解析社区命中结果失败,跳过展开: {parse_err}") + try: raw_results = raw_results['results'] + # 写回展开结果,接口返回中可见(已在 helper 中清洗过字段) + if _expanded_stmts_to_write and isinstance(raw_results, dict): + raw_results.setdefault('reranked_results', {})['expanded_statements'] = _expanded_stmts_to_write except Exception: raw_results = [] diff --git a/api/app/core/memory/agent/langgraph_graph/nodes/summary_nodes.py b/api/app/core/memory/agent/langgraph_graph/nodes/summary_nodes.py index 030acc9a..d967a285 100644 --- a/api/app/core/memory/agent/langgraph_graph/nodes/summary_nodes.py +++ b/api/app/core/memory/agent/langgraph_graph/nodes/summary_nodes.py @@ -334,13 +334,22 @@ async def Input_Summary(state: ReadState) -> ReadState: "end_user_id": end_user_id, "question": data, "return_raw_results": True, - "include": ["summaries"] # Only search summary nodes for faster performance + "include": ["summaries", "communities"] # MemorySummary 和 Community 同为高维度概括节点 } try: if storage_type != "rag": - retrieve_info, question, raw_results = await SearchService().execute_hybrid_search(**search_params, - memory_config=memory_config) + retrieve_info, question, raw_results = await SearchService().execute_hybrid_search( + **search_params, + memory_config=memory_config, + expand_communities=False, # 路径 "2" 只需要 community 的 summary 文本,不展开到 Statement + ) + # 调试:打印 community 检索结果数量 + if raw_results and isinstance(raw_results, dict): + reranked = raw_results.get('reranked_results', {}) + community_hits = reranked.get('communities', []) + logger.debug(f"[Input_Summary] community 命中数: {len(community_hits)}, " + f"summary 命中数: {len(reranked.get('summaries', []))}") else: retrieval_knowledge, retrieve_info, question, raw_results = await rag_knowledge(state, data) except Exception as e: diff --git a/api/app/core/memory/agent/langgraph_graph/routing/write_router.py b/api/app/core/memory/agent/langgraph_graph/routing/write_router.py index 6176caf5..2074b6ca 100644 --- a/api/app/core/memory/agent/langgraph_graph/routing/write_router.py +++ b/api/app/core/memory/agent/langgraph_graph/routing/write_router.py @@ -178,7 +178,7 @@ async def window_dialogue(end_user_id, langchain_messages, memory_config, scope) count_store.update_sessions_count(end_user_id, is_end_user_id, langchain_messages) elif int(is_end_user_id) == int(scope): logger.info('写入长期记忆NEO4J') - formatted_messages = (redis_messages) + formatted_messages = redis_messages # Get config_id (if memory_config is an object, extract config_id; otherwise use directly) if hasattr(memory_config, 'config_id'): config_id = memory_config.config_id diff --git a/api/app/core/memory/agent/langgraph_graph/tools/tool.py b/api/app/core/memory/agent/langgraph_graph/tools/tool.py index 9bd2b2cf..ae2c5772 100644 --- a/api/app/core/memory/agent/langgraph_graph/tools/tool.py +++ b/api/app/core/memory/agent/langgraph_graph/tools/tool.py @@ -252,9 +252,10 @@ def create_hybrid_retrieval_tool_async(memory_config, **search_params): # TODO: fact_summary functionality temporarily disabled, will be enabled after future development fields_to_remove = { 'invalid_at', 'valid_at', 'chunk_id_from_rel', 'entity_ids', - 'expired_at', 'created_at', 'chunk_id', 'id', 'apply_id', + 'expired_at', 'created_at', 'chunk_id', 'apply_id', 'user_id', 'statement_ids', 'updated_at', "chunk_ids", "fact_summary" } + # 注意:'id' 字段保留,community 展开时需要用 community id 查询成员 statements if isinstance(data, dict): # Clean dictionary @@ -310,7 +311,7 @@ def create_hybrid_retrieval_tool_async(memory_config, **search_params): "search_type": search_type, "end_user_id": end_user_id or search_params.get("end_user_id"), "limit": limit or search_params.get("limit", 10), - "include": search_params.get("include", ["summaries", "statements", "chunks", "entities"]), + "include": search_params.get("include", ["summaries", "statements", "chunks", "entities", "communities"]), "output_path": None, # Don't save to file "memory_config": memory_config, "rerank_alpha": rerank_alpha, diff --git a/api/app/core/memory/agent/services/search_service.py b/api/app/core/memory/agent/services/search_service.py index 4fc4256e..90b1c088 100644 --- a/api/app/core/memory/agent/services/search_service.py +++ b/api/app/core/memory/agent/services/search_service.py @@ -13,6 +13,72 @@ from app.core.memory.utils.data.text_utils import escape_lucene_query logger = get_agent_logger(__name__) +# 需要从展开结果中过滤的字段(含 Neo4j DateTime,不可 JSON 序列化) +_EXPAND_FIELDS_TO_REMOVE = { + 'invalid_at', 'valid_at', 'chunk_id_from_rel', 'entity_ids', + 'expired_at', 'created_at', 'chunk_id', 'apply_id', + 'user_id', 'statement_ids', 'updated_at', 'chunk_ids', 'fact_summary' +} + + +def _clean_expand_fields(obj): + """递归过滤展开结果中不可序列化的字段(DateTime 等)。""" + if isinstance(obj, dict): + return {k: _clean_expand_fields(v) for k, v in obj.items() if k not in _EXPAND_FIELDS_TO_REMOVE} + if isinstance(obj, list): + return [_clean_expand_fields(i) for i in obj] + return obj + + +async def expand_communities_to_statements( + community_results: List[dict], + end_user_id: str, + existing_content: str = "", + limit: int = 10, +) -> Tuple[List[dict], List[str]]: + """ + 社区展开 helper:给定命中的 community 列表,拉取关联 Statement。 + + - 对展开结果去重(过滤已在 existing_content 中出现的文本) + - 过滤不可序列化字段 + - 返回 (cleaned_expanded_stmts, new_texts) + - cleaned_expanded_stmts: 可直接写回 raw_results 的列表 + - new_texts: 去重后新增的 statement 文本列表,用于追加到 clean_content + """ + community_ids = [r.get("id") for r in community_results if r.get("id")] + if not community_ids or not end_user_id: + return [], [] + + from app.repositories.neo4j.graph_search import search_graph_community_expand + from app.repositories.neo4j.neo4j_connector import Neo4jConnector + + connector = Neo4jConnector() + try: + result = await search_graph_community_expand( + connector=connector, + community_ids=community_ids, + end_user_id=end_user_id, + limit=limit, + ) + except Exception as e: + logger.warning(f"[expand_communities] 社区展开检索失败,跳过: {e}") + return [], [] + finally: + await connector.close() + + expanded_stmts = result.get("expanded_statements", []) + if not expanded_stmts: + return [], [] + + existing_lines = set(existing_content.splitlines()) + new_texts = [ + s["statement"] for s in expanded_stmts + if s.get("statement") and s["statement"] not in existing_lines + ] + cleaned = _clean_expand_fields(expanded_stmts) + logger.info(f"[expand_communities] 展开 {len(expanded_stmts)} 条 statements,新增 {len(new_texts)} 条,community_ids={community_ids}") + return cleaned, new_texts + class SearchService: """Service for executing hybrid search and processing results.""" @@ -21,7 +87,7 @@ class SearchService: """Initialize the search service.""" logger.info("SearchService initialized") - def extract_content_from_result(self, result: dict) -> str: + def extract_content_from_result(self, result: dict, node_type: str = "") -> str: """ Extract only meaningful content from search results, dropping all metadata. @@ -30,9 +96,11 @@ class SearchService: - Entities: extract 'name' and 'fact_summary' fields - Summaries: extract 'content' field - Chunks: extract 'content' field + - Communities: extract 'content' field (c.summary), prefixed with community name Args: result: Search result dictionary + node_type: Hint for node type ("community", "summary", etc.) Returns: Clean content string without metadata @@ -46,8 +114,21 @@ class SearchService: if 'statement' in result and result['statement']: content_parts.append(result['statement']) - # Summaries/Chunks: extract content field - if 'content' in result and result['content']: + # Community 节点:有 member_count 或 core_entities 字段,或 node_type 明确指定 + # 用 "[主题:{name}]" 前缀区分,让 LLM 知道这是主题级摘要 + is_community = ( + node_type == "community" + or 'member_count' in result + or 'core_entities' in result + ) + if is_community: + name = result.get('name', '') + content = result.get('content', '') + if content: + prefix = f"[主题:{name}] " if name else "" + content_parts.append(f"{prefix}{content}") + elif 'content' in result and result['content']: + # Summaries / Chunks content_parts.append(result['content']) # Entities: extract name and fact_summary (commented out in original) @@ -99,7 +180,8 @@ class SearchService: rerank_alpha: float = 0.4, output_path: str = "search_results.json", return_raw_results: bool = False, - memory_config = None + memory_config = None, + expand_communities: bool = True, ) -> Tuple[str, str, Optional[dict]]: """ Execute hybrid search and return clean content. @@ -114,13 +196,15 @@ class SearchService: output_path: Path to save search results (default: "search_results.json") return_raw_results: If True, also return the raw search results as third element (default: False) memory_config: Memory configuration object (required) + expand_communities: If True, expand community hits to member statements (default: True). + Set to False for quick-summary paths that only need community-level text. Returns: Tuple of (clean_content, cleaned_query, raw_results) raw_results is None if return_raw_results=False """ if include is None: - include = ["statements", "chunks", "entities", "summaries"] + include = ["statements", "chunks", "entities", "summaries", "communities"] # Clean query cleaned_query = self.clean_query(question) @@ -146,8 +230,8 @@ class SearchService: if search_type == "hybrid": reranked_results = answer.get('reranked_results', {}) - # Priority order: summaries first (most contextual), then statements, chunks, entities - priority_order = ['summaries', 'statements', 'chunks', 'entities'] + # Priority order: summaries first (most contextual), then communities, statements, chunks, entities + priority_order = ['summaries', 'communities', 'statements', 'chunks', 'entities'] for category in priority_order: if category in include and category in reranked_results: @@ -157,19 +241,33 @@ class SearchService: else: # For keyword or embedding search, results are directly in answer dict # Apply same priority order - priority_order = ['summaries', 'statements', 'chunks', 'entities'] + priority_order = ['summaries', 'communities', 'statements', 'chunks', 'entities'] for category in priority_order: if category in include and category in answer: category_results = answer[category] if isinstance(category_results, list): answer_list.extend(category_results) + + # 对命中的 community 节点展开其成员 statements(路径 "0"/"1" 需要,路径 "2" 不需要) + if expand_communities and "communities" in include: + community_results = ( + answer.get('reranked_results', {}).get('communities', []) + if search_type == "hybrid" + else answer.get('communities', []) + ) + cleaned_stmts, new_texts = await expand_communities_to_statements( + community_results=community_results, + end_user_id=end_user_id, + ) + answer_list.extend(cleaned_stmts) - # Extract clean content from all results - content_list = [ - self.extract_content_from_result(ans) - for ans in answer_list - ] + # Extract clean content from all results,按类型传入 node_type 区分 community + content_list = [] + for ans in answer_list: + # community 节点有 member_count 或 core_entities 字段 + ntype = "community" if ('member_count' in ans or 'core_entities' in ans) else "" + content_list.append(self.extract_content_from_result(ans, node_type=ntype)) # Filter out empty strings and join with newlines diff --git a/api/app/core/memory/agent/utils/get_dialogs.py b/api/app/core/memory/agent/utils/get_dialogs.py index ea44d0a5..4c667061 100644 --- a/api/app/core/memory/agent/utils/get_dialogs.py +++ b/api/app/core/memory/agent/utils/get_dialogs.py @@ -11,7 +11,7 @@ async def get_chunked_dialogs( chunker_strategy: str = "RecursiveChunker", end_user_id: str = "group_1", messages: list = None, - ref_id: str = "wyl_20251027", + ref_id: str = "", config_id: str = None ) -> List[DialogData]: """Generate chunks from structured messages using the specified chunker strategy. @@ -40,12 +40,13 @@ async def get_chunked_dialogs( role = msg['role'] content = msg['content'] + files = msg.get("file_content", []) if role not in ['user', 'assistant']: raise ValueError(f"Message {idx} role must be 'user' or 'assistant', got: {role}") if content.strip(): - conversation_messages.append(ConversationMessage(role=role, msg=content.strip())) + conversation_messages.append(ConversationMessage(role=role, msg=content.strip(), files=files)) if not conversation_messages: raise ValueError("Message list cannot be empty after filtering") @@ -84,7 +85,7 @@ async def get_chunked_dialogs( pruning_scene=memory_config.pruning_scene or "education", pruning_threshold=memory_config.pruning_threshold, scene_id=str(memory_config.scene_id) if memory_config.scene_id else None, - ontology_classes=memory_config.ontology_classes, + ontology_class_infos=memory_config.ontology_class_infos, ) logger.info(f"[剪枝] 加载配置: switch={pruning_config.pruning_switch}, scene={pruning_config.pruning_scene}, threshold={pruning_config.pruning_threshold}") diff --git a/api/app/core/memory/agent/utils/prompt/Problem_Extension_prompt.jinja2 b/api/app/core/memory/agent/utils/prompt/Problem_Extension_prompt.jinja2 index a0e21fbd..c78cbaac 100644 --- a/api/app/core/memory/agent/utils/prompt/Problem_Extension_prompt.jinja2 +++ b/api/app/core/memory/agent/utils/prompt/Problem_Extension_prompt.jinja2 @@ -39,6 +39,30 @@ 比如:输入历史信息内容:[{'Query': '4月27日,我和你推荐过一本书,书名是什么?', 'ANswer': '张曼玉推荐了《小王子》'}] 拆分问题:4月27日,我和你推荐过一本书,书名是什么?,可以拆分为:4月27日,张曼玉推荐过一本书,书名是什么? +## 指代消歧规则(Coreference Resolution): +在拆分问题时,必须解析并替换所有指代词和抽象称呼,使问题具体化: + +1. **"用户"的消歧**: + - "用户是谁?" → 分析历史记录,找出对话发起者的姓名 + - 如果历史中有"我叫X"、"我的名字是X"、或多次提到某个人物,则"用户"指的就是这个人 + - 示例:历史中有"老李的原名叫李建国",则"用户是谁?"应拆分为"李建国是谁?"或"老李(李建国)是谁?" + +2. **"我"的消歧**: + - "我喜欢什么?" → 从历史中找出对话发起者的姓名,替换为"X喜欢什么?" + - 示例:历史中有"张曼玉推荐了《小王子》",则"我推荐的书是什么?"应拆分为"张曼玉推荐的书是什么?" + +3. **"他/她/它"的消歧**: + - 从上下文或历史中找出最近提到的同类实体 + - 示例:历史中有"老李的同事叫他建国哥",则"他的同事怎么称呼他?"应拆分为"老李的同事怎么称呼他?" + +4. **"那个人/这个人"的消歧**: + - 从历史中找出最近提到的人物 + - 示例:历史中有"李建国",则"那个人的原名是什么?"应拆分为"李建国的原名是什么?" + +5. **优先级**: + - 如果历史记录中反复出现某个人物(如"老李"、"李建国"、"建国哥"),则"用户"很可能指的就是这个人 + - 如果无法从历史中确定指代对象,保留原问题,但在reason中说明"无法确定指代对象" + 输出要求: @@ -71,6 +95,34 @@ "reason": "输出原问题的关键要素" } ] + +## 指代消歧示例(重要): +示例1 - "用户"的消歧: +输入历史:[{'Query': '老李的原名叫什么?', 'Answer': '李建国'}, {'Query': '老李的同事叫他什么?', 'Answer': '建国哥'}] +输入问题:"用户是谁?" +输出: +[ + { + "original_question": "用户是谁?", + "extended_question": "李建国是谁?", + "type": "单跳", + "reason": "历史中反复提到'老李/李建国/建国哥','用户'指的就是对话发起者李建国" + } +] + +示例2 - "我"的消歧: +输入历史:[{'Query': '张曼玉推荐了什么书?', 'Answer': '《小王子》'}] +输入问题:"我推荐的书是什么?" +输出: +[ + { + "original_question": "我推荐的书是什么?", + "extended_question": "张曼玉推荐的书是什么?", + "type": "单跳", + "reason": "历史中提到张曼玉推荐了书,'我'指的就是张曼玉" + } +] + **Output format** **CRITICAL JSON FORMATTING REQUIREMENTS:** 1. Use only standard ASCII double quotes (") for JSON structure - never use Chinese quotation marks ("") or other Unicode quotes diff --git a/api/app/core/memory/agent/utils/prompt/problem_breakdown_prompt.jinja2 b/api/app/core/memory/agent/utils/prompt/problem_breakdown_prompt.jinja2 index aca716a4..ff134ddb 100644 --- a/api/app/core/memory/agent/utils/prompt/problem_breakdown_prompt.jinja2 +++ b/api/app/core/memory/agent/utils/prompt/problem_breakdown_prompt.jinja2 @@ -27,6 +27,30 @@ 比如:输入历史信息内容:[{'Query': '4月27日,我和你推荐过一本书,书名是什么?', 'ANswer': '张曼玉推荐了《小王子》'}] 拆分问题:4月27日,我和你推荐过一本书,书名是什么?,可以拆分为:4月27日,张曼玉推荐过一本书,书名是什么? +## 指代消歧规则(Coreference Resolution): +在拆分问题时,必须解析并替换所有指代词和抽象称呼,使问题具体化: + +1. **"用户"的消歧**: + - "用户是谁?" → 分析历史记录,找出对话发起者的姓名 + - 如果历史中有"我叫X"、"我的名字是X"、或多次提到某个人物(如"老李"、"李建国"),则"用户"指的就是这个人 + - 示例:历史中反复出现"老李/李建国/建国哥",则"用户是谁?"应拆分为"李建国是谁?"或"老李(李建国)是谁?" + +2. **"我"的消歧**: + - "我喜欢什么?" → 从历史中找出对话发起者的姓名,替换为"X喜欢什么?" + - 示例:历史中有"张曼玉推荐了《小王子》",则"我推荐的书是什么?"应拆分为"张曼玉推荐的书是什么?" + +3. **"他/她/它"的消歧**: + - 从上下文或历史中找出最近提到的同类实体 + - 示例:历史中有"老李的同事叫他建国哥",则"他的同事怎么称呼他?"应拆分为"老李的同事怎么称呼他?" + +4. **"那个人/这个人"的消歧**: + - 从历史中找出最近提到的人物 + - 示例:历史中有"李建国",则"那个人的原名是什么?"应拆分为"李建国的原名是什么?" + +5. **优先级**: + - 如果历史记录中反复出现某个人物(如"老李"、"李建国"、"建国哥"),则"用户"很可能指的就是这个人 + - 如果无法从历史中确定指代对象,保留原问题,但在reason中说明"无法确定指代对象" + ## 指令: 你是一个智能数据拆分助手,请根据数据特性判断输入属于哪种类型: 单跳(Single-hop) @@ -151,6 +175,34 @@ ] - 必须通过json.loads()的格式支持的形式输出 - 必须通过json.loads()的格式支持的形式输出,响应必须是与此确切模式匹配的有效JSON对象。不要在JSON之前或之后包含任何文本。 + +## 指代消歧示例(重要): +示例1 - "用户"的消歧: +输入历史:[{'Query': '老李的原名叫什么?', 'Answer': '李建国'}, {'Query': '老李的同事叫他什么?', 'Answer': '建国哥'}] +输入问题:"用户是谁?" +输出: +[ + { + "id": "Q1", + "question": "李建国是谁?", + "type": "单跳", + "reason": "历史中反复提到'老李/李建国/建国哥','用户'指的就是对话发起者李建国" + } +] + +示例2 - "我"的消歧: +输入历史:[{'Query': '张曼玉推荐了什么书?', 'Answer': '《小王子》'}] +输入问题:"我推荐的书是什么?" +输出: +[ + { + "id": "Q1", + "question": "张曼玉推荐的书是什么?", + "type": "单跳", + "reason": "历史中提到张曼玉推荐了书,'我'指的就是张曼玉" + } +] + - 关键的JSON格式要求 1.JSON结构仅使用标准ASCII双引号(“)-切勿使用中文引号(“”)或其他Unicode引号 2.如果提取的语句文本包含引号,请使用反斜杠(\“)正确转义它们 diff --git a/api/app/core/memory/agent/utils/write_tools.py b/api/app/core/memory/agent/utils/write_tools.py index b3707083..1f437973 100644 --- a/api/app/core/memory/agent/utils/write_tools.py +++ b/api/app/core/memory/agent/utils/write_tools.py @@ -6,14 +6,17 @@ pipeline. Only MemoryConfig is needed - clients are constructed internally. """ import asyncio import time +import uuid from datetime import datetime +from typing import List, Optional from dotenv import load_dotenv from app.core.logging_config import get_agent_logger from app.core.memory.agent.utils.get_dialogs import get_chunked_dialogs from app.core.memory.storage_services.extraction_engine.extraction_orchestrator import ExtractionOrchestrator -from app.core.memory.storage_services.extraction_engine.knowledge_extraction.memory_summary import memory_summary_generation +from app.core.memory.storage_services.extraction_engine.knowledge_extraction.memory_summary import \ + memory_summary_generation from app.core.memory.utils.llm.llm_utils import MemoryClientFactory from app.core.memory.utils.log.logging_utils import log_time from app.db import get_db_context @@ -23,18 +26,17 @@ from app.repositories.neo4j.graph_saver import save_dialog_and_statements_to_neo from app.repositories.neo4j.neo4j_connector import Neo4jConnector from app.schemas.memory_config_schema import MemoryConfig - load_dotenv() logger = get_agent_logger(__name__) async def write( - end_user_id: str, - memory_config: MemoryConfig, - messages: list, - ref_id: str = "wyl20251027", - language: str = "zh", + end_user_id: str, + memory_config: MemoryConfig, + messages: list, + ref_id: str = "", + language: str = "zh", ) -> None: """ Execute the complete knowledge extraction pipeline. @@ -43,9 +45,11 @@ async def write( end_user_id: Group identifier memory_config: MemoryConfig object containing all configuration messages: Structured message list [{"role": "user", "content": "..."}, ...] - ref_id: Reference ID, defaults to "wyl20251027" + ref_id: Reference ID, defaults to "" language: 语言类型 ("zh" 中文, "en" 英文),默认中文 """ + if not ref_id: + ref_id = uuid.uuid4().hex # Extract config values embedding_model_id = str(memory_config.embedding_model_id) chunker_strategy = memory_config.chunker_strategy @@ -99,14 +103,14 @@ async def write( if memory_config.scene_id: try: from app.core.memory.ontology_services.ontology_type_loader import load_ontology_types_for_scene - + with get_db_context() as db: ontology_types = load_ontology_types_for_scene( scene_id=memory_config.scene_id, workspace_id=memory_config.workspace_id, db=db ) - + if ontology_types: logger.info( f"Loaded {len(ontology_types.types)} ontology types for scene_id: {memory_config.scene_id}" @@ -135,9 +139,11 @@ async def write( all_chunk_nodes, all_statement_nodes, all_entity_nodes, + all_perceptual_nodes, all_statement_chunk_edges, all_statement_entity_edges, all_entity_entity_edges, + all_perceptual_edges, all_dedup_details, ) = await orchestrator.run(chunked_dialogs, is_pilot_run=False) @@ -145,11 +151,6 @@ async def write( # Step 3: Save all data to Neo4j database step_start = time.time() - from app.repositories.neo4j.create_indexes import create_fulltext_indexes - try: - await create_fulltext_indexes() - except Exception as e: - logger.error(f"Error creating indexes: {e}", exc_info=True) # 添加死锁重试机制 max_retries = 3 @@ -162,15 +163,43 @@ async def write( chunk_nodes=all_chunk_nodes, statement_nodes=all_statement_nodes, entity_nodes=all_entity_nodes, + perceptual_nodes=all_perceptual_nodes, statement_chunk_edges=all_statement_chunk_edges, statement_entity_edges=all_statement_entity_edges, entity_edges=all_entity_entity_edges, + perceptual_edges=all_perceptual_edges, connector=neo4j_connector, - config_id=config_id, - llm_model_id=str(memory_config.llm_model_id) if memory_config.llm_model_id else None, ) if success: logger.info("Successfully saved all data to Neo4j") + + # 使用 Celery 异步任务触发聚类(不阻塞主流程) + if all_entity_nodes: + try: + from app.tasks import run_incremental_clustering + + end_user_id = all_entity_nodes[0].end_user_id + new_entity_ids = [e.id for e in all_entity_nodes] + + # 异步提交 Celery 任务 + task = run_incremental_clustering.apply_async( + kwargs={ + "end_user_id": end_user_id, + "new_entity_ids": new_entity_ids, + "llm_model_id": str(memory_config.llm_model_id) if memory_config.llm_model_id else None, + "embedding_model_id": str(memory_config.embedding_model_id) if memory_config.embedding_model_id else None, + }, + # 设置任务优先级(低优先级,不影响主业务) + priority=3, + ) + logger.info( + f"[Clustering] 增量聚类任务已提交到 Celery - " + f"task_id={task.id}, end_user_id={end_user_id}, entity_count={len(new_entity_ids)}" + ) + except Exception as e: + # 聚类任务提交失败不影响主流程 + logger.error(f"[Clustering] 提交聚类任务失败(不影响主流程): {e}", exc_info=True) + break else: logger.warning("Failed to save some data to Neo4j") @@ -204,9 +233,8 @@ async def write( summaries = await memory_summary_generation( chunked_dialogs, llm_client=llm_client, embedder_client=embedder_client, language=language ) - + ms_connector = Neo4jConnector() try: - ms_connector = Neo4jConnector() await add_memory_summary_nodes(summaries, ms_connector) await add_memory_summary_statement_edges(summaries, ms_connector) finally: @@ -246,5 +274,21 @@ async def write( except Exception as cache_err: logger.warning(f"[WRITE] 写入活动统计缓存失败(不影响主流程): {cache_err}", exc_info=True) + # Close LLM/Embedder underlying httpx clients to prevent + # 'RuntimeError: Event loop is closed' during garbage collection + for client_obj in (llm_client, embedder_client): + try: + underlying = getattr(client_obj, 'client', None) or getattr(client_obj, 'model', None) + if underlying is None: + continue + # Unwrap RedBearLLM / RedBearEmbeddings to get the LangChain model + inner = getattr(underlying, '_model', underlying) + # LangChain OpenAI models expose async_client (httpx.AsyncClient) + http_client = getattr(inner, 'async_client', None) + if http_client is not None and hasattr(http_client, 'aclose'): + await http_client.aclose() + except Exception: + pass + logger.info("=== Pipeline Complete ===") - logger.info(f"Total execution time: {total_time:.2f} seconds") \ No newline at end of file + logger.info(f"Total execution time: {total_time:.2f} seconds") diff --git a/api/app/core/memory/llm_tools/chunker_client.py b/api/app/core/memory/llm_tools/chunker_client.py index 93a2df82..51d15aab 100644 --- a/api/app/core/memory/llm_tools/chunker_client.py +++ b/api/app/core/memory/llm_tools/chunker_client.py @@ -1,10 +1,10 @@ -from typing import Any, List -import re -import os import asyncio import json -import numpy as np import logging +import os +from typing import Any, List + +import numpy as np # Fix tokenizer parallelism warning os.environ["TOKENIZERS_PARALLELISM"] = "false" @@ -246,6 +246,7 @@ class ChunkerClient: "total_sub_chunks": len(sub_chunks), "chunker_strategy": self.chunker_config.chunker_strategy, }, + files=msg.files ) dialogue.chunks.append(chunk) else: @@ -258,6 +259,7 @@ class ChunkerClient: "message_role": msg.role, "chunker_strategy": self.chunker_config.chunker_strategy, }, + files=msg.files ) dialogue.chunks.append(chunk) diff --git a/api/app/core/memory/llm_tools/openai_client.py b/api/app/core/memory/llm_tools/openai_client.py index 43c2b445..c70fef5f 100644 --- a/api/app/core/memory/llm_tools/openai_client.py +++ b/api/app/core/memory/llm_tools/openai_client.py @@ -65,7 +65,7 @@ class OpenAIClient(LLMClient): type=type_ ) - logger.info(f"OpenAI 客户端初始化完成: type={type_}") + logger.debug(f"OpenAI 客户端初始化完成: type={type_}") async def chat(self, messages: List[Dict[str, str]], **kwargs) -> Any: """ diff --git a/api/app/core/memory/llm_tools/openai_embedder.py b/api/app/core/memory/llm_tools/openai_embedder.py index 2d6fccbc..6ae87887 100644 --- a/api/app/core/memory/llm_tools/openai_embedder.py +++ b/api/app/core/memory/llm_tools/openai_embedder.py @@ -2,6 +2,7 @@ OpenAI Embedder 客户端实现 基于 LangChain 和 RedBearEmbeddings 的 OpenAI 嵌入模型客户端实现。 +自动支持火山引擎的多模态 Embedding。 """ from typing import List @@ -13,6 +14,7 @@ from app.core.memory.llm_tools.embedder_client import ( ) from app.core.models.base import RedBearModelConfig from app.core.models.embedding import RedBearEmbeddings +from app.models.models_model import ModelProvider logger = logging.getLogger(__name__) @@ -25,6 +27,7 @@ class OpenAIEmbedderClient(EmbedderClient): - 批量文本嵌入 - 自动重试机制 - 错误处理 + - 火山引擎多模态 Embedding(自动识别) """ def __init__(self, model_config: RedBearModelConfig): @@ -36,7 +39,7 @@ class OpenAIEmbedderClient(EmbedderClient): """ super().__init__(model_config) - # 初始化 RedBearEmbeddings 模型 + # 初始化 RedBearEmbeddings(自动支持火山引擎多模态) self.model = RedBearEmbeddings( RedBearModelConfig( model_name=self.model_name, @@ -47,8 +50,9 @@ class OpenAIEmbedderClient(EmbedderClient): timeout=self.timeout, ) ) + self.is_multimodal = self.model.is_multimodal_supported() - logger.info("OpenAI Embedder 客户端初始化完成") + logger.info(f"OpenAI Embedder 客户端初始化完成 (provider={self.provider}, multimodal={self.is_multimodal})") async def response( self, @@ -77,7 +81,14 @@ class OpenAIEmbedderClient(EmbedderClient): return [] # 生成嵌入向量 - embeddings = await self.model.aembed_documents(texts) + if self.is_multimodal: + # 火山引擎多模态 Embedding + embeddings = await self.model.aembed_multimodal( + [{"type": "text", "text": text} for text in texts] + ) + else: + # 普通 Embedding + embeddings = await self.model.aembed_documents(texts) logger.debug(f"成功生成 {len(embeddings)} 个嵌入向量") return embeddings diff --git a/api/app/core/memory/models/config_models.py b/api/app/core/memory/models/config_models.py index c2d62ac1..5ed50b7f 100644 --- a/api/app/core/memory/models/config_models.py +++ b/api/app/core/memory/models/config_models.py @@ -6,6 +6,7 @@ of the memory system including LLM, chunking, pruning, and search. Classes: LLMConfig: Configuration for LLM client ChunkerConfig: Configuration for dialogue chunking + OntologyClassInfo: Single ontology class with name and description PruningConfig: Configuration for semantic pruning TemporalSearchParams: Parameters for temporal search queries """ @@ -50,30 +51,41 @@ class ChunkerConfig(BaseModel): min_characters_per_chunk: Optional[int] = Field(24, ge=0, description="The minimum number of characters in each chunk.") +class OntologyClassInfo(BaseModel): + """本体类型的名称与语义描述,用于剪枝提示词注入。 + + Attributes: + class_name: 本体类型名称(如"患者"、"课程") + class_description: 本体类型语义描述,告知 LLM 该类型在当前场景下的含义 + """ + class_name: str = Field(..., description="本体类型名称") + class_description: str = Field(default="", description="本体类型语义描述") + + class PruningConfig(BaseModel): """Configuration for semantic pruning of dialogue content. Attributes: pruning_switch: Enable or disable semantic pruning - pruning_scene: Scene name for pruning, either a built-in key - ('education', 'online_service', 'outbound') or a custom scene_name - from ontology_scene table + pruning_scene: Scene name for pruning from ontology_scene table pruning_threshold: Pruning ratio (0-0.9, max 0.9 to avoid complete removal) - scene_id: Optional ontology scene UUID, used to load custom ontology classes - ontology_classes: List of class_name strings from ontology_class table, - injected into the prompt when pruning_scene is not a built-in scene + scene_id: Optional ontology scene UUID + ontology_class_infos: Full ontology class info (name + description) from + ontology_class table, injected into the pruning prompt to drive + scene-aware preservation decisions """ pruning_switch: bool = Field(False, description="Enable semantic pruning when True.") pruning_scene: str = Field( "education", - description="Scene for pruning: built-in key or custom scene_name from ontology_scene.", + description="Scene name from ontology_scene table.", ) pruning_threshold: float = Field( 0.5, ge=0.0, le=0.9, description="Pruning ratio within 0-0.9 (max 0.9 to avoid termination).") scene_id: Optional[str] = Field(None, description="Ontology scene UUID (optional).") - ontology_classes: Optional[List[str]] = Field( - None, description="Class names from ontology_class table for custom scenes." + ontology_class_infos: List[OntologyClassInfo] = Field( + default_factory=list, + description="Full ontology class info (name + description) injected into pruning prompt." ) diff --git a/api/app/core/memory/models/graph_models.py b/api/app/core/memory/models/graph_models.py index 1880b9ab..1b8c9d52 100644 --- a/api/app/core/memory/models/graph_models.py +++ b/api/app/core/memory/models/graph_models.py @@ -44,21 +44,21 @@ def parse_historical_datetime(v): """ if v is None: return v - + # 处理 Neo4j DateTime 对象 if hasattr(v, 'to_native'): return v.to_native() - + # 处理 Python datetime 对象 if isinstance(v, datetime): return v - + if isinstance(v, str): # 匹配 ISO 8601 格式:YYYY-MM-DD 或 YYYY-MM-DDTHH:MM:SS[.ffffff][Z|±HH:MM] # 支持1-4位年份 pattern = r'^(\d{1,4})-(\d{2})-(\d{2})(?:T(\d{2}):(\d{2}):(\d{2})(?:\.(\d+))?(?:Z|([+-]\d{2}:\d{2}))?)?' match = re.match(pattern, v) - + if match: try: year = int(match.group(1)) @@ -68,31 +68,31 @@ def parse_historical_datetime(v): minute = int(match.group(5)) if match.group(5) else 0 second = int(match.group(6)) if match.group(6) else 0 microsecond = 0 - + # 处理微秒 if match.group(7): # 补齐或截断到6位 us_str = match.group(7).ljust(6, '0')[:6] microsecond = int(us_str) - + # 处理时区 tzinfo = None if 'Z' in v or match.group(8): tzinfo = timezone.utc - + # 创建 datetime 对象 return datetime(year, month, day, hour, minute, second, microsecond, tzinfo=tzinfo) - + except (ValueError, OverflowError): # 日期值无效(如月份13、日期32等) return None - + # 如果不匹配模式,尝试使用 fromisoformat(用于标准格式) try: return datetime.fromisoformat(v.replace('Z', '+00:00')) except Exception: return None - + return v @@ -114,7 +114,7 @@ class Edge(BaseModel): end_user_id: str = Field(..., description="The end user ID of the edge.") run_id: str = Field(default_factory=lambda: uuid4().hex, description="Unique identifier for this pipeline run.") created_at: datetime = Field(..., description="The valid time of the edge from system perspective.") - expired_at: Optional[datetime] = Field(None, description="The expired time of the edge from system perspective.") + expired_at: Optional[datetime] = Field(default=None, description="The expired time of the edge from system perspective.") class ChunkEdge(Edge): @@ -167,7 +167,7 @@ class EntityEntityEdge(Edge): source_statement_id: str = Field(..., description="Statement where this relationship was extracted") valid_at: Optional[datetime] = Field(None, description="Temporal validity start") invalid_at: Optional[datetime] = Field(None, description="Temporal validity end") - + @field_validator('valid_at', 'invalid_at', mode='before') @classmethod def validate_datetime(cls, v): @@ -175,6 +175,12 @@ class EntityEntityEdge(Edge): return parse_historical_datetime(v) +class PerceptualEdge(Edge): + """Edge connecting perceptual nodes to their source chunks + """ + pass + + class Node(BaseModel): """Base class for all graph nodes in the knowledge graph. @@ -206,7 +212,8 @@ class DialogueNode(Node): ref_id: str = Field(..., description="Reference identifier of the dialog") content: str = Field(..., description="Dialogue content") dialog_embedding: Optional[List[float]] = Field(None, description="Dialog embedding vector") - config_id: Optional[int | str] = Field(None, description="Configuration ID used to process this dialogue (integer or string)") + config_id: Optional[int | str] = Field(None, + description="Configuration ID used to process this dialogue (integer or string)") class StatementNode(Node): @@ -241,17 +248,17 @@ class StatementNode(Node): chunk_id: str = Field(..., description="ID of the parent chunk") stmt_type: str = Field(..., description="Type of the statement") statement: str = Field(..., description="The statement text content") - + # Speaker identification speaker: Optional[str] = Field( None, description="Speaker identifier: 'user' for user messages, 'assistant' for AI responses" ) - + # Emotion fields (ordered as requested, emotion_intensity first for display) emotion_intensity: Optional[float] = Field( - None, - ge=0.0, + None, + ge=0.0, le=1.0, description="Emotion intensity: 0.0-1.0 (displayed on node)" ) @@ -264,25 +271,26 @@ class StatementNode(Node): description="Emotion subject: self/other/object" ) emotion_type: Optional[str] = Field( - None, + None, description="Emotion type: joy/sadness/anger/fear/surprise/neutral" ) emotion_keywords: Optional[List[str]] = Field( default_factory=list, description="Emotion keywords list, max 3 items" ) - + # Temporal fields temporal_info: TemporalInfo = Field(..., description="Temporal information") valid_at: Optional[datetime] = Field(None, description="Temporal validity start") invalid_at: Optional[datetime] = Field(None, description="Temporal validity end") - + # Embedding and other fields statement_embedding: Optional[List[float]] = Field(None, description="Statement embedding vector") chunk_embedding: Optional[List[float]] = Field(None, description="Chunk embedding vector") connect_strength: str = Field(..., description="Strong VS Weak classification of this statement") - config_id: Optional[int | str] = Field(None, description="Configuration ID used to process this statement (integer or string)") - + config_id: Optional[int | str] = Field(None, + description="Configuration ID used to process this statement (integer or string)") + # ACT-R Memory Activation Properties importance_score: float = Field( default=0.5, @@ -309,13 +317,13 @@ class StatementNode(Node): ge=0, description="Total number of times this node has been accessed" ) - + @field_validator('valid_at', 'invalid_at', mode='before') @classmethod def validate_datetime(cls, v): """使用通用的历史日期解析函数""" return parse_historical_datetime(v) - + @field_validator('emotion_type', mode='before') @classmethod def validate_emotion_type(cls, v): @@ -326,7 +334,7 @@ class StatementNode(Node): if v not in valid_types: raise ValueError(f"emotion_type must be one of {valid_types}, got {v}") return v - + @field_validator('emotion_subject', mode='before') @classmethod def validate_emotion_subject(cls, v): @@ -337,7 +345,7 @@ class StatementNode(Node): if v not in valid_subjects: raise ValueError(f"emotion_subject must be one of {valid_subjects}, got {v}") return v - + @field_validator('emotion_keywords', mode='before') @classmethod def validate_emotion_keywords(cls, v): @@ -405,19 +413,20 @@ class ExtractedEntityNode(Node): entity_type: str = Field(..., description="Type of the entity") description: str = Field(..., description="Entity description") example: str = Field( - default="", + default="", description="A concise example (around 20 characters) to help understand the entity" ) aliases: List[str] = Field( - default_factory=list, + default_factory=list, description="Entity aliases - alternative names for this entity" ) name_embedding: Optional[List[float]] = Field(default_factory=list, description="Name embedding vector") # TODO: fact_summary 功能暂时禁用,待后续开发完善后启用 # fact_summary: str = Field(default="", description="Summary of the fact about this entity") connect_strength: str = Field(..., description="Strong VS Weak about this entity") - config_id: Optional[int | str] = Field(None, description="Configuration ID used to process this entity (integer or string)") - + config_id: Optional[int | str] = Field(None, + description="Configuration ID used to process this entity (integer or string)") + # ACT-R Memory Activation Properties importance_score: float = Field( default=0.5, @@ -444,16 +453,16 @@ class ExtractedEntityNode(Node): ge=0, description="Total number of times this node has been accessed" ) - + # Explicit Memory Classification is_explicit_memory: bool = Field( default=False, description="Whether this entity represents explicit/semantic memory (knowledge, concepts, definitions, theories, principles)" ) - + @field_validator('aliases', mode='before') @classmethod - def validate_aliases_field(cls, v): # 字段验证器 自动清理和验证 aliases 字段 + def validate_aliases_field(cls, v): # 字段验证器 自动清理和验证 aliases 字段 """Validate and clean aliases field using utility function. This validator ensures that the aliases field is always a valid list of strings. @@ -507,8 +516,9 @@ class MemorySummaryNode(Node): memory_type: Optional[str] = Field(None, description="Type/category of the episodic memory") summary_embedding: Optional[List[float]] = Field(None, description="Embedding vector for the summary") metadata: dict = Field(default_factory=dict, description="Additional metadata for the summary") - config_id: Optional[int | str] = Field(None, description="Configuration ID used to process this summary (integer or string)") - + config_id: Optional[int | str] = Field(None, + description="Configuration ID used to process this summary (integer or string)") + # ACT-R Forgetting Engine Properties original_statement_id: Optional[str] = Field( None, @@ -522,7 +532,7 @@ class MemorySummaryNode(Node): None, description="Timestamp when the nodes were merged" ) - + # ACT-R Memory Activation Properties importance_score: float = Field( default=0.5, @@ -549,3 +559,18 @@ class MemorySummaryNode(Node): ge=0, description="Total number of times this node has been accessed (reset to 1 on creation)" ) + + +class PerceptualNode(Node): + """Node representing a multimodal message in the knowledge graph. + """ + perceptual_type: int + file_path: str + file_name: str + file_ext: str + summary: str + keywords: list[str] + topic: str + domain: str + file_type: str + summary_embedding: list[float] | None diff --git a/api/app/core/memory/models/message_models.py b/api/app/core/memory/models/message_models.py index 2f8660af..66203067 100644 --- a/api/app/core/memory/models/message_models.py +++ b/api/app/core/memory/models/message_models.py @@ -30,6 +30,7 @@ class ConversationMessage(BaseModel): """ role: str = Field(..., description="The role of the speaker (e.g., 'user', 'assistant').") msg: str = Field(..., description="The text content of the message.") + files: list[tuple] = Field(default_factory=list, description="The file content of the message", exclude=True) class TemporalValidityRange(BaseModel): @@ -130,7 +131,8 @@ class Chunk(BaseModel): content: str = Field(..., description="The content of the chunk as a string.") speaker: Optional[str] = Field(None, description="The speaker/role for this chunk (user/assistant).") statements: List[Statement] = Field(default_factory=list, description="A list of statements in the chunk.") - chunk_embedding: Optional[List[float]] = Field(None, description="The embedding vector of the chunk.") + files: list[tuple] = Field(default_factory=list, description="List of files in the chunk.") + chunk_embedding: Optional[List[float]] = Field(default=None, description="The embedding vector of the chunk.") metadata: Dict[str, Any] = Field(default_factory=dict, description="Additional metadata for the chunk.") @classmethod diff --git a/api/app/core/memory/src/search.py b/api/app/core/memory/src/search.py index 0e1d8424..e4f0d4d0 100644 --- a/api/app/core/memory/src/search.py +++ b/api/app/core/memory/src/search.py @@ -238,7 +238,7 @@ def rerank_with_activation( reranked: Dict[str, List[Dict[str, Any]]] = {} - for category in ["statements", "chunks", "entities", "summaries"]: + for category in ["statements", "chunks", "entities", "summaries", "communities"]: keyword_items = keyword_results.get(category, []) embedding_items = embedding_results.get(category, []) @@ -281,21 +281,23 @@ def rerank_with_activation( for item in items_list: item_id = item.get("id") or item.get("uuid") or item.get("chunk_id") if item_id and item_id in combined_items: - combined_items[item_id]["normalized_activation_value"] = item.get("normalized_activation_value", 0) + combined_items[item_id]["normalized_activation_value"] = item.get("normalized_activation_value") # 步骤 4: 计算基础分数和最终分数 for item_id, item in combined_items.items(): bm25_norm = float(item.get("bm25_score", 0) or 0) emb_norm = float(item.get("embedding_score", 0) or 0) - act_norm = float(item.get("normalized_activation_value", 0) or 0) + # normalized_activation_value 为 None 表示该节点无激活值,保留 None 语义 + raw_act_norm = item.get("normalized_activation_value") + act_norm = float(raw_act_norm) if raw_act_norm is not None else None # 第一阶段:只考虑内容相关性(BM25 + Embedding) # alpha 控制 BM25 权重,(1-alpha) 控制 Embedding 权重 content_score = alpha * bm25_norm + (1 - alpha) * emb_norm base_score = content_score # 第一阶段用内容分数 - # 存储激活度分数供第二阶段使用 - item["activation_score"] = act_norm + # 存储激活度分数供第二阶段使用(None 表示无激活值,不参与激活值排序) + item["activation_score"] = act_norm # 可能为 None item["content_score"] = content_score item["base_score"] = base_score @@ -724,6 +726,8 @@ async def run_hybrid_search( try: keyword_task = None embedding_task = None + keyword_results: Dict[str, List] = {} + embedding_results: Dict[str, List] = {} if search_type in ["keyword", "hybrid"]: # Keyword-based search @@ -746,35 +750,42 @@ async def run_hybrid_search( # 从数据库读取嵌入器配置(按 ID)并构建 RedBearModelConfig config_load_start = time.time() - with get_db_context() as db: - config_service = MemoryConfigService(db) - embedder_config_dict = config_service.get_embedder_config(str(memory_config.embedding_model_id)) - rb_config = RedBearModelConfig( - model_name=embedder_config_dict["model_name"], - provider=embedder_config_dict["provider"], - api_key=embedder_config_dict["api_key"], - base_url=embedder_config_dict["base_url"], - type="llm" - ) - config_load_time = time.time() - config_load_start - logger.info(f"[PERF] Config loading took {config_load_time:.4f}s") - - # Init embedder - embedder_init_start = time.time() - embedder = OpenAIEmbedderClient(model_config=rb_config) - embedder_init_time = time.time() - embedder_init_start - logger.info(f"[PERF] Embedder init took {embedder_init_time:.4f}s") - - embedding_task = asyncio.create_task( - search_graph_by_embedding( - connector=connector, - embedder_client=embedder, - query_text=query_text, - end_user_id=end_user_id, - limit=limit, - include=include, + try: + with get_db_context() as db: + config_service = MemoryConfigService(db) + embedder_config_dict = config_service.get_embedder_config(str(memory_config.embedding_model_id)) + rb_config = RedBearModelConfig( + model_name=embedder_config_dict["model_name"], + provider=embedder_config_dict["provider"], + api_key=embedder_config_dict["api_key"], + base_url=embedder_config_dict["base_url"], + type="llm" ) - ) + config_load_time = time.time() - config_load_start + logger.info(f"[PERF] Config loading took {config_load_time:.4f}s") + + # Init embedder + embedder_init_start = time.time() + embedder = OpenAIEmbedderClient(model_config=rb_config) + embedder_init_time = time.time() - embedder_init_start + logger.info(f"[PERF] Embedder init took {embedder_init_time:.4f}s") + + embedding_task = asyncio.create_task( + search_graph_by_embedding( + connector=connector, + embedder_client=embedder, + query_text=query_text, + end_user_id=end_user_id, + limit=limit, + include=include, + ) + ) + except Exception as emb_init_err: + logger.warning( + f"[PERF] Embedding search skipped due to init error " + f"(embedding_model_id={memory_config.embedding_model_id}): {emb_init_err}" + ) + embedding_task = None if keyword_task: keyword_results = await keyword_task diff --git a/api/app/core/memory/storage_services/clustering_engine/label_propagation.py b/api/app/core/memory/storage_services/clustering_engine/label_propagation.py index 58fd7f86..246453c0 100644 --- a/api/app/core/memory/storage_services/clustering_engine/label_propagation.py +++ b/api/app/core/memory/storage_services/clustering_engine/label_propagation.py @@ -7,6 +7,7 @@ - 增量更新(incremental_update):新实体到达时,只处理新实体及其邻居 """ +import asyncio import logging import uuid from math import sqrt @@ -19,8 +20,9 @@ logger = logging.getLogger(__name__) # 全量迭代最大轮数,防止不收敛 MAX_ITERATIONS = 10 -# 社区摘要核心实体数量 -CORE_ENTITY_LIMIT = 5 + +# 社区核心实体取 top-N 数量 +CORE_ENTITY_LIMIT = 10 def _cosine_similarity(v1: Optional[List[float]], v2: Optional[List[float]]) -> float: @@ -67,15 +69,16 @@ class LabelPropagationEngine: def __init__( self, connector: Neo4jConnector, - config_id: Optional[str] = None, llm_model_id: Optional[str] = None, embedding_model_id: Optional[str] = None, ): self.connector = connector self.repo = CommunityRepository(connector) - self.config_id = config_id self.llm_model_id = llm_model_id self.embedding_model_id = embedding_model_id + # 缓存客户端实例,避免重复初始化 + self._llm_client = None + self._embedder_client = None # ────────────────────────────────────────────────────────────────────────── # 公开接口 @@ -105,58 +108,81 @@ class LabelPropagationEngine: async def full_clustering(self, end_user_id: str) -> None: """ - 全量标签传播初始化。 + 全量标签传播初始化(分批处理,控制内存峰值)。 - 1. 拉取所有实体,初始化每个实体为独立社区 - 2. 迭代:每轮对所有实体做邻居投票,更新社区标签 - 3. 直到标签不再变化或达到 MAX_ITERATIONS - 4. 将最终标签写入 Neo4j + 策略: + - 每次只加载 BATCH_SIZE 个实体及其邻居进内存 + - labels 字典跨批次共享(只存 id→community_id,内存极小) + - 每批独立跑 MAX_ITERATIONS 轮 LPA,批次间通过 labels 传递社区信息 + - 所有批次完成后统一 flush 和 merge """ - entities = await self.repo.get_all_entities(end_user_id) - if not entities: + BATCH_SIZE = 888 # 每批实体数,可按需调整 + + # 轻量查询:只获取总数和 ID 列表,不加载 embedding 等大字段 + total_count = await self.repo.get_entity_count(end_user_id) + if not total_count: logger.info(f"[Clustering] 用户 {end_user_id} 无实体,跳过全量聚类") return - # 初始化:每个实体持有自己 id 作为社区标签 - labels: Dict[str, str] = {e["id"]: e["id"] for e in entities} - embeddings: Dict[str, Optional[List[float]]] = { - e["id"]: e.get("name_embedding") for e in entities - } + all_entity_ids = await self.repo.get_all_entity_ids(end_user_id) + logger.info(f"[Clustering] 用户 {end_user_id} 共 {total_count} 个实体," + f"分批大小 {BATCH_SIZE},共 {(total_count + BATCH_SIZE - 1) // BATCH_SIZE} 批") - # 预加载所有实体的邻居,避免迭代内 O(iterations * |E|) 次 Neo4j 往返 - logger.info(f"[Clustering] 预加载 {len(entities)} 个实体的邻居图...") - neighbors_cache: Dict[str, List[Dict]] = await self.repo.get_all_entity_neighbors_batch(end_user_id) - logger.info(f"[Clustering] 邻居预加载完成,覆盖实体数: {len(neighbors_cache)}") + # labels 跨批次共享:只存 id→community_id,内存极小 + labels: Dict[str, str] = {eid: eid for eid in all_entity_ids} + del all_entity_ids # 释放 ID 列表,后续按批次加载完整数据 - for iteration in range(MAX_ITERATIONS): - changed = 0 - # 随机顺序(Python dict 在 3.7+ 保持插入顺序,这里直接遍历) - for entity in entities: - eid = entity["id"] - # 直接从缓存取邻居,不再发起 Neo4j 查询 - neighbors = neighbors_cache.get(eid, []) - - # 将邻居的当前内存标签注入(覆盖 Neo4j 中的旧值) - enriched = [] - for nb in neighbors: - nb_copy = dict(nb) - nb_copy["community_id"] = labels.get(nb["id"], nb.get("community_id")) - enriched.append(nb_copy) - - new_label = _weighted_vote(enriched, embeddings.get(eid)) - if new_label and new_label != labels[eid]: - labels[eid] = new_label - changed += 1 - - logger.info( - f"[Clustering] 全量迭代 {iteration + 1}/{MAX_ITERATIONS}," - f"标签变化数: {changed}" + for batch_start in range(0, total_count, BATCH_SIZE): + batch_entities = await self.repo.get_entities_page( + end_user_id, skip=batch_start, limit=BATCH_SIZE ) - if changed == 0: - logger.info("[Clustering] 标签已收敛,提前结束迭代") + if not batch_entities: break - # 将最终标签写入 Neo4j + batch_ids = [e["id"] for e in batch_entities] + batch_embeddings: Dict[str, Optional[List[float]]] = { + e["id"]: e.get("name_embedding") for e in batch_entities + } + + logger.info( + f"[Clustering] 批次 {batch_start // BATCH_SIZE + 1}:" + f"加载 {len(batch_entities)} 个实体的邻居图..." + ) + neighbors_cache = await self.repo.get_entity_neighbors_for_ids( + batch_ids, end_user_id + ) + logger.info(f"[Clustering] 邻居预加载完成,覆盖实体数: {len(neighbors_cache)}") + + for iteration in range(MAX_ITERATIONS): + changed = 0 + for entity in batch_entities: + eid = entity["id"] + neighbors = neighbors_cache.get(eid, []) + + # 注入跨批次的最新标签(邻居可能在其他批次,labels 里有其最新值) + enriched = [] + for nb in neighbors: + nb_copy = dict(nb) + nb_copy["community_id"] = labels.get(nb["id"], nb.get("community_id")) + enriched.append(nb_copy) + + new_label = _weighted_vote(enriched, batch_embeddings.get(eid)) + if new_label and new_label != labels[eid]: + labels[eid] = new_label + changed += 1 + + logger.info( + f"[Clustering] 批次 {batch_start // BATCH_SIZE + 1} " + f"迭代 {iteration + 1}/{MAX_ITERATIONS},标签变化数: {changed}" + ) + if changed == 0: + logger.info("[Clustering] 标签已收敛,提前结束本批迭代") + break + + # 释放本批次的大对象 + del neighbors_cache, batch_embeddings, batch_entities + + # 所有批次完成,统一写入 Neo4j await self._flush_labels(labels, end_user_id) pre_merge_count = len(set(labels.values())) logger.info( @@ -164,7 +190,6 @@ class LabelPropagationEngine: f"{len(labels)} 个实体,开始后处理合并" ) - # 全量初始化后做一轮社区合并(基于 name_embedding 余弦相似度) all_community_ids = list(set(labels.values())) await self._evaluate_merge(all_community_ids, end_user_id) @@ -172,17 +197,15 @@ class LabelPropagationEngine: f"[Clustering] 全量聚类完成,合并前 {pre_merge_count} 个社区," f"{len(labels)} 个实体" ) - # 为所有社区生成元数据 - # 注意:_evaluate_merge 后部分社区已被合并消解,需重新从 Neo4j 查询实际存活的社区 - # 不能复用 labels.values(),那里包含已被 dissolve 的旧社区 ID + + # 查询存活社区并生成元数据 surviving_communities = await self.repo.get_all_entities(end_user_id) surviving_community_ids = list({ e.get("community_id") for e in surviving_communities if e.get("community_id") }) logger.info(f"[Clustering] 合并后实际存活社区数: {len(surviving_community_ids)}") - for cid in surviving_community_ids: - await self._generate_community_metadata(cid, end_user_id) + await self._generate_community_metadata(surviving_community_ids, end_user_id) async def incremental_update( self, new_entity_ids: List[str], end_user_id: str @@ -195,8 +218,17 @@ class LabelPropagationEngine: 3. 若邻居无社区 → 创建新社区 4. 若邻居分属多个社区 → 评估是否合并 """ + # 收集所有需要生成元数据的社区ID + communities_to_update = set() + for entity_id in new_entity_ids: - await self._process_single_entity(entity_id, end_user_id) + cid = await self._process_single_entity(entity_id, end_user_id) + if cid: + communities_to_update.add(cid) + + # 批量生成所有社区的元数据 + if communities_to_update: + await self._generate_community_metadata(list(communities_to_update), end_user_id, force=True) # ────────────────────────────────────────────────────────────────────────── # 内部方法 @@ -204,8 +236,21 @@ class LabelPropagationEngine: async def _process_single_entity( self, entity_id: str, end_user_id: str - ) -> None: - """处理单个新实体的社区分配。""" + ) -> Optional[str]: + """ + 处理单个新实体的社区分配。 + + 该函数会为新实体分配社区,可能的情况包括: + 1. 孤立实体(无邻居):创建新的单成员社区 + 2. 邻居都没有社区:创建新社区并将实体和邻居都加入 + 3. 邻居有社区:通过加权投票选择最合适的社区加入 + + Returns: + Optional[str]: 分配到的社区ID。当前实现总是返回一个有效的社区ID, + 但返回类型保留为Optional以支持未来可能的扩展场景 + (例如:实体无法分配到任何社区的情况)。 + 调用方应检查返回值的真假性(truthiness)。 + """ neighbors = await self.repo.get_entity_neighbors(entity_id, end_user_id) # 查询自身 embedding(从邻居查询结果中无法获取,需单独查) @@ -217,7 +262,7 @@ class LabelPropagationEngine: await self.repo.upsert_community(new_cid, end_user_id, member_count=1) await self.repo.assign_entity_to_community(entity_id, new_cid, end_user_id) logger.debug(f"[Clustering] 孤立实体 {entity_id} → 新社区 {new_cid}") - return + return new_cid # 统计邻居社区分布 community_ids_in_neighbors = set( @@ -239,7 +284,7 @@ class LabelPropagationEngine: logger.debug( f"[Clustering] 新实体 {entity_id} 与 {len(neighbors)} 个无社区邻居 → 新社区 {new_cid}" ) - await self._generate_community_metadata(new_cid, end_user_id) + return new_cid else: # 加入得票最多的社区 await self.repo.assign_entity_to_community(entity_id, target_cid, end_user_id) @@ -251,7 +296,8 @@ class LabelPropagationEngine: await self._evaluate_merge( list(community_ids_in_neighbors), end_user_id ) - await self._generate_community_metadata(target_cid, end_user_id) + # 返回目标社区ID,稍后批量生成元数据 + return target_cid async def _evaluate_merge( self, community_ids: List[str], end_user_id: str @@ -415,94 +461,223 @@ class LabelPropagationEngine: except Exception: return None + @staticmethod + def _build_entity_lines(members: List[Dict]) -> List[str]: + """将实体列表格式化为 prompt 行,包含 name、aliases、description、example。""" + lines = [] + for m in members: + m_name = m.get("name", "") + aliases = m.get("aliases") or [] + description = m.get("description") or "" + example = m.get("example") or "" + aliases_str = f"(别名:{'、'.join(aliases)})" if aliases else "" + desc_str = f":{description}" if description else "" + example_str = f"(示例:{example})" if example else "" + lines.append(f"- {m_name}{aliases_str}{desc_str}{example_str}") + return lines + async def _generate_community_metadata( - self, community_id: str, end_user_id: str + self, community_ids: List[str], end_user_id: str, force: bool = False ) -> None: """ - 为社区生成并写入元数据:名称、摘要、核心实体。 + 为一个或多个社区生成并写入元数据(优化版:批量 LLM 调用)。 - - core_entities:按 activation_value 排序取 top-N 实体名称列表(无需 LLM) - - name / summary:若有 llm_model_id 则调用 LLM 生成,否则用实体名称拼接兜底 + 流程: + 1. 批量准备所有社区的 prompt + 2. 并发调用 LLM 生成所有社区的 name / summary + 3. 批量 embed 所有 summary + 4. 批量写入数据库 + + Args: + force: 为 True 时跳过完整性检查,强制重新生成(用于增量更新成员变化后) """ - try: - # 先检查属性是否已完整,完整则跳过,避免重复生成 - check_embedding = bool(self.embedding_model_id) - if await self.repo.is_community_complete(community_id, end_user_id, check_embedding=check_embedding): - logger.debug(f"[Clustering] 社区 {community_id} 属性已完整,跳过生成") - return + async def _prepare_one(cid: str) -> Optional[Dict]: + """准备单个社区的数据和 prompt""" + try: + if not force: + check_embedding = bool(self.embedding_model_id) + if await self.repo.is_community_complete(cid, end_user_id, check_embedding=check_embedding): + return None - members = await self.repo.get_community_members(community_id, end_user_id) - if not members: - return + members = await self.repo.get_community_members(cid, end_user_id) + if not members: + logger.warning(f"[Clustering] 社区 {cid} 无成员,跳过元数据生成") + return None - # 核心实体:按 activation_value 降序取 top-N - sorted_members = sorted( - members, - key=lambda m: m.get("activation_value") or 0, - reverse=True, - ) - core_entities = [m["name"] for m in sorted_members[:CORE_ENTITY_LIMIT] if m.get("name")] - all_names = [m["name"] for m in members if m.get("name")] + sorted_members = sorted( + members, + key=lambda m: m.get("activation_value") or 0, + reverse=True, + ) + core_entities = [m["name"] for m in sorted_members[:CORE_ENTITY_LIMIT] if m.get("name")] + all_names = [m["name"] for m in members if m.get("name")] - name = "、".join(core_entities[:3]) if core_entities else community_id[:8] - summary = f"包含实体:{', '.join(all_names)}" + # 默认值 + name = "、".join(core_entities[:3]) if core_entities else cid[:8] + summary = f"包含实体:{', '.join(all_names)}" - # 若有 LLM 配置,调用 LLM 生成更好的名称和摘要 - if self.llm_model_id: - try: - from app.db import get_db_context - from app.core.memory.utils.llm.llm_utils import MemoryClientFactory - - entity_list_str = "、".join(all_names) + # 准备 LLM prompt(如果配置了 LLM) + prompt = None + if self.llm_model_id: + entity_list_str = "\n".join(self._build_entity_lines(members)) + relationships = await self.repo.get_community_relationships(cid, end_user_id) + rel_lines = [ + f"- {r['subject']} → {r['predicate']} → {r['object']}" + for r in relationships + if r.get("subject") and r.get("predicate") and r.get("object") + ] + rel_section = ( + f"\n实体间关系:\n" + "\n".join(rel_lines) + if rel_lines else "" + ) prompt = ( - f"以下是一组语义相关的实体:{entity_list_str}\n\n" + f"以下是一组语义相关的实体:\n{entity_list_str}{rel_section}\n\n" f"请为这组实体所代表的主题:\n" f"1. 起一个简洁的中文名称(不超过10个字)\n" - f"2. 写一句话摘要(不超过50个字)\n\n" + f"2. 写一句话摘要(不超过80个字)\n\n" f"严格按以下格式输出,不要有其他内容:\n" f"名称:<名称>\n摘要:<摘要>" ) - with get_db_context() as db: - factory = MemoryClientFactory(db) - llm_client = factory.get_llm_client(self.llm_model_id) - response = await llm_client.chat([{"role": "user", "content": prompt}]) - text = response.content if hasattr(response, "content") else str(response) - for line in text.strip().splitlines(): - if line.startswith("名称:"): - name = line[3:].strip() - elif line.startswith("摘要:"): - summary = line[3:].strip() - except Exception as e: - logger.warning(f"[Clustering] LLM 生成社区元数据失败,使用兜底值: {e}") + return { + "community_id": cid, + "end_user_id": end_user_id, + "name": name, + "summary": summary, + "core_entities": core_entities, + "prompt": prompt, + "summary_embedding": None, + } + except Exception as e: + logger.error(f"[Clustering] 社区 {cid} 元数据准备失败: {e}", exc_info=True) + return None - # 生成 summary_embedding - summary_embedding: Optional[List[float]] = None - if self.embedding_model_id and summary: + # --- 阶段1:并发准备所有社区数据 --- + results = await asyncio.gather( + *[_prepare_one(cid) for cid in community_ids], + return_exceptions=True, + ) + metadata_list = [] + for cid, res in zip(community_ids, results): + if isinstance(res, Exception): + logger.error(f"[Clustering] 社区 {cid} 元数据准备失败: {res}", exc_info=res) + elif res is not None: + metadata_list.append(res) + + if not metadata_list: + logger.warning(f"[Clustering] 无有效元数据可写入,community_ids={community_ids}") + return + + # --- 阶段2:批量调用 LLM 生成 name 和 summary --- + if self.llm_model_id: + llm_client = self._get_llm_client() + if not llm_client: + logger.warning( + f"[Clustering] LLM 已配置(model_id={self.llm_model_id})但客户端初始化失败," + f"将跳过社区元数据的 LLM 富化。请检查 model_id 是否正确或数据库连接是否正常。" + ) + if llm_client: + prompts_to_process = [(i, m) for i, m in enumerate(metadata_list) if m.get("prompt")] + + if prompts_to_process: + logger.info(f"[Clustering] 批量调用 LLM 生成 {len(prompts_to_process)} 个社区元数据") + + async def _call_llm(idx: int, meta: Dict) -> tuple: + """单个 LLM 调用""" + try: + response = await llm_client.chat([{"role": "user", "content": meta["prompt"]}]) + text = response.content if hasattr(response, "content") else str(response) + return (idx, text, None) + except Exception as e: + logger.warning(f"[Clustering] 社区 {meta['community_id']} LLM 生成失败: {e}") + return (idx, None, e) + + # 并发调用所有 LLM 请求 + llm_results = await asyncio.gather( + *[_call_llm(idx, meta) for idx, meta in prompts_to_process], + return_exceptions=True + ) + + # 解析 LLM 响应 + for result in llm_results: + if isinstance(result, Exception): + continue + idx, text, error = result + if error or not text: + continue + + meta = metadata_list[idx] + for line in text.strip().splitlines(): + if line.startswith("名称:"): + meta["name"] = line[3:].strip() + elif line.startswith("摘要:"): + meta["summary"] = line[3:].strip() + + logger.info(f"[Clustering] LLM 批量生成完成") + + # --- 阶段3:批量生成 summary_embedding --- + if self.embedding_model_id: + embedder = self._get_embedder_client() + if not embedder: + logger.warning( + f"[Clustering] Embedding 已配置(model_id={self.embedding_model_id})但客户端初始化失败," + f"将跳过社区摘要的向量化。请检查 model_id 是否正确或数据库连接是否正常。" + ) + if embedder: try: - from app.db import get_db_context - from app.core.memory.utils.llm.llm_utils import MemoryClientFactory - - with get_db_context() as db: - embedder = MemoryClientFactory(db).get_embedder_client(self.embedding_model_id) - vectors = await embedder.response([summary]) - if vectors: - summary_embedding = vectors[0] + summaries = [m["summary"] for m in metadata_list] + logger.info(f"[Clustering] 批量生成 {len(summaries)} 个 summary embedding") + embeddings = await embedder.response(summaries) + for i, meta in enumerate(metadata_list): + meta["summary_embedding"] = embeddings[i] if i < len(embeddings) else None + logger.info(f"[Clustering] Embedding 批量生成完成") except Exception as e: - logger.warning(f"[Clustering] 社区 {community_id} 生成 summary_embedding 失败: {e}") + logger.error(f"[Clustering] 批量生成 summary_embedding 失败: {e}", exc_info=True) - await self.repo.update_community_metadata( - community_id=community_id, - end_user_id=end_user_id, - name=name, - summary=summary, - core_entities=core_entities, - summary_embedding=summary_embedding, + # --- 阶段4:批量写入数据库 --- + # 移除 prompt 字段(不需要存储) + for m in metadata_list: + m.pop("prompt", None) + + if len(metadata_list) == 1: + m = metadata_list[0] + result = await self.repo.update_community_metadata( + community_id=m["community_id"], + end_user_id=m["end_user_id"], + name=m["name"], + summary=m["summary"], + core_entities=m["core_entities"], + summary_embedding=m["summary_embedding"], ) - logger.debug(f"[Clustering] 社区 {community_id} 元数据已更新: name={name}") - except Exception as e: - logger.error(f"[Clustering] _generate_community_metadata failed for {community_id}: {e}") + if not result: + logger.error(f"[Clustering] 社区 {m['community_id']} 元数据写入失败") + else: + ok = await self.repo.batch_update_community_metadata(metadata_list) + if not ok: + logger.error(f"[Clustering] 批量写入 {len(metadata_list)} 个社区元数据失败") + else: + logger.info(f"[Clustering] 批量写入 {len(metadata_list)} 个社区元数据成功") + + def _get_llm_client(self): + """获取或创建 LLM 客户端(单例模式)""" + if self._llm_client is None and self.llm_model_id: + from app.db import get_db_context + from app.core.memory.utils.llm.llm_utils import MemoryClientFactory + with get_db_context() as db: + self._llm_client = MemoryClientFactory(db).get_llm_client(self.llm_model_id) + logger.info(f"[Clustering] LLM 客户端初始化完成(单例): model_id={self.llm_model_id}") + return self._llm_client + + def _get_embedder_client(self): + """获取或创建 Embedder 客户端(单例模式)""" + if self._embedder_client is None and self.embedding_model_id: + from app.db import get_db_context + from app.core.memory.utils.llm.llm_utils import MemoryClientFactory + with get_db_context() as db: + self._embedder_client = MemoryClientFactory(db).get_embedder_client(self.embedding_model_id) + logger.info(f"[Clustering] Embedder 客户端初始化完成(单例): model_id={self.embedding_model_id}") + return self._embedder_client @staticmethod def _new_community_id() -> str: - return str(uuid.uuid4()) + return str(uuid.uuid4()) \ No newline at end of file diff --git a/api/app/core/memory/storage_services/extraction_engine/data_preprocessing/data_pruning.py b/api/app/core/memory/storage_services/extraction_engine/data_preprocessing/data_pruning.py index 28f7d8e0..5390197a 100644 --- a/api/app/core/memory/storage_services/extraction_engine/data_preprocessing/data_pruning.py +++ b/api/app/core/memory/storage_services/extraction_engine/data_preprocessing/data_pruning.py @@ -9,6 +9,7 @@ """ import asyncio +import logging import os import hashlib import json @@ -20,13 +21,26 @@ from pydantic import BaseModel, Field from app.core.memory.models.message_models import DialogData, ConversationMessage, ConversationContext from app.core.memory.models.config_models import PruningConfig -from app.core.memory.utils.config.config_utils import get_pruning_config from app.core.memory.utils.prompt.prompt_utils import prompt_env, log_prompt_rendering, log_template_rendering from app.core.memory.storage_services.extraction_engine.data_preprocessing.scene_config import ( SceneConfigRegistry, ScenePatterns ) +logger = logging.getLogger(__name__) + + +def message_has_files(message: "ConversationMessage") -> bool: + """检查消息是否包含文件。 + + Args: + message: 待检查的消息对象 + + Returns: + bool: 如果消息包含文件则返回 True,否则返回 False + """ + return message.files and len(message.files) > 0 + class DialogExtractionResponse(BaseModel): """对话级一次性抽取的结构化返回,用于加速剪枝。 @@ -34,6 +48,8 @@ class DialogExtractionResponse(BaseModel): - is_related:对话与场景的相关性判定。 - times / ids / amounts / contacts / addresses / keywords:重要信息片段,用来在不相关对话中保留关键消息。 - preserve_keywords:情绪/兴趣/爱好/个人观点相关词,包含这些词的消息必须强制保留。 + - scene_unrelated_snippets:与当前场景无关且无语义关联的消息片段(原文截取), + 用于高阈值阶段精准删除跨场景内容。 """ is_related: bool = Field(...) times: List[str] = Field(default_factory=list) @@ -43,6 +59,7 @@ class DialogExtractionResponse(BaseModel): addresses: List[str] = Field(default_factory=list) keywords: List[str] = Field(default_factory=list) preserve_keywords: List[str] = Field(default_factory=list, description="情绪/兴趣/爱好/个人观点相关词,包含这些词的消息强制保留") + scene_unrelated_snippets: List[str] = Field(default_factory=list,description="与当前场景无关且无语义关联的消息原文片段,高阈值阶段用于精准删除跨场景内容") class MessageImportanceResponse(BaseModel): @@ -91,12 +108,14 @@ class SemanticPruner: # 加载统一填充词库 self.scene_config: ScenePatterns = SceneConfigRegistry.get_config(self.config.pruning_scene) - # 本体类型列表(用于注入提示词,所有场景均支持) - self._ontology_classes = getattr(self.config, "ontology_classes", None) or [] + # 本体类型列表:直接使用 ontology_class_infos(name + description) + self._ontology_class_infos = getattr(self.config, "ontology_class_infos", None) or [] + # _ontology_classes 仅用于日志统计 + self._ontology_classes = [info.class_name for info in self._ontology_class_infos] self._log(f"[剪枝-初始化] 场景={self.config.pruning_scene}") - if self._ontology_classes: - self._log(f"[剪枝-初始化] 注入本体类型: {self._ontology_classes}") + if self._ontology_class_infos: + self._log(f"[剪枝-初始化] 注入本体类型({len(self._ontology_class_infos)}个): {self._ontology_classes}") else: self._log(f"[剪枝-初始化] 未找到本体类型,将使用通用提示词") @@ -121,7 +140,8 @@ class SemanticPruner: 1. 空消息 2. 场景特定填充词库精确匹配 3. 常见寒暄精确匹配 - 4. 纯表情/标点 + 4. 组合寒暄模式(前缀 + 后缀组合,如"好的谢谢"、"同学你好"、"明白了") + 5. 纯表情/标点 """ t = message.msg.strip() if not t: @@ -143,6 +163,55 @@ class SemanticPruner: if t in common_greetings: return True + # 组合寒暄模式:短消息(≤15字)且完全由寒暄成分构成 + # 策略:将消息拆分后,每个片段都能在填充词库或常见寒暄中找到,则整体为填充 + if len(t) <= 15: + # 确认+称呼/感谢组合,如"好的谢谢"、"明白了"、"知道了谢谢" + _confirm_prefixes = {"好的", "好", "嗯", "嗯嗯", "哦", "明白", "明白了", "知道了", "了解", "收到", "没问题"} + _thanks_suffixes = {"谢谢", "谢谢你", "谢谢您", "多谢", "感谢", "谢了"} + _greeting_suffixes = {"你好", "您好", "老师好", "同学好", "大家好"} + _greeting_prefixes = {"同学", "老师", "您好", "你好"} + _close_patterns = { + "没有了", "没事了", "没问题了", "好了", "行了", "可以了", + "不用了", "不需要了", "就这样", "就这样吧", "那就这样", + } + _polite_responses = { + "不客气", "不用谢", "没关系", "没事", "应该的", "这是我应该做的", + } + + # 规则1:确认词 + 感谢词(如"好的谢谢"、"嗯谢谢") + for cp in _confirm_prefixes: + for ts in _thanks_suffixes: + if t == cp + ts or t == cp + "," + ts or t == cp + "," + ts: + return True + + # 规则2:称呼前缀 + 问候(如"同学你好"、"老师好") + for gp in _greeting_prefixes: + for gs in _greeting_suffixes: + if t == gp + gs or t.startswith(gp) and t.endswith("好"): + return True + + # 规则3:结束语 + 感谢(如"没有了,谢谢老师"、"没有了谢谢") + for cp in _close_patterns: + if t.startswith(cp): + remainder = t[len(cp):].lstrip(",,、 ") + if not remainder or any(remainder.startswith(ts) for ts in _thanks_suffixes): + return True + + # 规则4:礼貌回应(如"不客气,祝你考试顺利"——前缀是礼貌词,后半是祝福套话) + for pr in _polite_responses: + if t.startswith(pr): + remainder = t[len(pr):].lstrip(",,、 ") + # 后半是祝福/套话(不含实质信息) + if not remainder or re.match(r"^(祝|希望|期待|加油|顺利|好好|保重)", remainder): + return True + + # 规则5:纯确认词加"了"后缀(如"明白了"、"知道了"、"好了") + _confirm_base = {"明白", "知道", "了解", "收到", "好", "行", "可以", "没问题"} + for cb in _confirm_base: + if t == cb + "了" or t == cb + "了。" or t == cb + "了!": + return True + # 检查是否为纯表情符号(方括号包裹) if re.fullmatch(r"(\[[^\]]+\])+", t): return True @@ -331,13 +400,13 @@ class SemanticPruner: rendered = self.template.render( pruning_scene=self.config.pruning_scene, - ontology_classes=self._ontology_classes, + ontology_class_infos=self._ontology_class_infos, dialog_text=dialog_text, language=self.language ) log_template_rendering("extracat_Pruning.jinja2", { "pruning_scene": self.config.pruning_scene, - "ontology_classes_count": len(self._ontology_classes), + "ontology_class_infos_count": len(self._ontology_class_infos), "language": self.language }) log_prompt_rendering("pruning-extract", rendered) @@ -377,6 +446,193 @@ class SemanticPruner: ) return fallback_response + def _get_pruning_mode(self) -> str: + """根据 pruning_threshold 返回当前剪枝阶段。 + + - 低阈值 [0.0, 0.3):conservative 只删填充,保留所有实质内容 + - 中阈值 [0.3, 0.6):semantic 保留场景相关 + 有语义关联的内容,删除无关联内容 + - 高阈值 [0.6, 0.9]:strict 只保留场景相关内容,跨场景内容可被删除 + """ + t = float(self.config.pruning_threshold) + if t < 0.3: + return "conservative" + elif t < 0.6: + return "semantic" + else: + return "strict" + + def _apply_related_dialog_pruning( + self, + msgs: List[ConversationMessage], + extraction: "DialogExtractionResponse", + dialog_label: str, + pruning_mode: str, + ) -> List[ConversationMessage]: + """相关对话统一剪枝入口,消除 prune_dialog / prune_dataset 中的重复逻辑。 + + - conservative:只删填充 + - semantic / strict:场景感知剪枝 + """ + if pruning_mode == "conservative": + preserve_tokens = self._build_preserve_tokens(extraction) + return self._prune_fillers_only(msgs, preserve_tokens, dialog_label) + else: + return self._prune_with_scene_filter(msgs, extraction, dialog_label, pruning_mode) + + def _prune_fillers_only( + self, + msgs: List[ConversationMessage], + preserve_tokens: List[str], + dialog_label: str, + ) -> List[ConversationMessage]: + """相关对话专用:只删填充消息,LLM 保护消息和实质内容一律保留。 + + 不受 pruning_threshold 约束,删多少算多少(填充有多少删多少)。 + 至少保留 1 条消息。 + 注意:填充检测优先于 preserve_tokens 保护——填充消息本身无信息价值, + 即使 LLM 误将其关键词放入 preserve_tokens 也应删除。 + """ + to_delete_ids: set = set() + for m in msgs: + # 最高优先级保护:带有文件的消息一律保留,不参与任何剪枝判断 + if message_has_files(m): + self._log(f" [保护] 带文件的消息(不参与剪枝):'{m.msg[:40]}',文件数={len(m.files)}") + continue + + # 填充检测优先:先判断是否为填充,再看 LLM 保护 + if self._is_filler_message(m): + to_delete_ids.add(id(m)) + self._log(f" [填充] '{m.msg[:40]}' → 删除") + continue + if self._msg_matches_tokens(m, preserve_tokens): + self._log(f" [保护] '{m.msg[:40]}' → LLM保护,跳过") + + kept = [m for m in msgs if id(m) not in to_delete_ids] + if not kept and msgs: + kept = [msgs[0]] + + deleted = len(msgs) - len(kept) + self._log( + f"[剪枝-相关] {dialog_label} 总消息={len(msgs)} " + f"填充删除={deleted} 保留={len(kept)}" + ) + return kept + + def _prune_with_scene_filter( + self, + msgs: List[ConversationMessage], + extraction: "DialogExtractionResponse", + dialog_label: str, + mode: str, + ) -> List[ConversationMessage]: + """场景感知剪枝,供 semantic / strict 两个阈值档位调用。 + + 本函数体现剪枝系统的三层递进逻辑: + + 第一层(conservative,阈值 < 0.3): + 不进入本函数,由 _prune_fillers_only 处理。 + 保留标准:只问"有没有信息量",填充消息(嗯/好的/哈哈等)删除,其余一律保留。 + + 第二层(semantic,阈值 [0.3, 0.6)): + 保留标准:内容价值优先,场景相关性是参考而非唯一标准。 + - 填充消息 → 删除(最高优先级) + - 场景相关消息 → 保留 + - 场景无关消息 → 有两次豁免机会: + 1. 命中 scene_preserve_tokens(LLM 标记的关键词/时间/金额等)→ 保留 + 2. 含情感词(感觉/压力/开心等)→ 保留(情感内容有记忆价值) + 3. 两次豁免均未命中 → 删除 + + 第三层(strict,阈值 [0.6, 0.9]): + 保留标准:场景相关性优先,无任何豁免。 + - 填充消息 → 删除(最高优先级) + - 场景相关消息 → 保留 + - 场景无关消息 → 直接删除,preserve_keywords 和情感词在此模式下均不生效 + + 至少保留 1 条消息(兜底取第一条)。 + """ + # strict 模式收窄保护范围:只保护结构化关键信息(时间/编号/金额/联系方式/地址), + # 不保护 keywords / preserve_keywords,让场景过滤能删掉更多内容。 + # semantic 模式完整保护:包含 LLM 抽取的所有重要片段(含 keywords 和 preserve_keywords)。 + if mode == "strict": + scene_preserve_tokens = ( + extraction.times + extraction.ids + extraction.amounts + + extraction.contacts + extraction.addresses + ) + else: + scene_preserve_tokens = self._build_preserve_tokens(extraction) + + unrelated_snippets = extraction.scene_unrelated_snippets or [] + + to_delete_ids: set = set() + for m in msgs: + msg_text = m.msg.strip() + + # 最高优先级保护:带有文件的消息一律保留,不参与任何剪枝判断 + if message_has_files(m): + self._log(f" [保护] 带文件的消息(不参与剪枝):'{msg_text[:40]}',文件数={len(m.files)}") + continue + + # 第一优先级:填充消息无论模式直接删除,不参与后续场景判断 + if self._is_filler_message(m): + to_delete_ids.add(id(m)) + self._log(f" [填充] '{msg_text[:40]}' → 删除") + continue + + # 双向包含匹配:处理 LLM 返回片段与原始消息文本长度不完全一致的情况 + is_scene_unrelated = any( + snip and (snip in msg_text or msg_text in snip) + for snip in unrelated_snippets + ) + + if is_scene_unrelated: + if mode == "strict": + # strict:场景无关直接删除,不做任何豁免 + # 场景相关性是唯一裁决标准,preserve_keywords 在此模式下不生效 + to_delete_ids.add(id(m)) + self._log(f" [场景无关-严格] '{msg_text[:40]}' → 删除") + elif mode == "semantic": + # semantic:场景无关但有内容价值 → 保留 + # 豁免第一层:命中 scene_preserve_tokens(关键词/结构化信息保护) + if self._msg_matches_tokens(m, scene_preserve_tokens): + self._log(f" [保护] '{msg_text[:40]}' → 场景关键词保护,保留") + else: + # 豁免第二层:含情感词,认为有情境记忆价值,即使场景无关也保留 + has_contextual_emotion = any( + word in msg_text + for word in ["感觉", "觉得", "心情", "开心", "难过", "高兴", "沮丧", + "喜欢", "讨厌", "爱", "恨", "担心", "害怕", "兴奋", + "压力", "累", "疲惫", "烦", "焦虑", "委屈", "感动"] + ) + if not has_contextual_emotion: + to_delete_ids.add(id(m)) + self._log(f" [场景无关-语义] '{msg_text[:40]}' → 删除(无情感关联)") + else: + self._log(f" [场景关联-保留] '{msg_text[:40]}' → 有情感关联,保留") + else: + # 不在 scene_unrelated_snippets 中 → 场景相关,直接保留 + if self._msg_matches_tokens(m, scene_preserve_tokens): + self._log(f" [保护] '{msg_text[:40]}' → LLM保护,跳过") + # else: 普通场景相关消息,保留,不输出日志 + + kept = [m for m in msgs if id(m) not in to_delete_ids] + if not kept and msgs: + kept = [msgs[0]] + + deleted = len(msgs) - len(kept) + self._log( + f"[剪枝-{mode}] {dialog_label} 总消息={len(msgs)} " + f"删除={deleted} 保留={len(kept)}" + ) + return kept + + def _build_preserve_tokens(self, extraction: "DialogExtractionResponse") -> List[str]: + """统一构建 preserve_tokens,合并 LLM 抽取的所有重要片段。""" + return ( + extraction.times + extraction.ids + extraction.amounts + + extraction.contacts + extraction.addresses + extraction.keywords + + extraction.preserve_keywords + ) + def _msg_matches_tokens(self, message: ConversationMessage, tokens: List[str]) -> bool: """判断消息是否包含任意抽取到的重要片段。""" if not tokens: @@ -397,16 +653,18 @@ class SemanticPruner: proportion = float(self.config.pruning_threshold) extraction = await self._extract_dialog_important(dialog.content) + pruning_mode = self._get_pruning_mode() + self._log(f"[剪枝-模式] 阈值={proportion} → 模式={pruning_mode}") + if extraction.is_related: - # 相关对话不剪枝 + kept = self._apply_related_dialog_pruning( + dialog.context.msgs, extraction, f"对话ID={dialog.id}", pruning_mode + ) + dialog.context = ConversationContext(msgs=kept) return dialog # 在不相关对话中,LLM 已通过 preserve_tokens 标记需要保护的内容 - preserve_tokens = ( - extraction.times + extraction.ids + extraction.amounts + - extraction.contacts + extraction.addresses + extraction.keywords + - extraction.preserve_keywords - ) + preserve_tokens = self._build_preserve_tokens(extraction) msgs = dialog.context.msgs # 分类:填充 / 其他可删(LLM保护消息通过不加入任何桶来隐式保护) @@ -473,7 +731,7 @@ class SemanticPruner: # 阈值保护:最高0.9 proportion = float(self.config.pruning_threshold) if proportion > 0.9: - print(f"[剪枝-数据集] 阈值{proportion}超过上限0.9,已自动调整为0.9") + logger.warning(f"[剪枝-数据集] 阈值{proportion}超过上限0.9,已自动调整为0.9") proportion = 0.9 if proportion < 0.0: proportion = 0.0 @@ -481,11 +739,30 @@ class SemanticPruner: self._log( f"[剪枝-数据集] 对话总数={len(dialogs)} 场景={self.config.pruning_scene} 删除比例={proportion} 开关={self.config.pruning_switch} 模式=消息级独立判断" ) - + + pruning_mode = self._get_pruning_mode() + self._log(f"[剪枝-数据集] 阈值={proportion} → 剪枝阶段={pruning_mode}") + result: List[DialogData] = [] total_original_msgs = 0 total_deleted_msgs = 0 + # 统计对象:直接收集结构化数据,无需事后正则解析 + stats = { + "scene": self.config.pruning_scene, + "dialog_total": len(dialogs), + "deletion_ratio": proportion, + "enabled": self.config.pruning_switch, + "pruning_mode": pruning_mode, + "related_count": 0, + "unrelated_count": 0, + "related_indices": [], + "unrelated_indices": [], + "total_deleted_messages": 0, + "remaining_dialogs": 0, + "dialogs": [], + } + # 并发执行所有对话的 LLM 抽取(获取 preserve_keywords 等保护信息) semaphore = asyncio.Semaphore(self.max_concurrent) @@ -505,12 +782,31 @@ class SemanticPruner: original_count = len(msgs) total_original_msgs += original_count + # 相关对话:根据阶段决定处理力度 + if extraction.is_related: + stats["related_count"] += 1 + stats["related_indices"].append(d_idx + 1) + kept = self._apply_related_dialog_pruning( + msgs, extraction, f"对话 {d_idx+1}", pruning_mode + ) + deleted_count = original_count - len(kept) + total_deleted_msgs += deleted_count + dd.context.msgs = kept + result.append(dd) + stats["dialogs"].append({ + "index": d_idx + 1, + "is_related": True, + "total_messages": original_count, + "deleted": deleted_count, + "kept": len(kept), + }) + continue + + stats["unrelated_count"] += 1 + stats["unrelated_indices"].append(d_idx + 1) + # 从 LLM 抽取结果中获取所有需要保留的 token - preserve_tokens = ( - extraction.times + extraction.ids + extraction.amounts + - extraction.contacts + extraction.addresses + extraction.keywords + - extraction.preserve_keywords # 情绪/兴趣/爱好关键词 - ) + preserve_tokens = self._build_preserve_tokens(extraction) # 判断是否需要详细日志 should_log_details = self._detailed_prune_logging and original_count <= self._max_debug_msgs_per_dialog @@ -527,6 +823,12 @@ class SemanticPruner: for idx, m in enumerate(msgs): msg_text = m.msg.strip() + + # 最高优先级保护:带有文件的消息一律保留,不参与分类 + if message_has_files(m): + self._log(f" [保护] 带文件的消息(不参与分类,直接保留):索引{idx}, '{msg_text[:40]}', 文件数={len(m.files)}") + llm_protected_msgs.append((idx, m)) # 放入保护列表 + continue if self._msg_matches_tokens(m, preserve_tokens): llm_protected_msgs.append((idx, m)) @@ -543,16 +845,16 @@ class SemanticPruner: # important_msgs 仅用于日志统计 important_msgs = llm_protected_msgs - + # 计算删除配额 delete_target = int(original_count * proportion) if proportion > 0 and original_count > 0 and delete_target == 0: delete_target = 1 - + # 确保至少保留1条消息 max_deletable = max(0, original_count - 1) delete_target = min(delete_target, max_deletable) - + # 删除策略:优先删填充消息,再按出现顺序删其余可删消息 to_delete_indices = set() deleted_details = [] @@ -570,58 +872,73 @@ class SemanticPruner: break to_delete_indices.add(idx) deleted_details.append(f"[{idx}] 可删: '{msg.msg[:50]}'") - + # 执行删除 kept_msgs = [] for idx, m in enumerate(msgs): if idx not in to_delete_indices: kept_msgs.append(m) - + # 确保至少保留1条 if not kept_msgs and msgs: kept_msgs = [msgs[0]] - + dd.context.msgs = kept_msgs deleted_count = original_count - len(kept_msgs) total_deleted_msgs += deleted_count - + # 输出删除详情 if deleted_details: self._log(f"[剪枝-删除详情] 对话 {d_idx+1} 删除了以下消息:") for detail in deleted_details: self._log(f" {detail}") - + # ========== 问答对统计(已注释) ========== # qa_info = f",问答对={len(qa_pairs)}" if qa_pairs else "" # ======================================== - + self._log( f"[剪枝-对话] 对话 {d_idx+1} 总消息={original_count} " f"(保护={len(important_msgs)} 填充={len(filler_msgs)} 可删={len(deletable_msgs)}) " f"删除={deleted_count} 保留={len(kept_msgs)}" ) - - result.append(dd) - - self._log(f"[剪枝-数据集] 剩余对话数={len(result)}") - # 保存日志 + stats["dialogs"].append({ + "index": d_idx + 1, + "is_related": False, + "total_messages": original_count, + "protected": len(important_msgs), + "fillers": len(filler_msgs), + "deletable": len(deletable_msgs), + "deleted": deleted_count, + "kept": len(kept_msgs), + }) + + result.append(dd) + + # 补全统计对象 + stats["total_deleted_messages"] = total_deleted_msgs + stats["remaining_dialogs"] = len(result) + + self._log(f"[剪枝-数据集] 剩余对话数={len(result)}") + self._log(f"[剪枝-数据集] 相关对话数={stats['related_count']} 不相关对话数={stats['unrelated_count']}") + self._log(f"[剪枝-数据集] 总删除 {total_deleted_msgs} 条") + + # 直接序列化统计对象,无需正则解析 try: from app.core.config import settings settings.ensure_memory_output_dir() log_output_path = settings.get_memory_output_path("pruned_terminal.json") - sanitized_logs = [self._sanitize_log_line(l) for l in self.run_logs] - payload = self._parse_logs_to_structured(sanitized_logs) with open(log_output_path, "w", encoding="utf-8") as f: - json.dump(payload, f, ensure_ascii=False, indent=2) + json.dump(stats, f, ensure_ascii=False, indent=2) except Exception as e: self._log(f"[剪枝-数据集] 保存终端输出日志失败:{e}") # Safety: avoid empty dataset if not result: - print("警告: 语义剪枝后数据集为空,已回退为未剪枝数据以避免流程中断") + logger.warning("语义剪枝后数据集为空,已回退为未剪枝数据以避免流程中断") return dialogs - + return result def _log(self, msg: str) -> None: @@ -629,118 +946,7 @@ class SemanticPruner: try: self.run_logs.append(msg) except Exception: - # 任何异常都不影响打印 pass - print(msg) + logger.debug(msg) - def _sanitize_log_line(self, line: str) -> str: - """移除行首的方括号标签前缀,例如 [剪枝-数据集] 或 [剪枝-对话]。""" - try: - return re.sub(r"^\[[^\]]+\]\s*", "", line) - except Exception: - return line - def _parse_logs_to_structured(self, logs: List[str]) -> dict: - """将已去前缀的日志列表解析为结构化 JSON,便于数据对接。""" - summary = { - "scene": self.config.pruning_scene, - "dialog_total": None, - "deletion_ratio": None, - "enabled": None, - "related_count": None, - "unrelated_count": None, - "related_indices": [], - "unrelated_indices": [], - "total_deleted_messages": None, - "remaining_dialogs": None, - } - dialogs = [] - - # 解析函数 - def parse_int(value: str) -> Optional[int]: - try: - return int(value) - except Exception: - return None - - def parse_float(value: str) -> Optional[float]: - try: - return float(value) - except Exception: - return None - - def parse_indices(s: str) -> List[int]: - s = s.strip() - if not s: - return [] - parts = [p.strip() for p in s.split(",") if p.strip()] - out: List[int] = [] - for p in parts: - try: - out.append(int(p)) - except Exception: - pass - return out - - # 正则 - re_header = re.compile(r"对话总数=(\d+)\s+场景=([^\s]+)\s+删除比例=([0-9.]+)\s+开关=(True|False)") - re_counts = re.compile(r"相关对话数=(\d+)\s+不相关对话数=(\d+)") - re_indices = re.compile(r"相关对话:第\[(.*?)\]段;不相关对话:第\[(.*?)\]段") - re_dialog = re.compile(r"对话\s+(\d+)\s+总消息=(\d+)\s+分配删除=(\d+)\s+实删=(\d+)\s+保留=(\d+)") - re_total_del = re.compile(r"总删除\s+(\d+)\s+条") - re_remaining = re.compile(r"剩余对话数=(\d+)") - - for line in logs: - # 第一行:总览 - m = re_header.search(line) - if m: - summary["dialog_total"] = parse_int(m.group(1)) - # 顶层 scene 依配置,这里不覆盖,但也可校验 m.group(2) - summary["deletion_ratio"] = parse_float(m.group(3)) - summary["enabled"] = True if m.group(4) == "True" else False - continue - - # 第二行:相关/不相关数量 - m = re_counts.search(line) - if m: - summary["related_count"] = parse_int(m.group(1)) - summary["unrelated_count"] = parse_int(m.group(2)) - continue - - # 第三行:相关/不相关索引 - m = re_indices.search(line) - if m: - summary["related_indices"] = parse_indices(m.group(1)) - summary["unrelated_indices"] = parse_indices(m.group(2)) - continue - - # 对话级统计 - m = re_dialog.search(line) - if m: - dialogs.append({ - "index": parse_int(m.group(1)), - "total_messages": parse_int(m.group(2)), - "quota_delete": parse_int(m.group(3)), - "actual_deleted": parse_int(m.group(4)), - "kept": parse_int(m.group(5)), - }) - continue - - # 全局删除总数 - m = re_total_del.search(line) - if m: - summary["total_deleted_messages"] = parse_int(m.group(1)) - continue - - # 剩余对话数 - m = re_remaining.search(line) - if m: - summary["remaining_dialogs"] = parse_int(m.group(1)) - continue - - return { - "scene": summary["scene"], - "timestamp": datetime.now().isoformat(), - "summary": {k: v for k, v in summary.items() if k != "scene"}, - "dialogs": dialogs, - } diff --git a/api/app/core/memory/storage_services/extraction_engine/deduplication/deduped_and_disamb.py b/api/app/core/memory/storage_services/extraction_engine/deduplication/deduped_and_disamb.py index f2f14d9e..622f6e05 100644 --- a/api/app/core/memory/storage_services/extraction_engine/deduplication/deduped_and_disamb.py +++ b/api/app/core/memory/storage_services/extraction_engine/deduplication/deduped_and_disamb.py @@ -203,6 +203,7 @@ def accurate_match( ) -> Tuple[List[ExtractedEntityNode], Dict[str, str], Dict[str, Dict]]: """ 精确匹配:按 (end_user_id, name, entity_type) 合并实体并建立重定向与合并记录。 + 同时检测某实体的 name 是否命中另一实体的 aliases,若命中则直接合并。 返回: (deduped_entities, id_redirect, exact_merge_map) """ exact_merge_map: Dict[str, Dict] = {} @@ -240,6 +241,48 @@ def accurate_match( pass deduped_entities = list(canonical_map.values()) + + # 2) 第二轮:检测某实体的 name 是否命中另一实体的 aliases(alias-to-name 精确合并) + # 场景:LLM 把 aliases 中的词(如"齐齐")又单独抽取为独立实体,需在此阶段合并掉 + # 优化:先构建 (end_user_id, alias_lower) -> canonical 的反向索引,查找 O(1) + alias_index: Dict[tuple, ExtractedEntityNode] = {} + for canonical in deduped_entities: + uid = getattr(canonical, "end_user_id", None) + for alias in (getattr(canonical, "aliases", []) or []): + alias_lower = alias.strip().lower() + if alias_lower: + alias_index[(uid, alias_lower)] = canonical + + i = 0 + while i < len(deduped_entities): + ent = deduped_entities[i] + ent_name = (getattr(ent, "name", "") or "").strip().lower() + ent_uid = getattr(ent, "end_user_id", None) + canonical = alias_index.get((ent_uid, ent_name)) + # 确保不是自身 + if canonical is not None and canonical.id != ent.id: + _merge_attribute(canonical, ent) + id_redirect[ent.id] = canonical.id + for k, v in list(id_redirect.items()): + if v == ent.id: + id_redirect[k] = canonical.id + try: + k = f"{canonical.end_user_id}|{(canonical.name or '').strip()}|{(canonical.entity_type or '').strip()}" + if k not in exact_merge_map: + exact_merge_map[k] = { + "canonical_id": canonical.id, + "end_user_id": canonical.end_user_id, + "name": canonical.name, + "entity_type": canonical.entity_type, + "merged_ids": set(), + } + exact_merge_map[k]["merged_ids"].add(ent.id) + except Exception: + pass + deduped_entities.pop(i) + else: + i += 1 + return deduped_entities, id_redirect, exact_merge_map def fuzzy_match( diff --git a/api/app/core/memory/storage_services/extraction_engine/deduplication/two_stage_dedup.py b/api/app/core/memory/storage_services/extraction_engine/deduplication/two_stage_dedup.py index f28b8a5f..4b9c5718 100644 --- a/api/app/core/memory/storage_services/extraction_engine/deduplication/two_stage_dedup.py +++ b/api/app/core/memory/storage_services/extraction_engine/deduplication/two_stage_dedup.py @@ -25,17 +25,17 @@ from app.repositories.neo4j.neo4j_connector import Neo4jConnector async def dedup_layers_and_merge_and_return( - dialogue_nodes: List[DialogueNode], - chunk_nodes: List[ChunkNode], - statement_nodes: List[StatementNode], - entity_nodes: List[ExtractedEntityNode], - statement_chunk_edges: List[StatementChunkEdge], - statement_entity_edges: List[StatementEntityEdge], - entity_entity_edges: List[EntityEntityEdge], - dialog_data_list: List[DialogData], - pipeline_config: ExtractionPipelineConfig, - connector: Optional[Neo4jConnector] = None, - llm_client = None, + dialogue_nodes: List[DialogueNode], + chunk_nodes: List[ChunkNode], + statement_nodes: List[StatementNode], + entity_nodes: List[ExtractedEntityNode], + statement_chunk_edges: List[StatementChunkEdge], + statement_entity_edges: List[StatementEntityEdge], + entity_entity_edges: List[EntityEntityEdge], + dialog_data_list: List[DialogData], + pipeline_config: ExtractionPipelineConfig, + connector: Optional[Neo4jConnector] = None, + llm_client=None, ) -> Tuple[ List[DialogueNode], List[ChunkNode], @@ -44,7 +44,7 @@ async def dedup_layers_and_merge_and_return( List[StatementChunkEdge], List[StatementEntityEdge], List[EntityEntityEdge], - dict, # 新增:返回去重详情 + dict ]: """ 执行两层实体去重与融合: diff --git a/api/app/core/memory/storage_services/extraction_engine/extraction_orchestrator.py b/api/app/core/memory/storage_services/extraction_engine/extraction_orchestrator.py index 1242e4e6..b20112a2 100644 --- a/api/app/core/memory/storage_services/extraction_engine/extraction_orchestrator.py +++ b/api/app/core/memory/storage_services/extraction_engine/extraction_orchestrator.py @@ -19,6 +19,7 @@ import asyncio import logging import os +import uuid from datetime import datetime from typing import Any, Awaitable, Callable, Dict, List, Optional, Tuple @@ -32,10 +33,11 @@ from app.core.memory.models.graph_models import ( StatementChunkEdge, StatementEntityEdge, StatementNode, + PerceptualEdge, + PerceptualNode ) from app.core.memory.models.message_models import DialogData from app.core.memory.models.ontology_extraction_models import OntologyTypeList -from app.core.memory.models.ontology_extraction_models import OntologyTypeList from app.core.memory.models.variate_config import ( ExtractionPipelineConfig, ) @@ -46,7 +48,6 @@ from app.core.memory.storage_services.extraction_engine.knowledge_extraction.emb embedding_generation, generate_entity_embeddings_from_triplets, ) - # 导入各个提取模块 from app.core.memory.storage_services.extraction_engine.knowledge_extraction.statement_extraction import ( StatementExtractor, @@ -62,6 +63,10 @@ from app.core.memory.storage_services.extraction_engine.pipeline_help import ( export_test_input_doc, ) from app.core.memory.utils.data.ontology import TemporalInfo +from app.db import get_db_context +from app.models.end_user_info_model import EndUserInfo +from app.repositories.end_user_info_repository import EndUserInfoRepository +from app.repositories.end_user_repository import EndUserRepository from app.repositories.neo4j.neo4j_connector import Neo4jConnector # 配置日志 @@ -90,16 +95,16 @@ class ExtractionOrchestrator: """ def __init__( - self, - llm_client: LLMClient, - embedder_client: OpenAIEmbedderClient, - connector: Neo4jConnector, - config: Optional[ExtractionPipelineConfig] = None, - progress_callback: Optional[Callable[[str, str, Optional[Dict[str, Any]]], Awaitable[None]]] = None, - embedding_id: Optional[str] = None, - ontology_types: Optional[OntologyTypeList] = None, - enable_general_types: bool = True, - language: str = "zh", + self, + llm_client: LLMClient, + embedder_client: OpenAIEmbedderClient, + connector: Neo4jConnector, + config: Optional[ExtractionPipelineConfig] = None, + progress_callback: Optional[Callable[[str, str, Optional[Dict[str, Any]]], Awaitable[None]]] = None, + embedding_id: Optional[str] = None, + ontology_types: Optional[OntologyTypeList] = None, + enable_general_types: bool = True, + language: str = "zh", ): """ 初始化流水线编排器 @@ -123,7 +128,7 @@ class ExtractionOrchestrator: self.progress_callback = progress_callback # 保存进度回调函数 self.embedding_id = embedding_id # 保存嵌入模型ID self.language = language # 保存语言配置 - + # 处理本体类型配置 # 根据 enable_general_types 参数决定是否将通用本体类型与场景特定类型合并 # 如果启用合并且配置中开启了通用本体功能,则使用 OntologyTypeMerger 进行融合 @@ -146,7 +151,7 @@ class ExtractionOrchestrator: self.ontology_types = ontology_types if not enable_general_types and ontology_types: logger.info("enable_general_types=False,仅使用场景类型") - + # 保存去重消歧的详细记录(内存中的数据结构) self.dedup_merge_records: List[Dict[str, Any]] = [] # 实体合并记录 self.dedup_disamb_records: List[Dict[str, Any]] = [] # 实体消歧记录 @@ -157,19 +162,27 @@ class ExtractionOrchestrator: llm_client=llm_client, config=self.config.statement_extraction, ) - self.triplet_extractor = TripletExtractor(llm_client=llm_client,ontology_types=self.ontology_types, language=language) + self.triplet_extractor = TripletExtractor(llm_client=llm_client, ontology_types=self.ontology_types, + language=language) self.temporal_extractor = TemporalExtractor(llm_client=llm_client) logger.info("ExtractionOrchestrator 初始化完成") async def run( - self, - dialog_data_list: List[DialogData], - is_pilot_run: bool = False, - ) -> Tuple[ - Tuple[List[DialogueNode], List[ChunkNode], List[StatementNode]], - Tuple[List[ExtractedEntityNode], List[StatementEntityEdge], List[EntityEntityEdge]], - Tuple[List[ExtractedEntityNode], List[StatementEntityEdge], List[EntityEntityEdge]], + self, + dialog_data_list: List[DialogData], + is_pilot_run: bool = False, + ) -> tuple[ + list[DialogueNode], + list[ChunkNode], + list[StatementNode], + list[ExtractedEntityNode], + list[PerceptualNode], + list[StatementChunkEdge], + list[StatementEntityEdge], + list[EntityEntityEdge], + list[PerceptualEdge], + list[DialogData] ]: """ 运行完整的知识提取流水线(优化版:并行执行) @@ -202,13 +215,12 @@ class ExtractionOrchestrator: # 步骤 1: 陈述句提取 logger.info("步骤 1/6: 陈述句提取(全局分块级并行)") dialog_data_list = await self._extract_statements(dialog_data_list) - + # 收集陈述句内容和统计数量 all_statements_list = [] for dialog in dialog_data_list: for chunk in dialog.chunks: all_statements_list.extend(chunk.statements) - len(all_statements_list) # 步骤 2: 并行执行三元组提取、时间信息提取、情绪提取和基础嵌入生成 logger.info("步骤 2/6: 并行执行三元组提取、时间信息提取、情绪提取和嵌入生成") @@ -220,7 +232,7 @@ class ExtractionOrchestrator: chunk_embedding_maps, dialog_embeddings, ) = await self._parallel_extract_and_embed(dialog_data_list) - + # 收集实体和三元组内容,并统计数量 all_entities_list = [] all_triplets_list = [] @@ -229,10 +241,6 @@ class ExtractionOrchestrator: if triplet_info: all_entities_list.extend(triplet_info.entities) all_triplets_list.extend(triplet_info.triplets) - - len(all_entities_list) - len(all_triplets_list) - sum(len(temporal_map) for temporal_map in temporal_maps) # 步骤 3: 生成实体嵌入(依赖三元组提取结果) logger.info("步骤 3/6: 生成实体嵌入") @@ -252,17 +260,19 @@ class ExtractionOrchestrator: # 步骤 5: 创建节点和边 logger.info("步骤 5/6: 创建节点和边") - + # 注意:creating_nodes_edges 消息已在知识抽取完成后立即发送 - + ( dialogue_nodes, chunk_nodes, statement_nodes, entity_nodes, + perceptual_nodes, statement_chunk_edges, statement_entity_edges, entity_entity_edges, + perceptual_edges ) = await self._create_nodes_and_edges(dialog_data_list) # 导出去重前的测试输入文档(试运行和正式模式都需要,用于生成结果汇总) @@ -273,10 +283,20 @@ class ExtractionOrchestrator: logger.info("步骤 6/6: 去重和消歧(试运行模式:仅第一层去重)") else: logger.info("步骤 6/6: 两阶段去重和消歧") - + # 注意:deduplication 消息已在创建节点和边完成后立即发送 - - result = await self._run_dedup_and_write_summary( + + ( + dialogue_nodes, + chunk_nodes, + statement_nodes, + entity_nodes, + statement_chunk_edges, + statement_entity_edges, + entity_entity_edges, + dialog_data_list, + dedup_details, + ) = await self._run_dedup_and_write_summary( dialogue_nodes, chunk_nodes, statement_nodes, @@ -287,17 +307,31 @@ class ExtractionOrchestrator: dialog_data_list, ) - + # 步骤 7: 同步用户别名到数据库表(仅正式模式) + if not is_pilot_run: + logger.info("步骤 7: 同步用户别名到 end_user 和 end_user_info 表") + await self._update_end_user_other_name(entity_nodes, dialog_data_list) logger.info(f"知识提取流水线运行完成({mode_str})") - return result + return ( + dialogue_nodes, + chunk_nodes, + statement_nodes, + entity_nodes, + perceptual_nodes, + statement_chunk_edges, + statement_entity_edges, + entity_entity_edges, + perceptual_edges, + dialog_data_list, + ) except Exception as e: logger.error(f"知识提取流水线运行失败: {e}", exc_info=True) raise async def _extract_statements( - self, dialog_data_list: List[DialogData] + self, dialog_data_list: List[DialogData] ) -> List[DialogData]: """ 从对话中提取陈述句(流式输出版本:边提取边发送进度) @@ -313,7 +347,7 @@ class ExtractionOrchestrator: # 收集所有分块及其元数据 all_chunks = [] chunk_metadata = [] # (dialog_idx, chunk_idx) - + for d_idx, dialog in enumerate(dialog_data_list): dialogue_content = dialog.content if self.config.statement_extraction.include_dialogue_context else None for c_idx, chunk in enumerate(dialog.chunks): @@ -321,7 +355,7 @@ class ExtractionOrchestrator: chunk_metadata.append((d_idx, c_idx)) logger.info(f"收集到 {len(all_chunks)} 个分块,开始全局并行提取") - + # 用于跟踪已完成的分块数量 completed_chunks = 0 total_chunks = len(all_chunks) @@ -332,7 +366,7 @@ class ExtractionOrchestrator: chunk, end_user_id, dialogue_content = chunk_data try: statements = await self.statement_extractor._extract_statements(chunk, end_user_id, dialogue_content) - + # 流式输出:每提取完一个分块的陈述句,立即发送进度 # 注意:只在试运行模式下发送陈述句详情,正式模式不发送 completed_chunks += 1 @@ -347,11 +381,11 @@ class ExtractionOrchestrator: "statement_index_in_chunk": idx + 1 } await self.progress_callback( - "knowledge_extraction_result", - f"陈述句提取中 ({completed_chunks}/{total_chunks})", + "knowledge_extraction_result", + f"陈述句提取中 ({completed_chunks}/{total_chunks})", stmt_result ) - + return statements except Exception as e: logger.error(f"分块 {chunk.id} 陈述句提取失败: {e}") @@ -381,13 +415,21 @@ class ExtractionOrchestrator: # 保存陈述句到文件(试运行和正式模式都需要) self.statement_extractor.save_statements(all_statements) - + logger.info(f"陈述句提取完成,共提取 {len(all_statements)} 条陈述句") + # 试运行模式下,所有分块提取完成后发送完成事件 + if self.progress_callback and self.is_pilot_run: + await self.progress_callback( + "knowledge_extraction_complete", + f"陈述句提取完成,共提取 {len(all_statements)} 条", + {"total_statements": len(all_statements), "total_chunks": total_chunks} + ) + return dialog_data_list async def _extract_triplets( - self, dialog_data_list: List[DialogData] + self, dialog_data_list: List[DialogData] ) -> List[Dict[str, Any]]: """ 从对话中提取三元组(流式输出版本:边提取边发送进度) @@ -403,7 +445,7 @@ class ExtractionOrchestrator: # 收集所有陈述句及其元数据 all_statements = [] statement_metadata = [] # (dialog_idx, statement_id, chunk_content) - + for d_idx, dialog in enumerate(dialog_data_list): for chunk in dialog.chunks: for statement in chunk.statements: @@ -411,7 +453,7 @@ class ExtractionOrchestrator: statement_metadata.append((d_idx, statement.id)) logger.info(f"收集到 {len(all_statements)} 个陈述句,开始全局并行提取三元组") - + # 用于跟踪已完成的陈述句数量 completed_statements = 0 len(all_statements) @@ -422,11 +464,11 @@ class ExtractionOrchestrator: statement, chunk_content = stmt_data try: triplet_info = await self.triplet_extractor._extract_triplets(statement, chunk_content) - + # 注意:不再发送三元组提取的流式输出 # 三元组提取在后台执行,但不向前端发送详细信息 completed_statements += 1 - + return triplet_info except Exception as e: logger.error(f"陈述句 {statement.id} 三元组提取失败: {e}") @@ -442,7 +484,7 @@ class ExtractionOrchestrator: # 将结果组织成对话级别的映射 triplet_maps = [{} for _ in dialog_data_list] all_responses = [] - + for i, result in enumerate(results): d_idx, stmt_id = statement_metadata[i] if isinstance(result, Exception): @@ -470,7 +512,7 @@ class ExtractionOrchestrator: return triplet_maps async def _extract_temporal( - self, dialog_data_list: List[DialogData] + self, dialog_data_list: List[DialogData] ) -> List[Dict[str, Any]]: """ 从对话中提取时间信息(流式输出版本:边提取边发送进度) @@ -494,13 +536,13 @@ class ExtractionOrchestrator: temporal_map[statement.id] = TemporalValidityRange(valid_at=None, invalid_at=None) temporal_maps.append(temporal_map) return temporal_maps - + logger.info("开始时间信息提取(全局陈述句级并行 + 流式输出)") # 收集所有需要提取时间的陈述句 all_statements = [] statement_metadata = [] # (dialog_idx, statement_id, ref_dates) - + for d_idx, dialog in enumerate(dialog_data_list): # 获取参考日期 ref_dates = {} @@ -509,11 +551,11 @@ class ExtractionOrchestrator: ref_dates['conversation_date'] = dialog.metadata['conversation_date'] if 'publication_date' in dialog.metadata: ref_dates['publication_date'] = dialog.metadata['publication_date'] - + if not ref_dates: from datetime import datetime ref_dates = {"today": datetime.now().strftime("%Y-%m-%d")} - + for chunk in dialog.chunks: for statement in chunk.statements: # 跳过 ATEMPORAL 类型的陈述句 @@ -523,7 +565,7 @@ class ExtractionOrchestrator: statement_metadata.append((d_idx, statement.id)) logger.info(f"收集到 {len(all_statements)} 个需要时间提取的陈述句,开始全局并行提取") - + # 用于跟踪已完成的时间提取数量 completed_temporal = 0 len(all_statements) @@ -534,11 +576,11 @@ class ExtractionOrchestrator: statement, ref_dates = stmt_data try: temporal_range = await self.temporal_extractor._extract_temporal_ranges(statement, ref_dates) - + # 注意:不再发送时间提取的流式输出 # 时间提取在后台执行,但不向前端发送详细信息 completed_temporal += 1 - + return temporal_range except Exception as e: logger.error(f"陈述句 {statement.id} 时间信息提取失败: {e}") @@ -551,7 +593,7 @@ class ExtractionOrchestrator: # 将结果组织成对话级别的映射 temporal_maps = [{} for _ in dialog_data_list] - + for i, result in enumerate(results): d_idx, stmt_id = statement_metadata[i] if isinstance(result, Exception): @@ -577,7 +619,7 @@ class ExtractionOrchestrator: return temporal_maps async def _extract_emotions( - self, dialog_data_list: List[DialogData] + self, dialog_data_list: List[DialogData] ) -> List[Dict[str, Any]]: """ 从对话中提取情绪信息(仅针对用户消息,全局陈述句级并行) @@ -593,36 +635,36 @@ class ExtractionOrchestrator: # 收集所有陈述句及其配置 all_statements = [] statement_metadata = [] # (dialog_idx, statement_id) - + # 获取第一个对话的config_id来加载配置 config_id = None if dialog_data_list and hasattr(dialog_data_list[0], 'config_id'): config_id = dialog_data_list[0].config_id - + # 加载MemoryConfig memory_config = None if config_id: try: from app.db import SessionLocal from app.repositories.memory_config_repository import MemoryConfigRepository - + db = SessionLocal() try: memory_config = MemoryConfigRepository.get_by_id(db, config_id) finally: db.close() - + if memory_config and not memory_config.emotion_enabled: logger.info("情绪提取已在配置中禁用,跳过情绪提取") return [{} for _ in dialog_data_list] - + except Exception as e: logger.warning(f"加载MemoryConfig失败: {e},将跳过情绪提取") return [{} for _ in dialog_data_list] else: logger.info("未找到config_id,跳过情绪提取") return [{} for _ in dialog_data_list] - + # 如果配置未启用情绪提取,直接返回空映射 if not memory_config or not memory_config.emotion_enabled: logger.info("情绪提取未启用,跳过") @@ -631,7 +673,7 @@ class ExtractionOrchestrator: # 收集所有陈述句(只收集 speaker 为 "user" 的) total_statements = 0 filtered_statements = 0 - + for d_idx, dialog in enumerate(dialog_data_list): for chunk in dialog.chunks: for statement in chunk.statements: @@ -647,12 +689,12 @@ class ExtractionOrchestrator: # 初始化情绪提取服务 # 如果 emotion_model_id 为空,回退到工作空间默认 LLM from app.services.emotion_extraction_service import EmotionExtractionService - + emotion_model_id = memory_config.emotion_model_id if not emotion_model_id and memory_config.workspace_id: from app.repositories.workspace_repository import get_workspace_models_configs from app.db import SessionLocal - + db = SessionLocal() try: workspace_models = get_workspace_models_configs(db, memory_config.workspace_id) @@ -661,7 +703,7 @@ class ExtractionOrchestrator: logger.info(f"emotion_model_id 为空,使用工作空间默认 LLM: {emotion_model_id}") finally: db.close() - + emotion_service = EmotionExtractionService( llm_id=emotion_model_id if emotion_model_id else None ) @@ -681,7 +723,7 @@ class ExtractionOrchestrator: # 将结果组织成对话级别的映射 emotion_maps = [{} for _ in dialog_data_list] successful_extractions = 0 - + for i, result in enumerate(results): d_idx, stmt_id = statement_metadata[i] if isinstance(result, Exception): @@ -698,7 +740,7 @@ class ExtractionOrchestrator: return emotion_maps async def _parallel_extract_and_embed( - self, dialog_data_list: List[DialogData] + self, dialog_data_list: List[DialogData] ) -> Tuple[ List[Dict[str, Any]], List[Dict[str, Any]], @@ -749,7 +791,7 @@ class ExtractionOrchestrator: triplet_maps = results[0] if not isinstance(results[0], Exception) else [{} for _ in dialog_data_list] temporal_maps = results[1] if not isinstance(results[1], Exception) else [{} for _ in dialog_data_list] emotion_maps = results[2] if not isinstance(results[2], Exception) else [{} for _ in dialog_data_list] - + if isinstance(results[3], Exception): logger.error(f"基础嵌入生成失败: {results[3]}") statement_embedding_maps = [{} for _ in dialog_data_list] @@ -769,7 +811,7 @@ class ExtractionOrchestrator: ) async def _generate_basic_embeddings( - self, dialog_data_list: List[DialogData] + self, dialog_data_list: List[DialogData] ) -> Tuple[List[Dict[str, List[float]]], List[Dict[str, List[float]]], List[List[float]]]: """ 生成基础嵌入向量(陈述句、分块、对话) @@ -802,7 +844,7 @@ class ExtractionOrchestrator: if not self.embedding_id: logger.error("embedding_id is required but was not provided to ExtractionOrchestrator") raise ValueError("embedding_id is required but was not provided") - + # 只生成陈述句、分块和对话的嵌入(不包括实体) statement_embedding_maps, chunk_embedding_maps, dialog_embeddings = await embedding_generation( dialog_data_list, self.embedding_id @@ -828,7 +870,7 @@ class ExtractionOrchestrator: ) async def _generate_entity_embeddings( - self, triplet_maps: List[Dict[str, Any]] + self, triplet_maps: List[Dict[str, Any]] ) -> List[Dict[str, Any]]: """ 生成实体嵌入向量 @@ -853,7 +895,7 @@ class ExtractionOrchestrator: if not self.embedding_id: logger.error("embedding_id is required but was not provided to ExtractionOrchestrator") return triplet_maps - + # 生成实体嵌入 updated_triplet_maps = await generate_entity_embeddings_from_triplets( triplet_maps, self.embedding_id @@ -866,17 +908,15 @@ class ExtractionOrchestrator: logger.error(f"实体嵌入生成失败: {e}", exc_info=True) return triplet_maps - - async def _assign_extracted_data( - self, - dialog_data_list: List[DialogData], - temporal_maps: List[Dict[str, Any]], - triplet_maps: List[Dict[str, Any]], - emotion_maps: List[Dict[str, Any]], - statement_embedding_maps: List[Dict[str, List[float]]], - chunk_embedding_maps: List[Dict[str, List[float]]], - dialog_embeddings: List[List[float]], + self, + dialog_data_list: List[DialogData], + temporal_maps: List[Dict[str, Any]], + triplet_maps: List[Dict[str, Any]], + emotion_maps: List[Dict[str, Any]], + statement_embedding_maps: List[Dict[str, List[float]]], + chunk_embedding_maps: List[Dict[str, List[float]]], + dialog_embeddings: List[List[float]], ) -> List[DialogData]: """ 将提取的数据赋值到语句 @@ -898,12 +938,12 @@ class ExtractionOrchestrator: # 确保列表长度匹配 expected_length = len(dialog_data_list) if ( - len(temporal_maps) != expected_length - or len(triplet_maps) != expected_length - or len(emotion_maps) != expected_length - or len(statement_embedding_maps) != expected_length - or len(chunk_embedding_maps) != expected_length - or len(dialog_embeddings) != expected_length + len(temporal_maps) != expected_length + or len(triplet_maps) != expected_length + or len(emotion_maps) != expected_length + or len(statement_embedding_maps) != expected_length + or len(chunk_embedding_maps) != expected_length + or len(dialog_embeddings) != expected_length ): logger.warning( f"数据大小不匹配 - 对话: {len(dialog_data_list)}, " @@ -991,15 +1031,17 @@ class ExtractionOrchestrator: return dialog_data_list async def _create_nodes_and_edges( - self, dialog_data_list: List[DialogData] + self, dialog_data_list: List[DialogData] ) -> Tuple[ List[DialogueNode], List[ChunkNode], List[StatementNode], List[ExtractedEntityNode], + List[PerceptualNode], List[StatementChunkEdge], List[StatementEntityEdge], List[EntityEntityEdge], + List[PerceptualEdge] ]: """ 创建图数据库节点和边 @@ -1013,7 +1055,7 @@ class ExtractionOrchestrator: 包含所有节点和边的元组 """ logger.info("开始创建节点和边") - + # 注意:开始消息已在 run 方法中发送,这里不再重复发送 dialogue_nodes = [] @@ -1023,10 +1065,12 @@ class ExtractionOrchestrator: statement_chunk_edges = [] statement_entity_edges = [] entity_entity_edges = [] + perceptual_nodes = [] + perceptual_edges = [] # 用于去重的集合 entity_id_set = set() - + # 用于跟踪进度 total_dialogs = len(dialog_data_list) processed_dialogs = 0 @@ -1067,6 +1111,45 @@ class ExtractionOrchestrator: ) chunk_nodes.append(chunk_node) + for p, file_type in chunk.files: + + meta = p.meta_data or {} + content_meta = meta.get("content", {}) + + # 生成 summary embedding(如果有 embedder_client) + summary_embedding = None + if self.embedder_client and p.summary: + try: + summary_embedding = (await self.embedder_client.response([p.summary]))[0] + except Exception as emb_err: + print(f"Failed to embed perceptual summary: {emb_err}") + + perceptual = PerceptualNode( + name=f"Perceptual_{p.id}", + **{ + "id": str(p.id), + "end_user_id": str(p.end_user_id), + "perceptual_type": p.perceptual_type, + "file_path": p.file_path or "", + "file_name": p.file_name or "", + "file_ext": p.file_ext or "", + "summary": p.summary or "", + "keywords": content_meta.get("keywords", []), + "topic": content_meta.get("topic", ""), + "domain": content_meta.get("domain", ""), + "created_at": p.created_time.isoformat() if p.created_time else None, + "file_type": file_type, + "summary_embedding": summary_embedding, + }) + perceptual_nodes.append(perceptual) + perceptual_edges.append(PerceptualEdge( + source=perceptual.id, + target=chunk.id, + end_user_id=dialog_data.end_user_id, + run_id=dialog_data.run_id, + created_at=dialog_data.created_at, + )) + # 处理每个陈述句 for statement in chunk.statements: # 创建陈述句节点 @@ -1075,15 +1158,19 @@ class ExtractionOrchestrator: name=f"Statement_{statement.id}", # 添加必需的 name 字段 chunk_id=chunk.id, stmt_type=getattr(statement, 'stmt_type', 'general'), # 添加必需的 stmt_type 字段 - temporal_info=getattr(statement, 'temporal_info', TemporalInfo.ATEMPORAL), # 添加必需的 temporal_info 字段 - connect_strength=statement.connect_strength if statement.connect_strength is not None else 'Strong', # 添加必需的 connect_strength 字段 + temporal_info=getattr(statement, 'temporal_info', TemporalInfo.ATEMPORAL), + # 添加必需的 temporal_info 字段 + connect_strength=statement.connect_strength if statement.connect_strength is not None else 'Strong', + # 添加必需的 connect_strength 字段 end_user_id=dialog_data.end_user_id, run_id=dialog_data.run_id, # 使用 dialog_data 的 run_id statement=statement.statement, speaker=getattr(statement, 'speaker', None), # 添加 speaker 字段 statement_embedding=statement.statement_embedding, - valid_at=statement.temporal_validity.valid_at if hasattr(statement, 'temporal_validity') and statement.temporal_validity else None, - invalid_at=statement.temporal_validity.invalid_at if hasattr(statement, 'temporal_validity') and statement.temporal_validity else None, + valid_at=statement.temporal_validity.valid_at if hasattr(statement, + 'temporal_validity') and statement.temporal_validity else None, + invalid_at=statement.temporal_validity.invalid_at if hasattr(statement, + 'temporal_validity') and statement.temporal_validity else None, created_at=dialog_data.created_at, expired_at=dialog_data.expired_at, config_id=dialog_data.config_id if hasattr(dialog_data, 'config_id') else None, @@ -1112,7 +1199,7 @@ class ExtractionOrchestrator: # 创建实体索引到ID的映射(支持多种索引方式) entity_idx_to_id = {} - + # 创建实体节点 for entity_idx, entity in enumerate(triplet_info.entities): # 映射实体索引到实体ID(使用多个键以提高容错性) @@ -1120,7 +1207,7 @@ class ExtractionOrchestrator: entity_idx_to_id[entity.entity_idx] = entity.id # 2. 使用枚举索引(从0开始) entity_idx_to_id[entity_idx] = entity.id - + if entity.id not in entity_id_set: entity_connect_strength = getattr(entity, 'connect_strength', 'Strong') entity_node = ExtractedEntityNode( @@ -1133,7 +1220,8 @@ class ExtractionOrchestrator: example=getattr(entity, 'example', ''), # 新增:传递示例字段 # TODO: fact_summary 功能暂时禁用,待后续开发完善后启用 # fact_summary=getattr(entity, 'fact_summary', ''), # 添加必需的 fact_summary 字段 - connect_strength=entity_connect_strength if entity_connect_strength is not None else 'Strong', # 添加必需的 connect_strength 字段 + connect_strength=entity_connect_strength if entity_connect_strength is not None else 'Strong', + # 添加必需的 connect_strength 字段 aliases=getattr(entity, 'aliases', []) or [], # 传递从三元组提取阶段获取的aliases name_embedding=getattr(entity, 'name_embedding', None), is_explicit_memory=getattr(entity, 'is_explicit_memory', False), # 新增:传递语义记忆标记 @@ -1163,7 +1251,7 @@ class ExtractionOrchestrator: # 将三元组中的整数索引映射到实体ID subject_entity_id = entity_idx_to_id.get(triplet.subject_id) object_entity_id = entity_idx_to_id.get(triplet.object_id) - + # 只有当两个实体ID都存在时才创建边 if subject_entity_id and object_entity_id: entity_entity_edge = EntityEntityEdge( @@ -1178,7 +1266,7 @@ class ExtractionOrchestrator: expired_at=dialog_data.expired_at, ) entity_entity_edges.append(entity_entity_edge) - + # 流式输出:每创建一个关系边,立即发送进度(限制发送数量) if self.progress_callback and len(entity_entity_edges) <= 10: # 获取实体名称 @@ -1194,8 +1282,8 @@ class ExtractionOrchestrator: "dialog_progress": f"{processed_dialogs}/{total_dialogs}" } await self.progress_callback( - "creating_nodes_edges_result", - f"关系创建中 ({processed_dialogs}/{total_dialogs})", + "creating_nodes_edges_result", + f"关系创建中 ({processed_dialogs}/{total_dialogs})", relationship_result ) else: @@ -1203,7 +1291,7 @@ class ExtractionOrchestrator: missing_subject = "subject" if not subject_entity_id else "" missing_object = "object" if not object_entity_id else "" missing_both = " and " if (not subject_entity_id and not object_entity_id) else "" - + logger.debug( f"跳过三元组 - 无法找到{missing_subject}{missing_both}{missing_object}实体ID: " f"subject_id={triplet.subject_id} ({triplet.subject_name}), " @@ -1220,7 +1308,7 @@ class ExtractionOrchestrator: f"陈述句-实体边: {len(statement_entity_edges)}, " f"实体-实体边: {len(entity_entity_edges)}" ) - + # 进度回调:创建节点和边完成,传递结果统计 # 注意:具体的关系创建结果已经在创建过程中实时发送了 if self.progress_callback: @@ -1240,25 +1328,197 @@ class ExtractionOrchestrator: chunk_nodes, statement_nodes, entity_nodes, + perceptual_nodes, statement_chunk_edges, statement_entity_edges, entity_entity_edges, + perceptual_edges ) + async def _update_end_user_other_name( + self, + entity_nodes: List[ExtractedEntityNode], + dialog_data_list: List[DialogData] + ) -> None: + """ + 从 Neo4j 读取用户实体的最终 aliases,同步到 end_user 和 end_user_info 表 + + 注意: + 1. other_name 使用本次对话提取的第一个别名(保持时间顺序) + 2. aliases 从 Neo4j 读取(保持完整性) + + Args: + entity_nodes: 实体节点列表 + dialog_data_list: 对话数据列表 + """ + try: + if not dialog_data_list: + logger.warning("dialog_data_list 为空,跳过用户别名同步") + return + + end_user_id = dialog_data_list[0].end_user_id + if not end_user_id: + logger.warning("end_user_id 为空,跳过用户别名同步") + return + + # 1. 提取本次对话的用户别名(保持 LLM 提取的原始顺序,不排序) + current_aliases = self._extract_current_aliases(entity_nodes) + + # 2. 从 Neo4j 获取完整 aliases(权威数据源) + neo4j_aliases = await self._fetch_neo4j_user_aliases(end_user_id) + + if not neo4j_aliases: + # Neo4j 中没有别名,使用本次对话提取的别名 + neo4j_aliases = current_aliases + if not neo4j_aliases: + logger.debug(f"aliases 为空,跳过同步: end_user_id={end_user_id}") + return + + logger.info(f"本次对话提取的 aliases: {current_aliases}") + logger.info(f"Neo4j 中的完整 aliases: {neo4j_aliases}") + + # 3. 同步到数据库 + end_user_uuid = uuid.UUID(end_user_id) + with get_db_context() as db: + # 更新 end_user 表 + end_user = EndUserRepository(db).get_by_id(end_user_uuid) + if not end_user: + logger.warning(f"未找到 end_user_id={end_user_id} 的用户记录") + return + + new_name = self._resolve_other_name(end_user.other_name, current_aliases, neo4j_aliases) + if new_name is not None: + end_user.other_name = new_name + logger.info(f"更新 end_user 表 other_name → {new_name}") + else: + logger.debug(f"end_user 表 other_name 保持不变: {end_user.other_name}") + + # 更新或创建 end_user_info 记录 + info = EndUserInfoRepository(db).get_by_end_user_id(end_user_uuid) + if info: + new_name_info = self._resolve_other_name(info.other_name, current_aliases, neo4j_aliases) + if new_name_info is not None: + info.other_name = new_name_info + logger.info(f"更新 end_user_info 表 other_name → {new_name_info}") + if info.aliases != neo4j_aliases: + info.aliases = neo4j_aliases + logger.info(f"同步 Neo4j aliases 到 end_user_info: {neo4j_aliases}") + else: + first_alias = current_aliases[0].strip() if current_aliases else "" + # 确保 first_alias 不是占位名称 + if first_alias and first_alias not in self.USER_PLACEHOLDER_NAMES: + db.add(EndUserInfo( + end_user_id=end_user_uuid, + other_name=first_alias, + aliases=neo4j_aliases, + meta_data={} + )) + logger.info(f"创建 end_user_info 记录,other_name={first_alias}, aliases={neo4j_aliases}") + + db.commit() + + except Exception as e: + logger.error(f"更新 end_user other_name 失败: {e}", exc_info=True) + + + + # 用户实体占位名称,不允许作为 other_name 或出现在 aliases 中 + USER_PLACEHOLDER_NAMES = {'用户', '我', 'User', 'I'} + + def _extract_current_aliases(self, entity_nodes: List[ExtractedEntityNode]) -> List[str]: + """从实体节点提取用户别名(保持 LLM 提取的原始顺序,不进行任何排序) + + 这个方法直接返回 LLM 提取的别名列表,并过滤掉占位名称("用户"、"我"、"User"、"I")。 + 第一个别名将被用作 other_name。 + + Args: + entity_nodes: 实体节点列表 + + Returns: + 别名列表(保持 LLM 提取的原始顺序,已过滤占位名称) + """ + for entity in entity_nodes: + if getattr(entity, 'name', '').strip() in self.USER_PLACEHOLDER_NAMES: + aliases = getattr(entity, 'aliases', []) or [] + # 过滤掉占位名称,防止 "用户"/"我"/"User"/"I" 被存入 aliases 和 other_name + filtered = [a for a in aliases if a.strip() not in self.USER_PLACEHOLDER_NAMES] + logger.debug(f"提取到用户别名(原始顺序,已过滤占位名称): {filtered}") + return filtered + return [] + + + async def _fetch_neo4j_user_aliases(self, end_user_id: str) -> List[str]: + """从 Neo4j 查询用户实体的完整 aliases 列表(已过滤占位名称)""" + cypher = """ + MATCH (e:ExtractedEntity) + WHERE e.end_user_id = $end_user_id AND e.name IN ['用户', '我', 'User', 'I'] + RETURN e.aliases AS aliases + LIMIT 1 + """ + result = await Neo4jConnector().execute_query(cypher, end_user_id=end_user_id) + if not result: + logger.debug(f"Neo4j 中未找到用户实体: end_user_id={end_user_id}") + return [] + aliases = result[0].get('aliases') or [] + if not aliases: + logger.debug(f"Neo4j 用户实体 aliases 为空: end_user_id={end_user_id}") + return [] + # 过滤掉占位名称,防止历史脏数据传播 + filtered = [a for a in aliases if a.strip() not in self.USER_PLACEHOLDER_NAMES] + return filtered + + def _resolve_other_name( + self, + current: Optional[str], + current_aliases: List[str], + neo4j_aliases: List[str] + ) -> Optional[str]: + """ + 决定 other_name 是否需要更新,返回新值;无需更新返回 None。 + + 决策规则: + - 为空或为占位名称 → 用本次对话第一个别名 + - 不在 Neo4j aliases 中 → 用 Neo4j 第一个别名(说明已被删除) + - 否则 → 保持不变(返回 None) + + 注意:返回值不允许是占位名称("用户"、"我"、"User"、"I") + """ + # 当前值为空或为占位名称时,需要更新 + if not current or not current.strip() or current.strip() in self.USER_PLACEHOLDER_NAMES: + candidate = current_aliases[0].strip() if current_aliases else None + # 确保候选值不是占位名称 + if candidate and candidate in self.USER_PLACEHOLDER_NAMES: + return None + return candidate + if current not in neo4j_aliases: + candidate = neo4j_aliases[0].strip() if neo4j_aliases else None + # 确保候选值不是占位名称 + if candidate and candidate in self.USER_PLACEHOLDER_NAMES: + return None + return candidate + + return None + async def _run_dedup_and_write_summary( - self, - dialogue_nodes: List[DialogueNode], - chunk_nodes: List[ChunkNode], - statement_nodes: List[StatementNode], - entity_nodes: List[ExtractedEntityNode], - statement_chunk_edges: List[StatementChunkEdge], - statement_entity_edges: List[StatementEntityEdge], - entity_entity_edges: List[EntityEntityEdge], - dialog_data_list: List[DialogData], - ) -> Tuple[ - Tuple[List[DialogueNode], List[ChunkNode], List[StatementNode]], - Tuple[List[ExtractedEntityNode], List[StatementEntityEdge], List[EntityEntityEdge]], - Tuple[List[ExtractedEntityNode], List[StatementEntityEdge], List[EntityEntityEdge]], + self, + dialogue_nodes: List[DialogueNode], + chunk_nodes: List[ChunkNode], + statement_nodes: List[StatementNode], + entity_nodes: List[ExtractedEntityNode], + statement_chunk_edges: List[StatementChunkEdge], + statement_entity_edges: List[StatementEntityEdge], + entity_entity_edges: List[EntityEntityEdge], + dialog_data_list: List[DialogData], + ) -> tuple[ + list[DialogueNode], + list[ChunkNode], + list[StatementNode], + list[ExtractedEntityNode], + list[StatementChunkEdge], + list[StatementEntityEdge], + list[EntityEntityEdge], + list[DialogData], + dict ]: """ 执行两阶段去重并写入汇总 @@ -1280,11 +1540,11 @@ class ExtractionOrchestrator: - 第三个元组:去重后的 (实体节点列表, 陈述句-实体边列表, 实体-实体边列表) """ logger.info("开始两阶段实体去重和消歧") - + # 进度回调:发送去重消歧开始消息 if self.progress_callback: await self.progress_callback("deduplication", "正在去重消歧...") - + logger.info( f"去重前: {len(entity_nodes)} 个实体节点, " f"{len(statement_entity_edges)} 条陈述句-实体边, " @@ -1299,7 +1559,7 @@ class ExtractionOrchestrator: from app.core.memory.storage_services.extraction_engine.deduplication.deduped_and_disamb import ( deduplicate_entities_and_edges, ) - + dedup_entity_nodes, dedup_statement_entity_edges, dedup_entity_entity_edges, dedup_details = await deduplicate_entities_and_edges( entity_nodes, statement_entity_edges, @@ -1309,10 +1569,10 @@ class ExtractionOrchestrator: dedup_config=self.config.deduplication, llm_client=self.llm_client, ) - + # 保存去重消歧的详细记录到实例变量 self._save_dedup_details(dedup_details, entity_nodes, dedup_entity_nodes) - + result_tuple = ( dialogue_nodes, chunk_nodes, @@ -1321,14 +1581,25 @@ class ExtractionOrchestrator: statement_chunk_edges, dedup_statement_entity_edges, dedup_entity_entity_edges, + dialog_data_list, + dedup_details, ) - + final_entity_nodes = dedup_entity_nodes final_statement_entity_edges = dedup_statement_entity_edges final_entity_entity_edges = dedup_entity_entity_edges else: # 正式模式:执行完整的两阶段去重 - result_tuple = await dedup_layers_and_merge_and_return( + ( + dialogue_nodes, + chunk_nodes, + statement_nodes, + final_entity_nodes, + statement_chunk_edges, + final_statement_entity_edges, + final_entity_entity_edges, + dedup_details, + ) = await dedup_layers_and_merge_and_return( dialogue_nodes, chunk_nodes, statement_nodes, @@ -1342,21 +1613,21 @@ class ExtractionOrchestrator: llm_client=self.llm_client, ) - # 解包返回值 - ( - _, - _, - _, - final_entity_nodes, - _, - final_statement_entity_edges, - final_entity_entity_edges, - dedup_details, - ) = result_tuple - # 保存去重消歧的详细记录到实例变量 self._save_dedup_details(dedup_details, entity_nodes, final_entity_nodes) + result_tuple = ( + dialogue_nodes, + chunk_nodes, + statement_nodes, + final_entity_nodes, + statement_chunk_edges, + final_statement_entity_edges, + final_entity_entity_edges, + dialog_data_list, + dedup_details, + ) + logger.info( f"去重后: {len(final_entity_nodes)} 个实体节点, " f"{len(final_statement_entity_edges)} 条陈述句-实体边, " @@ -1367,12 +1638,12 @@ class ExtractionOrchestrator: f"陈述句-实体边减少 {len(statement_entity_edges) - len(final_statement_entity_edges)}, " f"实体-实体边减少 {len(entity_entity_edges) - len(final_entity_entity_edges)}" ) - + # 流式输出:实时输出去重消歧的具体结果 if self.progress_callback: # 分析实体合并情况(使用内存中的记录) merge_info = await self._analyze_entity_merges(entity_nodes, final_entity_nodes) - + # 逐个输出去重合并的实体示例 for i, merge_detail in enumerate(merge_info[:5]): # 输出前5个去重结果 dedup_result = { @@ -1383,10 +1654,10 @@ class ExtractionOrchestrator: "message": f"{merge_detail['main_entity_name']}合并{merge_detail['merged_count']}个:相似实体已合并" } await self.progress_callback("dedup_disambiguation_result", "实体去重中", dedup_result) - + # 分析实体消歧情况(使用内存中的记录) disamb_info = await self._analyze_entity_disambiguation(entity_nodes, final_entity_nodes) - + # 逐个输出实体消歧的结果 for i, disamb_detail in enumerate(disamb_info[:5]): # 输出前5个消歧结果 disamb_result = { @@ -1399,14 +1670,13 @@ class ExtractionOrchestrator: "message": f"{disamb_detail['entity_name']}消歧完成:{disamb_detail['disamb_type']}" } await self.progress_callback("dedup_disambiguation_result", "实体消歧中", disamb_result) - + # 进度回调:去重消歧完成,传递去重和消歧的具体效果 await self._send_dedup_progress_callback( len(entity_nodes), len(final_entity_nodes), len(statement_entity_edges), len(final_statement_entity_edges), len(entity_entity_edges), len(final_entity_entity_edges) ) - # 写入提取结果汇总(试运行和正式模式都需要生成) try: @@ -1428,10 +1698,10 @@ class ExtractionOrchestrator: raise def _save_dedup_details( - self, - dedup_details: Dict[str, Any], - original_entities: List[ExtractedEntityNode], - final_entities: List[ExtractedEntityNode] + self, + dedup_details: Dict[str, Any], + original_entities: List[ExtractedEntityNode], + final_entities: List[ExtractedEntityNode] ): """ 保存去重消歧的详细记录到实例变量(基于内存数据结构) @@ -1444,7 +1714,7 @@ class ExtractionOrchestrator: try: # 保存ID重定向映射 self.id_redirect_map = dedup_details.get("id_redirect", {}) - + # 处理精确匹配的合并记录 exact_merge_map = dedup_details.get("exact_merge_map", {}) for key, info in exact_merge_map.items(): @@ -1458,7 +1728,7 @@ class ExtractionOrchestrator: "merged_count": len(merged_ids), "merged_ids": list(merged_ids) }) - + # 处理模糊匹配的合并记录 fuzzy_merge_records = dedup_details.get("fuzzy_merge_records", []) for record in fuzzy_merge_records: @@ -1478,7 +1748,7 @@ class ExtractionOrchestrator: }) except Exception as e: logger.debug(f"解析模糊匹配记录失败: {record}, 错误: {e}") - + # 处理LLM去重的合并记录 llm_decision_records = dedup_details.get("llm_decision_records", []) for record in llm_decision_records: @@ -1497,7 +1767,7 @@ class ExtractionOrchestrator: }) except Exception as e: logger.debug(f"解析LLM去重记录失败: {record}, 错误: {e}") - + # 处理消歧记录 disamb_records = dedup_details.get("disamb_records", []) for record in disamb_records: @@ -1512,14 +1782,14 @@ class ExtractionOrchestrator: entity1_type = match.group(2) match.group(3).strip() entity2_type = match.group(4) - + # 提取置信度和原因 conf_match = re.search(r"conf=([0-9.]+)", str(record)) confidence = conf_match.group(1) if conf_match else "unknown" - + reason_match = re.search(r"reason=([^|]+)", str(record)) reason = reason_match.group(1).strip() if reason_match else "" - + self.dedup_disamb_records.append({ "entity_name": entity1_name, "disamb_type": f"消歧阻断:{entity1_type} vs {entity2_type}", @@ -1528,16 +1798,17 @@ class ExtractionOrchestrator: }) except Exception as e: logger.debug(f"解析消歧记录失败: {record}, 错误: {e}") - - logger.info(f"保存去重消歧记录:{len(self.dedup_merge_records)} 个合并记录,{len(self.dedup_disamb_records)} 个消歧记录") - + + logger.info( + f"保存去重消歧记录:{len(self.dedup_merge_records)} 个合并记录,{len(self.dedup_disamb_records)} 个消歧记录") + except Exception as e: logger.error(f"保存去重消歧详情失败: {e}", exc_info=True) async def _analyze_entity_merges( - self, - original_entities: List[ExtractedEntityNode], - final_entities: List[ExtractedEntityNode] + self, + original_entities: List[ExtractedEntityNode], + final_entities: List[ExtractedEntityNode] ) -> List[Dict[str, Any]]: """ 分析实体合并情况,直接使用内存中的合并记录(不再解析日志文件) @@ -1558,28 +1829,28 @@ class ExtractionOrchestrator: key=lambda x: x.get("merged_count", 0), reverse=True ) - + merge_info = [] for record in sorted_records: merge_info.append({ "main_entity_name": record.get("entity_name", "未知实体"), "merged_count": record.get("merged_count", 1) }) - + return merge_info - + # 如果没有保存的记录,返回空列表 logger.info("未找到实体合并记录") return [] - + except Exception as e: logger.error(f"分析实体合并情况失败: {e}", exc_info=True) return [] async def _analyze_entity_disambiguation( - self, - original_entities: List[ExtractedEntityNode], - final_entities: List[ExtractedEntityNode] + self, + original_entities: List[ExtractedEntityNode], + final_entities: List[ExtractedEntityNode] ) -> List[Dict[str, Any]]: """ 分析实体消歧情况,直接使用内存中的消歧记录(不再解析日志文件) @@ -1595,11 +1866,11 @@ class ExtractionOrchestrator: # 直接使用保存的消歧记录 if self.dedup_disamb_records: return self.dedup_disamb_records - + # 如果没有保存的记录,返回空列表 logger.info("未找到实体消歧记录") return [] - + except Exception as e: logger.error(f"分析实体消歧情况失败: {e}", exc_info=True) return [] @@ -1616,7 +1887,7 @@ class ExtractionOrchestrator: """ type_mapping = { "Person": "人物实体节点", - "Organization": "组织实体节点", + "Organization": "组织实体节点", "ORG": "组织实体节点", "Location": "地点实体节点", "LOC": "地点实体节点", @@ -1637,9 +1908,9 @@ class ExtractionOrchestrator: return type_mapping.get(entity_type, f"{entity_type}实体节点") async def _output_relationship_creation_results( - self, - entity_entity_edges: List[EntityEntityEdge], - entity_nodes: List[ExtractedEntityNode] + self, + entity_entity_edges: List[EntityEntityEdge], + entity_nodes: List[ExtractedEntityNode] ): """ 输出关系创建结果 @@ -1651,13 +1922,13 @@ class ExtractionOrchestrator: try: # 创建实体ID到名称的映射 entity_id_to_name = {node.id: node.name for node in entity_nodes} - + # 输出关系创建结果 for i, edge in enumerate(entity_entity_edges[:10]): # 只输出前10个关系 source_name = entity_id_to_name.get(edge.source, f"Entity_{edge.source}") target_name = entity_id_to_name.get(edge.target, f"Entity_{edge.target}") relation_type = edge.relation_type - + relationship_result = { "result_type": "relationship_creation", "relationship_index": i + 1, @@ -1666,20 +1937,20 @@ class ExtractionOrchestrator: "target_entity": target_name, "relationship_text": f"{source_name} -[{relation_type}]-> {target_name}" } - + await self.progress_callback("creating_nodes_edges_result", "关系创建", relationship_result) - + except Exception as e: logger.error(f"输出关系创建结果失败: {e}", exc_info=True) async def _send_dedup_progress_callback( - self, - original_entities: int, - final_entities: int, - original_stmt_edges: int, - final_stmt_edges: int, - original_ent_edges: int, - final_ent_edges: int, + self, + original_entities: int, + final_entities: int, + original_stmt_edges: int, + final_stmt_edges: int, + original_ent_edges: int, + final_ent_edges: int, ): """ 发送去重消歧完成的进度回调,传递具体的去重和消歧效果 @@ -1695,19 +1966,20 @@ class ExtractionOrchestrator: try: # 解析去重消歧报告文件,获取具体的去重和消歧效果 dedup_details = await self._parse_dedup_report() - + # 计算去重效果统计 entities_reduced = original_entities - final_entities stmt_edges_reduced = original_stmt_edges - final_stmt_edges ent_edges_reduced = original_ent_edges - final_ent_edges - + # 构建进度回调数据 dedup_stats = { "entities": { "original_count": original_entities, "final_count": final_entities, "reduced_count": entities_reduced, - "reduction_rate": round(entities_reduced / original_entities * 100, 1) if original_entities > 0 else 0, + "reduction_rate": round(entities_reduced / original_entities * 100, + 1) if original_entities > 0 else 0, }, "statement_entity_edges": { "original_count": original_stmt_edges, @@ -1726,9 +1998,9 @@ class ExtractionOrchestrator: "total_disambiguations": dedup_details.get("total_disambiguations", 0), } } - + await self.progress_callback("dedup_disambiguation_complete", "去重消歧完成", dedup_stats) - + except Exception as e: logger.error(f"发送去重消歧进度回调失败: {e}", exc_info=True) # 即使解析失败,也发送基本的统计信息 @@ -1758,12 +2030,12 @@ class ExtractionOrchestrator: disamb_examples = [] total_merges = 0 total_disambiguations = 0 - + # 处理合并记录 for record in self.dedup_merge_records: merge_count = record.get("merged_count", 0) total_merges += merge_count - + dedup_examples.append({ "type": record.get("type", "未知"), "entity_name": record.get("entity_name", "未知实体"), @@ -1771,30 +2043,31 @@ class ExtractionOrchestrator: "merge_count": merge_count, "description": f"{record.get('entity_name', '未知实体')}实体去重合并{merge_count}个" }) - + # 处理消歧记录 for record in self.dedup_disamb_records: total_disambiguations += 1 - + # 从消歧类型中提取实体类型信息 disamb_type = record.get("disamb_type", "") entity_name = record.get("entity_name", "未知实体") - + disamb_examples.append({ "entity1_name": entity_name, - "entity1_type": disamb_type.split("vs")[0].replace("消歧阻断:", "").strip() if "vs" in disamb_type else "未知", + "entity1_type": disamb_type.split("vs")[0].replace("消歧阻断:", + "").strip() if "vs" in disamb_type else "未知", "entity2_name": entity_name, "entity2_type": disamb_type.split("vs")[1].strip() if "vs" in disamb_type else "未知", "description": f"{entity_name},消歧区分成功" }) - + return { "dedup_examples": dedup_examples[:5], # 只返回前5个示例 "disamb_examples": disamb_examples[:5], # 只返回前5个示例 "total_merges": total_merges, "total_disambiguations": total_disambiguations, } - + except Exception as e: logger.error(f"获取去重报告失败: {e}", exc_info=True) return {"dedup_examples": [], "disamb_examples": [], "total_merges": 0, "total_disambiguations": 0} @@ -1807,9 +2080,9 @@ class ExtractionOrchestrator: async def get_chunked_dialogs( - chunker_strategy: str = "RecursiveChunker", - end_user_id: str = "group_1", - indices: Optional[List[int]] = None, + chunker_strategy: str = "RecursiveChunker", + end_user_id: str = "group_1", + indices: Optional[List[int]] = None, ) -> List[DialogData]: """从测试数据生成分块对话 @@ -1823,7 +2096,7 @@ async def get_chunked_dialogs( """ import json import re - + # 加载测试数据 testdata_path = os.path.join(os.path.dirname(__file__), "../../data", "testdata.json") with open(testdata_path, "r", encoding="utf-8") as f: @@ -1837,7 +2110,7 @@ async def get_chunked_dialogs( else: # 默认使用所有数据 selected_data = test_data - + for data in selected_data: # 解析对话上下文 context_text = data["context"] @@ -1853,7 +2126,7 @@ async def get_chunked_dialogs( if m: y, mo, d = int(m.group(1)), int(m.group(2)), int(m.group(3)) conv_date = f"{y:04d}-{mo:02d}-{d:02d}" - + dialog_metadata: Dict[str, Any] = {} if conv_date: dialog_metadata["conversation_date"] = conv_date @@ -1882,7 +2155,7 @@ async def get_chunked_dialogs( end_user_id=end_user_id, metadata=dialog_metadata, ) - + # 创建分块器并处理对话 from app.core.memory.storage_services.extraction_engine.knowledge_extraction.chunk_extraction import ( DialogueChunker, @@ -1905,7 +2178,7 @@ async def get_chunked_dialogs( from app.core.config import settings settings.ensure_memory_output_dir() output_path = settings.get_memory_output_path("chunker_test_output.txt") - + import json with open(output_path, "w", encoding="utf-8") as f: json.dump( @@ -1916,10 +2189,10 @@ async def get_chunked_dialogs( def preprocess_data( - input_path: Optional[str] = None, - output_path: Optional[str] = None, - skip_cleaning: bool = True, - indices: Optional[List[int]] = None + input_path: Optional[str] = None, + output_path: Optional[str] = None, + skip_cleaning: bool = True, + indices: Optional[List[int]] = None ) -> List[DialogData]: """数据预处理 @@ -1938,7 +2211,8 @@ def preprocess_data( ) preprocessor = DataPreprocessor() try: - cleaned_data = preprocessor.preprocess(input_path=input_path, output_path=output_path, skip_cleaning=skip_cleaning, indices=indices) + cleaned_data = preprocessor.preprocess(input_path=input_path, output_path=output_path, + skip_cleaning=skip_cleaning, indices=indices) logger.debug(f"数据预处理完成!共处理了 {len(cleaned_data)} 条对话数据") return cleaned_data except Exception as e: @@ -1947,9 +2221,9 @@ def preprocess_data( async def get_chunked_dialogs_from_preprocessed( - data: List[DialogData], - chunker_strategy: str = "RecursiveChunker", - llm_client: Optional[Any] = None, + data: List[DialogData], + chunker_strategy: str = "RecursiveChunker", + llm_client: Optional[Any] = None, ) -> List[DialogData]: """从预处理后的数据中生成分块 @@ -1964,31 +2238,31 @@ async def get_chunked_dialogs_from_preprocessed( logger.debug(f"=== 批量对话分块处理 (使用 {chunker_strategy}) ===") if not data: raise ValueError("预处理数据为空,无法进行分块") - + all_chunked_dialogs: List[DialogData] = [] from app.core.memory.storage_services.extraction_engine.knowledge_extraction.chunk_extraction import ( DialogueChunker, ) - + for dialog_data in data: chunker = DialogueChunker(chunker_strategy, llm_client=llm_client) chunks = await chunker.process_dialogue(dialog_data) dialog_data.chunks = chunks all_chunked_dialogs.append(dialog_data) - + return all_chunked_dialogs async def get_chunked_dialogs_with_preprocessing( - chunker_strategy: str = "RecursiveChunker", - end_user_id: str = "default", - user_id: str = "default", - apply_id: str = "default", - indices: Optional[List[int]] = None, - input_data_path: Optional[str] = None, - llm_client: Optional[Any] = None, - skip_cleaning: bool = True, - pruning_config: Optional[Dict] = None, + chunker_strategy: str = "RecursiveChunker", + end_user_id: str = "default", + user_id: str = "default", + apply_id: str = "default", + indices: Optional[List[int]] = None, + input_data_path: Optional[str] = None, + llm_client: Optional[Any] = None, + skip_cleaning: bool = True, + pruning_config: Optional[Dict] = None, ) -> List[DialogData]: """包含数据预处理步骤的完整分块流程 @@ -2012,7 +2286,7 @@ async def get_chunked_dialogs_with_preprocessing( input_data_path = os.path.join( os.path.dirname(__file__), "../../data", "testdata.json" ) - + # 步骤1: 数据预处理(包含索引筛选) from app.core.config import settings settings.ensure_memory_output_dir() @@ -2022,37 +2296,38 @@ async def get_chunked_dialogs_with_preprocessing( skip_cleaning=skip_cleaning, indices=indices, ) - + # 设置 end_user_id for dd in preprocessed_data: dd.end_user_id = end_user_id - + # 步骤2: 语义剪枝 try: from app.core.memory.storage_services.extraction_engine.data_preprocessing.data_pruning import ( SemanticPruner, ) from app.core.memory.models.config_models import PruningConfig - + # 构建剪枝配置 if pruning_config: # 使用传入的配置 config = PruningConfig(**pruning_config) - logger.debug(f"[剪枝] 使用传入配置: switch={config.pruning_switch}, scene={config.pruning_scene}, threshold={config.pruning_threshold}") + logger.debug( + f"[剪枝] 使用传入配置: switch={config.pruning_switch}, scene={config.pruning_scene}, threshold={config.pruning_threshold}") else: # 使用默认配置(关闭剪枝) config = None logger.debug("[剪枝] 未提供配置,使用默认配置(剪枝关闭)") - + pruner = SemanticPruner(config=config, llm_client=llm_client) - + # 记录单对话场景下剪枝前的消息数量 single_dialog_original_msgs = None if len(preprocessed_data) == 1 and preprocessed_data[0].context: single_dialog_original_msgs = len(preprocessed_data[0].context.msgs) preprocessed_data = await pruner.prune_dataset(preprocessed_data) - + # 单对话:打印清洗与剪枝信息 if len(preprocessed_data) == 1 and single_dialog_original_msgs is not None: remaining_msgs = len(preprocessed_data[0].context.msgs) if preprocessed_data[0].context else 0 @@ -2063,7 +2338,7 @@ async def get_chunked_dialogs_with_preprocessing( ) else: logger.debug(f"语义剪枝完成!剩余 {len(preprocessed_data)} 条对话") - + # 保存剪枝后的数据 try: from app.core.memory.storage_services.extraction_engine.data_preprocessing.data_preprocessor import ( @@ -2076,7 +2351,7 @@ async def get_chunked_dialogs_with_preprocessing( logger.error(f"保存剪枝结果失败:{se}") except Exception as e: logger.error(f"语义剪枝过程中出现错误,跳过剪枝: {e}") - + # 步骤3: 对话分块 return await get_chunked_dialogs_from_preprocessed( preprocessed_data, diff --git a/api/app/core/memory/storage_services/extraction_engine/knowledge_extraction/embedding_generation.py b/api/app/core/memory/storage_services/extraction_engine/knowledge_extraction/embedding_generation.py index 72f3641e..33838061 100644 --- a/api/app/core/memory/storage_services/extraction_engine/knowledge_extraction/embedding_generation.py +++ b/api/app/core/memory/storage_services/extraction_engine/knowledge_extraction/embedding_generation.py @@ -5,8 +5,11 @@ """ import asyncio +import logging from typing import Any, Dict, List, Tuple +logger = logging.getLogger(__name__) + from app.core.memory.llm_tools.openai_embedder import OpenAIEmbedderClient from app.core.memory.models.message_models import DialogData from app.core.models.base import RedBearModelConfig @@ -48,9 +51,9 @@ class EmbeddingGenerator: return await self.embedder_client.response(texts) # 分批并行处理 - print(f"文本数量 {len(texts)} 超过批次大小 {batch_size},分批并行处理") + logger.info(f"文本数量 {len(texts)} 超过批次大小 {batch_size},分批并行处理") batches = [texts[i:i+batch_size] for i in range(0, len(texts), batch_size)] - print(f"分成 {len(batches)} 批,每批最多 {batch_size} 个文本") + logger.info(f"分成 {len(batches)} 批,每批最多 {batch_size} 个文本") # 并行发送所有批次 batch_results = await asyncio.gather(*[ @@ -62,7 +65,7 @@ class EmbeddingGenerator: for batch_result in batch_results: embeddings.extend(batch_result) - print(f"分批并行处理完成,共生成 {len(embeddings)} 个嵌入向量") + logger.info(f"分批并行处理完成,共生成 {len(embeddings)} 个嵌入向量") return embeddings async def generate_statement_embeddings( @@ -77,7 +80,7 @@ class EmbeddingGenerator: Returns: 每个对话的陈述句嵌入向量映射列表 """ - print("\n=== 生成陈述句嵌入向量 ===") + logger.debug("=== 生成陈述句嵌入向量 ===") # 收集所有陈述句 all_statements = [] @@ -102,7 +105,7 @@ class EmbeddingGenerator: stmt_id = chunked_dialogs[d_idx].chunks[c_idx].statements[s_idx].id stmt_embedding_maps[d_idx][stmt_id] = embedding - print(f"为 {len(all_statements)} 个陈述句生成了嵌入向量") + logger.info(f"为 {len(all_statements)} 个陈述句生成了嵌入向量") return stmt_embedding_maps async def generate_chunk_embeddings( @@ -117,7 +120,7 @@ class EmbeddingGenerator: Returns: 每个对话的分块嵌入向量映射列表 """ - print("\n=== 生成分块嵌入向量 ===") + logger.debug("=== 生成分块嵌入向量 ===") # 收集所有分块 all_chunks = [] @@ -138,7 +141,7 @@ class EmbeddingGenerator: chunk_id = chunked_dialogs[d_idx].chunks[c_idx].id chunk_embedding_maps[d_idx][chunk_id] = embedding - print(f"为 {len(all_chunks)} 个分块生成了嵌入向量") + logger.info(f"为 {len(all_chunks)} 个分块生成了嵌入向量") return chunk_embedding_maps async def generate_dialog_embeddings( @@ -172,7 +175,7 @@ class EmbeddingGenerator: Returns: (陈述句嵌入映射列表, 分块嵌入映射列表, 对话嵌入列表) """ - print("\n=== 生成所有嵌入向量 ===") + logger.debug("=== 生成所有嵌入向量 ===") # 并发生成陈述句和分块嵌入向量 stmt_embedding_maps, chunk_embedding_maps = await asyncio.gather( @@ -183,9 +186,7 @@ class EmbeddingGenerator: # 对话嵌入向量(当前跳过) dialog_embeddings = await self.generate_dialog_embeddings(chunked_dialogs) - print( - f"生成完成:{len(chunked_dialogs)} 个对话的嵌入向量" - ) + logger.info(f"生成完成:{len(chunked_dialogs)} 个对话的嵌入向量") return stmt_embedding_maps, chunk_embedding_maps, dialog_embeddings @@ -201,7 +202,7 @@ class EmbeddingGenerator: Returns: 更新后的三元组映射列表(实体包含嵌入向量) """ - print("\n=== 生成实体嵌入向量 ===") + logger.debug("=== 生成实体嵌入向量 ===") entity_texts: List[str] = [] entity_refs: List[Any] = [] @@ -219,7 +220,7 @@ class EmbeddingGenerator: entity_refs.append(ent) if not entity_texts: - print("没有找到需要生成嵌入向量的实体") + logger.debug("没有找到需要生成嵌入向量的实体") return triplet_maps # 批量生成嵌入向量 @@ -227,13 +228,13 @@ class EmbeddingGenerator: # 打印前几个嵌入向量的维度 for i in range(min(5, len(embeddings))): - print(f"实体 '{entity_texts[i]}' 嵌入向量维度: {len(embeddings[i])}") + logger.debug(f"实体 '{entity_texts[i]}' 嵌入向量维度: {len(embeddings[i])}") # 将嵌入向量赋值给实体 for ent, emb in zip(entity_refs, embeddings): setattr(ent, "name_embedding", emb) - print(f"为 {len(entity_refs)} 个实体生成了嵌入向量") + logger.info(f"为 {len(entity_refs)} 个实体生成了嵌入向量") return triplet_maps @@ -296,7 +297,7 @@ async def embedding_generation_all( Returns: (陈述句嵌入映射列表, 分块嵌入映射列表, 对话嵌入列表, 更新后的三元组映射列表) """ - print("\n=== 综合嵌入向量生成(陈述句/分块/对话 + 实体)===") + logger.debug("=== 综合嵌入向量生成(陈述句/分块/对话 + 实体)===") generator = EmbeddingGenerator(embedding_id) diff --git a/api/app/core/memory/storage_services/extraction_engine/knowledge_extraction/memory_summary.py b/api/app/core/memory/storage_services/extraction_engine/knowledge_extraction/memory_summary.py index 443ee36a..5e39ba36 100644 --- a/api/app/core/memory/storage_services/extraction_engine/knowledge_extraction/memory_summary.py +++ b/api/app/core/memory/storage_services/extraction_engine/knowledge_extraction/memory_summary.py @@ -188,7 +188,6 @@ async def _process_chunk_summary( response_model=MemorySummaryResponse, ) summary_text = structured.summary.strip() - # Generate title and type for the summary title = None episodic_type = None diff --git a/api/app/core/memory/utils/prompt/prompts/extracat_Pruning.jinja2 b/api/app/core/memory/utils/prompt/prompts/extracat_Pruning.jinja2 index e204b7f9..3061e663 100644 --- a/api/app/core/memory/utils/prompt/prompts/extracat_Pruning.jinja2 +++ b/api/app/core/memory/utils/prompt/prompts/extracat_Pruning.jinja2 @@ -1,6 +1,7 @@ {# 对话级抽取与相关性判定模板(用于剪枝加速) - 输入:pruning_scene, ontology_classes, dialog_text, language + 输入:pruning_scene, ontology_class_infos, dialog_text, language + - ontology_class_infos: List[{class_name: str, class_description: str}] 输出:严格 JSON(不要包含任何多余文本),字段: - is_related: bool,是否与所选场景相关 - times: [string],从对话中抽取的时间相关文本(日期、时间、时间段、有效期等) @@ -18,20 +19,16 @@ #} {# ── 确定场景说明 ── #} -{% if ontology_classes and ontology_classes | length > 0 %} +{% if ontology_class_infos and ontology_class_infos | length > 0 %} {% if language == 'en' %} - {% set custom_types_str = ontology_classes | join(', ') %} - {% set instruction = 'Scene "' ~ pruning_scene ~ '": The dialogue is related to this scene if it involves any of the following entity types: ' ~ custom_types_str ~ '.' %} + {% set instruction = 'Scene "' ~ pruning_scene ~ '": The dialogue is relevant if it involves any of the following entity types.' %} {% else %} - {% set custom_types_str = ontology_classes | join('、') %} - {% set instruction = '场景「' ~ pruning_scene ~ '」:对话涉及以下任意实体类型时视为相关:' ~ custom_types_str ~ '。' %} + {% set instruction = '场景「' ~ pruning_scene ~ '」:对话涉及以下任意实体类型时视为相关。' %} {% endif %} {% else %} {% if language == 'en' %} - {% set custom_types_str = '' %} {% set instruction = 'Scene "' ~ pruning_scene ~ '": Determine whether the dialogue content is relevant to this scene based on overall context.' %} {% else %} - {% set custom_types_str = '' %} {% set instruction = '场景「' ~ pruning_scene ~ '」:根据对话整体内容判断是否与该场景相关。' %} {% endif %} {% endif %} @@ -42,8 +39,17 @@ 2. 从对话中抽取所有需要保留的重要信息片段。 场景说明:{{ instruction }} -{% if custom_types_str %} -重要提示:只要对话中出现与上述实体类型({{ custom_types_str }})相关的内容,即判定为相关(is_related=true)。 + +{% if ontology_class_infos and ontology_class_infos | length > 0 %} +【本场景实体类型定义】 +以下实体类型定义了本场景中哪些内容是重要的。 +凡是与以下任意类型相关的内容,都必须保留,并将关键词/短语提取到 keywords 字段: + +{% for info in ontology_class_infos %} +- {{ info.class_name }}:{{ info.class_description }} +{% endfor %} + +重要提示:只要对话中出现与上述任意实体类型相关的内容,即判定为相关(is_related=true)。 {% endif %} --- @@ -51,13 +57,40 @@ 以下类型的内容无论是否与场景直接相关,都必须保留,请将其关键词/短语抽取到对应字段: - 时间信息:日期、时间点、时间段、有效期 → times 字段 - 编号信息:学号、工号、订单号、申请号、账号、ID → ids 字段 -- 金额信息:价格、费用、金额(含货币符号或单位) → amounts 字段 +- 金额信息:价格、费用、金额(含货币符号或单位,如"100元"、"¥200")→ amounts 字段(注意:考试分数、成绩分数不属于金额,不要放入此字段) - 联系方式:电话、手机号、邮箱、微信、QQ → contacts 字段 - 地址信息:地点、地址、位置 → addresses 字段 -- 场景关键词:与场景强相关的专业术语、事件名称 → keywords 字段 +- 场景关键词:与**当前场景**强相关的专业术语、事件名称 → keywords 字段(注意:只放与当前场景直接相关的词,跨场景的内容不要放入此字段) - **情绪与情感**:喜悦、悲伤、愤怒、焦虑、开心、难过、委屈、兴奋、害怕、担心、压力、感动等情绪表达 → preserve_keywords 字段 - **兴趣与爱好**:喜欢、热爱、爱好、擅长、享受、沉迷、着迷、讨厌某事物等个人偏好表达 → preserve_keywords 字段 -- **个人观点与态度**:对某事物的明确看法、评价、立场 → preserve_keywords 字段 +- **个人情感态度**:对人际关系、情感状态的明确表达(如"我跟室友闹矛盾了"、"我都快抑郁了")→ preserve_keywords 字段 +- 注意:学业目标(如"我想考研")、成绩(如"87分")、学科偏好(如"喜欢数学")属于学业信息,不属于情绪/情感,不要放入 preserve_keywords 字段 + +【场景无关内容标记】 +请从对话中识别出与当前场景({{ pruning_scene }})**既不相关、也无语义关联**的消息片段,将其原文(或关键片段)提取到 scene_unrelated_snippets 字段。 +判断标准: +- 与场景实体类型完全无关 +- 与场景话题没有因果/时间/情境上的关联(例如:不是"因为上课所以累"这种关联) +- 纯粹是另一个话题的内容(如在教育场景中讨论购物、娱乐等) +注意:有情绪/感受表达的消息即使话题不同,也可能有语义关联,请谨慎标记。 + +**重要:scene_unrelated_snippets 必须认真填写,不能为空数组。** +如果对话中存在与场景无关的内容,必须将其原文片段提取出来。 + +示例(场景=在线教育): +- "我最近心情很差,跟室友闹矛盾了" → 与教育场景无关,加入 scene_unrelated_snippets +- "她总是很晚回来吵到我睡觉" → 与教育场景无关,加入 scene_unrelated_snippets +- "对,我都快抑郁了" → 与教育场景无关,加入 scene_unrelated_snippets +- "期末考试12月25日" → 与教育场景相关,不加入 scene_unrelated_snippets +- "我上次高数作业87分" → 与教育场景相关,不加入 scene_unrelated_snippets +- "我的目标是考研" → 与教育场景相关,不加入 scene_unrelated_snippets + +示例(场景=情感陪伴): +- "我最近心情很差,跟室友闹矛盾了" → 与情感陪伴场景相关(情绪+关系),不加入 scene_unrelated_snippets +- "对,我都快抑郁了" → 与情感陪伴场景相关(情绪),不加入 scene_unrelated_snippets +- "期末考试12月25日,3号教学楼201室" → 与情感陪伴场景无关(教育信息),加入 scene_unrelated_snippets +- "我上次高数作业87分,这次能考好吗" → 与情感陪伴场景无关(学业信息),加入 scene_unrelated_snippets +- "我的目标是考研,想读应用数学" → 与情感陪伴场景无关(学业目标),加入 scene_unrelated_snippets 【可以删除的内容】 以下类型的内容属于低价值信息,可以在剪枝时删除: @@ -88,7 +121,8 @@ "contacts": [...], "addresses": [...], "keywords": [...], - "preserve_keywords": [...] + "preserve_keywords": [...], + "scene_unrelated_snippets": [...] } {% else %} You are a dialogue content analysis assistant. Please analyze the full dialogue below in one pass and complete two tasks: @@ -96,8 +130,17 @@ You are a dialogue content analysis assistant. Please analyze the full dialogue 2. Extract all important information fragments that must be preserved. Scenario Description: {{ instruction }} -{% if custom_types_str %} -Important: If the dialogue contains content related to any of the entity types above ({{ custom_types_str }}), mark it as relevant (is_related=true). + +{% if ontology_class_infos and ontology_class_infos | length > 0 %} +[Scene Entity Type Definitions] +The following entity types define what content is important in this scene. +Content related to ANY of these types must be preserved and extracted into the keywords field: + +{% for info in ontology_class_infos %} +- {{ info.class_name }}: {{ info.class_description }} +{% endfor %} + +Important: If the dialogue contains content related to any of the entity types above, mark it as relevant (is_related=true). {% endif %} --- @@ -105,13 +148,22 @@ Important: If the dialogue contains content related to any of the entity types a The following types of content must always be preserved regardless of scene relevance. Extract their keywords/phrases into the corresponding fields: - Time information: dates, time points, durations, expiry dates → times field - ID information: student IDs, employee IDs, order numbers, application numbers, account IDs → ids field -- Amount information: prices, fees, amounts (with currency symbols or units) → amounts field +- Amount information: prices, fees, amounts (with currency symbols or units, e.g., "$100", "¥200") → amounts field (Note: exam scores and grades are NOT amounts, do not put them here) - Contact information: phone numbers, emails, WeChat, QQ → contacts field - Address information: locations, addresses, places → addresses field -- Scene keywords: professional terms and event names strongly related to the scene → keywords field +- Scene keywords: professional terms and event names strongly related to **the current scene** → keywords field (Note: only put terms directly related to the current scene; cross-scene content should not be placed here) - **Emotions and feelings**: joy, sadness, anger, anxiety, happiness, sadness, excitement, fear, worry, stress, being moved, etc. → preserve_keywords field - **Interests and hobbies**: likes, loves, hobbies, good at, enjoys, obsessed with, hates something, personal preferences → preserve_keywords field -- **Personal opinions and attitudes**: clear views, evaluations, or stances on something → preserve_keywords field +- **Personal emotional attitudes**: clear expressions about interpersonal relationships or emotional states (e.g., "I had a fight with my roommate", "I'm almost depressed") → preserve_keywords field +- Note: Academic goals (e.g., "I want to pursue a master's degree"), grades (e.g., "87 points"), and subject preferences (e.g., "I like math") are academic information, NOT emotions/feelings — do not put them in preserve_keywords + +[Scene-Unrelated Content Marking] +Please identify message snippets in the dialogue that are **neither relevant to nor semantically associated with** the current scene ({{ pruning_scene }}), and extract their original text (or key fragments) into the scene_unrelated_snippets field. +Criteria: +- Completely unrelated to the scene's entity types +- No causal/temporal/contextual association with the scene topic (e.g., "feeling tired because of class" IS associated) +- Purely belongs to a different topic (e.g., discussing shopping or entertainment in an education scene) +Note: Messages with emotional/feeling expressions may still have semantic association even if the topic differs — mark carefully. [CAN BE DELETED] The following types of content are low-value and can be removed during pruning: @@ -141,6 +193,7 @@ Output strict JSON only (fixed keys, order doesn't matter): "contacts": [...], "addresses": [...], "keywords": [...], - "preserve_keywords": [...] + "preserve_keywords": [...], + "scene_unrelated_snippets": [...] } {% endif %} diff --git a/api/app/core/memory/utils/prompt/prompts/extract_triplet.jinja2 b/api/app/core/memory/utils/prompt/prompts/extract_triplet.jinja2 index b2f287f4..6605532d 100644 --- a/api/app/core/memory/utils/prompt/prompts/extract_triplet.jinja2 +++ b/api/app/core/memory/utils/prompt/prompts/extract_triplet.jinja2 @@ -5,6 +5,15 @@ ===Task=== Extract entities and knowledge triplets from the given statement. +**⚠️ CRITICAL REQUIREMENTS:** +1. **ALIASES ORDER IS CRITICAL**: The FIRST alias in the array will be used as the user's primary display name (other_name). You MUST put the most important/frequently used name FIRST. +2. **ALWAYS include aliases field**: Even if empty, you MUST include "aliases": [] in EVERY entity. + + + {% if language == "zh" %} **重要:请使用中文生成实体名称(name)、描述(description)和示例(example)。** {% else %} @@ -18,34 +27,29 @@ Extract entities and knowledge triplets from the given statement. {% if ontology_types %} ===Ontology Type Guidance=== -**CRITICAL RULE: You MUST ONLY use the predefined ontology type names listed below for the entity "type" field. Do NOT use any other type names, even if they seem reasonable.** +**CRITICAL: Use ONLY predefined type names below. If no exact match, use CLOSEST type. NEVER invent new types.** -**If no predefined type fits an entity, use the CLOSEST matching predefined type. NEVER invent new type names.** +**Type Priority:** +1. [场景类型] Scene Types (domain-specific, prefer first) +2. [通用类型] General Types (standard ontologies) +3. [通用父类] Parent Types (hierarchy context) -**Type Priority (from highest to lowest):** -1. **[场景类型] Scene Types** - Domain-specific types, ALWAYS prefer these first -2. **[通用类型] General Types** - Common types from standard ontologies (DBpedia) -3. **[通用父类] Parent Types** - Provide type hierarchy context +**Rules:** +- Type MUST exactly match predefined names +- Do NOT modify, translate, or abbreviate type names +- Prefer scene types over general types -**Type Matching Rules:** -- Entity type MUST exactly match one of the predefined type names below -- Do NOT use types like "Equipment", "Component", "Concept", "Action", "Condition", "Data", "Duration" unless they appear in the predefined list -- Do NOT modify, translate, abbreviate, or create variations of type names -- Prefer scene types (marked [场景类型]) over general types when both could apply -- If uncertain, check the type description to find the best match - -**Predefined Ontology Types:** +**Predefined Types:** {{ ontology_types }} {% if type_hierarchy_hints %} -**Type Hierarchy Reference:** -The following shows type inheritance relationships (Child → Parent → Grandparent): +**Hierarchy:** {% for hint in type_hierarchy_hints %} - {{ hint }} {% endfor %} {% endif %} -**ALLOWED Type Names (use EXACTLY one of these, no exceptions):** +**ALLOWED Names:** {{ ontology_type_names | join(', ') }} {% endif %} @@ -62,66 +66,94 @@ The following shows type inheritance relationships (Child → Parent → Grandpa - **Entity descriptions must be in English** - **Examples must be in English** {% endif %} -- **Semantic Memory Classification (is_explicit_memory):** - * Set to `true` if the entity represents **explicit/semantic memory**: - - **Concepts:** "Machine Learning", "Photosynthesis", "Democracy" - - **Knowledge:** "Python Programming Language", "Theory of Relativity" - - **Definitions:** "API (Application Programming Interface)", "REST API" - - **Principles:** "SOLID Principles", "First Law of Thermodynamics" - - **Theories:** "Evolution Theory", "Quantum Mechanics" - - **Methods/Techniques:** "Agile Development", "Machine Learning Algorithm" - - **Technical Terms:** "Neural Network", "Database" - * Set to `false` for: - - **People:** "John Smith", "Dr. Wang" - - **Organizations:** "Microsoft", "Harvard University" - - **Locations:** "Beijing", "Central Park" - - **Events:** "2024 Conference", "Project Meeting" - - **Specific objects:** "iPhone 15", "Building A" -- **Example Generation (IMPORTANT for semantic memory entities):** - * For entities where `is_explicit_memory=true`, generate a **concise example (around 20 characters)** to help understand the concept - * The example should be: - - **Specific and concrete**: Use real-world scenarios or applications - - **Brief**: Around 20 characters (can be slightly longer if needed for clarity) +- **Semantic Memory (is_explicit_memory):** + * `true` for: Concepts, Knowledge, Definitions, Theories, Methods (e.g., "Machine Learning", "REST API") + * `false` for: People, Organizations, Locations, Events, Specific objects + * For `is_explicit_memory=true`, provide concise example (~20 chars{% if language == "zh" %},使用中文{% endif %}) + +**🚨🚨🚨 ALIASES & DENIED_ALIASES - MANDATORY FIELDS 🚨🚨🚨** + +**CRITICAL RULES (违反将导致提取失败):** + +1. **EVERY entity MUST have aliases field:** + - `"aliases": [...]` - REQUIRED, even if empty `[]` + +2. **ALIASES - 别名提取规则:** {% if language == "zh" %} - - **使用中文** + - 包含:昵称、全名、简称、别称、网名等 + - 顺序:**第一个别名将作为用户的主显示名称(other_name),必须把最重要/最常用的名字放在第一位** + - 提取顺序:严格按照对话中首次出现的顺序 + - 示例: + * "我叫张三,大家叫我小张" → aliases=["张三", "小张"](张三是第一个,将成为 other_name) + * "大家叫我小李,我全名叫李明" → aliases=["小李", "李明"](小李先出现,将成为 other_name) + - 空值:如果没有别名,使用 `[]` + - 重要:只提取本次对话中明确提到的别名,不要推测或添加未提及的名字 {% else %} - - **In English** + - Include: nicknames, full names, abbreviations, alternative names + - Order: **The FIRST alias will be used as the user's primary display name (other_name). Put the most important/frequently used name FIRST** + - Extraction order: Strictly follow the order of first appearance in conversation + - Examples: + * "I'm John, people call me Johnny" → aliases=["John", "Johnny"] (John is first, will become other_name) + * "People call me Mike, my full name is Michael" → aliases=["Mike", "Michael"] (Mike appears first, will become other_name) + - Empty: If no aliases, use `[]` + - Important: Only extract aliases explicitly mentioned in current conversation, do not infer or add unmentioned names {% endif %} - * For non-semantic entities (`is_explicit_memory=false`), the example field can be empty -- **Aliases Extraction:** + + + +3. **USER ENTITY SPECIAL HANDLING:** {% if language == "zh" %} - * 别名使用中文 + - 用户实体的 name 字段:使用 "用户" 或 "我" + - 用户的真实姓名:放入 aliases + - **🚨 禁止将 "用户"、"我" 放入 aliases 中,aliases 只能包含用户的真实姓名、昵称等** + - 示例: + * "我叫李明" → name="用户", aliases=["李明"] + * ❌ 错误:aliases=["用户", "李明"]("用户"不是真实姓名,禁止放入 aliases) + * ❌ 错误:aliases=["我", "李明"]("我"不是真实姓名,禁止放入 aliases) {% else %} - * Aliases should be in English + - User entity name field: use "User" or "I" + - User's real name: put in aliases + - **🚨 NEVER put "User" or "I" in aliases. Aliases must only contain real names, nicknames, etc.** + - Examples: + * "I'm John" → name="User", aliases=["John"] + * ❌ Wrong: aliases=["User", "John"] ("User" is not a real name, FORBIDDEN in aliases) + * ❌ Wrong: aliases=["I", "John"] ("I" is not a real name, FORBIDDEN in aliases) {% endif %} - * Include common alternative names, abbreviations and full names - * If no aliases exist, use empty array: [] -- Exclude lengthy quotes, calendar dates, temporal ranges, and temporal expressions -- For numeric values: extract as separate entities (instance_of: 'Numeric', name: units, numeric_value: value) - Example: £30 → name: 'GBP', numeric_value: 30, instance_of: 'Numeric' + + + +4. **ALIASES ORDER:** +{% if language == "zh" %} + - 顺序优先级:按出现顺序,先出现的在前 +{% else %} + - Order priority: by appearance order, first mentioned comes first +{% endif %} + +**EXAMPLES OF CORRECT EXTRACTION:** +{% if language == "zh" %} +- "我叫张三" → aliases=["张三"] (张三将成为 other_name) +- "大家叫我小明,我全名叫李明" → aliases=["小明", "李明"] (小明先出现,将成为 other_name) +- "我是李华,网名叫华仔" → aliases=["李华", "华仔"] (李华先出现,将成为 other_name) +{% else %} +- "I'm John" → aliases=["John"] (John will become other_name) +- "People call me Mike, my full name is Michael" → aliases=["Mike", "Michael"] (Mike appears first, will become other_name) +- "I'm John Smith, username JSmith" → aliases=["John Smith", "JSmith"] (John Smith appears first, will become other_name) +{% endif %} + +- Exclude lengthy quotes, dates, temporal expressions +- Numeric values: extract as entities (instance_of: 'Numeric', name: units, numeric_value: value) **Triplet Extraction:** -- Extract (subject, predicate, object) triplets where: - - Subject: main entity performing the action or being described - - Predicate: relationship between entities (e.g., 'is', 'works at', 'believes') - - Object: entity, value, or concept affected by the predicate +- Extract (subject, predicate, object) where subject/object are entities, predicate is relationship {% if language == "zh" %} -- subject_name 和 object_name 必须使用中文 +- subject_name 和 object_name 使用中文 {% else %} -- subject_name and object_name must be in English (translate if original is in another language) +- subject_name and object_name in English {% endif %} -- Exclude all temporal expressions from every field -- Use ONLY the predicates listed in "Predicate Instructions" (uppercase English tokens) -- Do NOT translate predicate tokens -- Do NOT include `statement_id` field (assigned automatically) - -**When NOT to extract triplets:** -- Non-propositional utterances (emotions, fillers, onomatopoeia) -- No clear predicate from the given definitions applies -- Standalone noun phrases or checklist items → extract as entities only -- Do NOT invent generic predicates (e.g., "IS_DOING", "FEELS", "MENTIONS") - -**If no valid triplet exists:** Return triplets: [], extract entities if present, otherwise both arrays empty. +- Use ONLY predicates from "Predicate Instructions" (uppercase tokens) +- Exclude temporal expressions, do NOT include `statement_id` +- **When NOT to extract:** emotions, fillers, no clear predicate, standalone nouns +- **If no valid triplet:** Return triplets: [] {%- if predicate_instructions -%} **Predicate Instructions:** @@ -207,26 +239,44 @@ Output: {"entity_idx": 0, "name": "三脚架", "type": "Equipment", "description": "摄影器材配件", "example": "", "aliases": ["相机三脚架"], "is_explicit_memory": false} ] } + +**Example 4 (别名 - Chinese):** "我的名字是乐力齐,我的小名是齐齐,同事们都叫我小乐" +Output: +{ + "triplets": [], + "entities": [ + {"entity_idx": 0, "name": "用户", "type": "Person", "description": "用户本人", "example": "", "aliases": ["乐力齐", "齐齐", "小乐"], "is_explicit_memory": false} + ] +} + +**Example 5 (别名顺序 - Chinese):** "我叫陈思远。对了,我的网名叫「远山」" +Output: +{ + "triplets": [], + "entities": [ + {"entity_idx": 0, "name": "用户", "type": "Person", "description": "用户本人", "example": "", "aliases": ["陈思远", "远山"], "is_explicit_memory": false} + ] +} + + {% endif %} ===End of Examples=== {% if ontology_types %} -**⚠️ REMINDER: The examples above use generic type names for illustration only. You MUST use ONLY the predefined ontology type names from the "ALLOWED Type Names" list above. For example, use "PredictiveMaintenance" instead of "Concept", use "ProductionLine" instead of "Equipment", etc. Map each entity to the closest matching predefined type.** +**⚠️ REMINDER: Examples use generic types for illustration. You MUST use predefined types from "ALLOWED Names" above.** {% endif %} ===Output Format=== **JSON Requirements:** -- Use only ASCII double quotes (") for JSON structure -- Never use Chinese quotation marks ("") or Unicode quotes -- Escape quotation marks in text with backslashes (\") -- Ensure proper string closure and comma separation -- No line breaks within JSON string values +- Use ASCII double quotes ("), escape with \" +- No Chinese quotes (""), no line breaks in strings {% if language == "zh" %} -- **语言要求:实体名称(name)、描述(description)、示例(example)、subject_name、object_name 必须使用中文** +- **语言:name、description、example、subject_name、object_name 使用中文** {% else %} -- **Language Requirement: Entity names, descriptions, examples, subject_name, object_name must be in English** -- **If the original text is in Chinese, translate all names to English** +- **Language: names, descriptions, examples in English (translate if needed)** {% endif %} +- **⚠️ ALIASES ORDER: preserve temporal order of appearance** +- **🚨 MANDATORY FIELD: EVERY entity MUST include "aliases" field, even if empty array []** {{ json_schema }} diff --git a/api/app/core/models/__init__.py b/api/app/core/models/__init__.py index f54afc08..f98d073f 100644 --- a/api/app/core/models/__init__.py +++ b/api/app/core/models/__init__.py @@ -2,6 +2,7 @@ from .base import RedBearModelConfig, get_provider_llm_class, RedBearModelFacto from .llm import RedBearLLM from .embedding import RedBearEmbeddings from .rerank import RedBearRerank +from .generation import RedBearImageGenerator, RedBearVideoGenerator __all__ = [ "RedBearModelConfig", @@ -9,5 +10,7 @@ __all__ = [ "RedBearEmbeddings", "RedBearRerank", "RedBearModelFactory", - "get_provider_llm_class" + "get_provider_llm_class", + "RedBearImageGenerator", + "RedBearVideoGenerator" ] \ No newline at end of file diff --git a/api/app/core/models/base.py b/api/app/core/models/base.py index 4a453c6b..80117f27 100644 --- a/api/app/core/models/base.py +++ b/api/app/core/models/base.py @@ -67,7 +67,7 @@ class RedBearModelFactory: **config.extra_params } - if provider in [ModelProvider.OPENAI, ModelProvider.XINFERENCE, ModelProvider.GPUSTACK, ModelProvider.OLLAMA]: + if provider in [ModelProvider.OPENAI, ModelProvider.XINFERENCE, ModelProvider.GPUSTACK, ModelProvider.OLLAMA, ModelProvider.VOLCANO]: # 使用 httpx.Timeout 对象来设置详细的超时配置 # 这样可以分别控制连接超时和读取超时 import httpx @@ -160,11 +160,13 @@ def get_provider_llm_class(config: RedBearModelConfig, type: ModelType = ModelTy # dashscope 的 omni 模型使用 OpenAI 兼容模式 if provider == ModelProvider.DASHSCOPE and config.is_omni: return ChatOpenAI - if provider in [ModelProvider.OPENAI, ModelProvider.XINFERENCE, ModelProvider.GPUSTACK]: + if provider in [ModelProvider.OPENAI, ModelProvider.XINFERENCE, ModelProvider.GPUSTACK, ModelProvider.VOLCANO]: if type == ModelType.LLM: return OpenAI elif type == ModelType.CHAT: return ChatOpenAI + else: + raise BusinessException(f"不支持的模型提供商及类型: {provider}-{type}", code=BizCode.PROVIDER_NOT_SUPPORTED) elif provider == ModelProvider.DASHSCOPE: return ChatTongyi elif provider == ModelProvider.OLLAMA: diff --git a/api/app/core/models/embedding.py b/api/app/core/models/embedding.py index 16af2567..3269e1d0 100644 --- a/api/app/core/models/embedding.py +++ b/api/app/core/models/embedding.py @@ -1,23 +1,190 @@ -from typing import Any, Dict, List, Optional, TypeVar, Callable +from typing import Any, Dict, List, Optional, Union from langchain_core.embeddings import Embeddings -from app.core.models.base import RedBearModelConfig,get_provider_embedding_class,RedBearModelFactory +from app.core.models.base import RedBearModelConfig, get_provider_embedding_class, RedBearModelFactory +from app.models.models_model import ModelProvider + class RedBearEmbeddings(Embeddings): - """Embedding → 完全符合 LangChain Embeddings""" + """统一的 Embedding 类,自动支持多模态(根据 provider 判断)""" + def __init__(self, config: RedBearModelConfig): - self._model = self._create_model(config) self._config = config + self._is_volcano = config.provider.lower() == ModelProvider.VOLCANO + + if self._is_volcano: + # 火山引擎使用 Ark SDK + self._client = self._create_volcano_client(config) + self._model = None + else: + # 其他 provider 使用 LangChain + self._model = self._create_model(config) + self._client = None def _create_model(self, config: RedBearModelConfig) -> Embeddings: - """根据配置创建模型""" + """根据配置创建 LangChain 模型""" embedding_class = get_provider_embedding_class(config.provider) model_params = RedBearModelFactory.get_model_params(config) return embedding_class(**model_params) + + def _create_volcano_client(self, config: RedBearModelConfig): + """创建火山引擎客户端""" + from volcenginesdkarkruntime import Ark + return Ark(api_key=config.api_key, base_url=config.base_url) + # ==================== LangChain 标准接口 ==================== + def embed_documents(self, texts: list[str]) -> list[list[float]]: - return self._model.embed_documents(texts) + """批量文本向量化(LangChain 标准接口)""" + if self._is_volcano: + # 火山引擎多模态 Embedding + contents = [{"type": "text", "text": text} for text in texts] + response = self._client.multimodal_embeddings.create( + model=self._config.model_name, + input=contents, + encoding_format="float" + ) + return [response.data.embedding] + else: + # 其他 provider + return self._model.embed_documents(texts) def embed_query(self, text: str) -> List[float]: - return self._model.embed_query(text) + """单个文本向量化(LangChain 标准接口)""" + if self._is_volcano: + # 火山引擎多模态 Embedding + result = self.embed_documents([text]) + return result[0] if result else [] + else: + # 其他 provider + return self._model.embed_query(text) + + # ==================== 多模态扩展方法 ==================== + + def embed_multimodal( + self, + contents: List[Dict[str, Any]], + **kwargs + ) -> List[List[float]]: + """ + 多模态向量化(仅火山引擎支持) + + Args: + contents: 内容列表,格式: + - 文本: {"type": "text", "text": "..."} + - 图片: {"type": "image_url", "image_url": {"url": "..."}} + - 视频: {"type": "video_url", "video_url": {"url": "..."}} + **kwargs: 其他参数 + + Returns: + 向量列表 + """ + if not self._is_volcano: + raise NotImplementedError( + f"多模态 Embedding 仅支持火山引擎,当前 provider: {self._config.provider}" + ) + + response = self._client.multimodal_embeddings.create( + model=self._config.model_name, + input=contents, + **kwargs + ) + return [response.data.embedding] + + async def aembed_multimodal( + self, + contents: List[Dict[str, Any]], + **kwargs + ) -> List[List[float]]: + """异步多模态向量化""" + # 火山引擎 SDK 暂不支持异步,使用同步方法 + return self.embed_multimodal(contents, **kwargs) + + def embed_text(self, text: str, **kwargs) -> List[float]: + """文本向量化(便捷方法)""" + if self._is_volcano: + result = self.embed_multimodal( + [{"type": "text", "text": text}], + **kwargs + ) + return result[0] if result else [] + else: + return self.embed_query(text) + + def embed_image(self, image_url: str, **kwargs) -> List[float]: + """图片向量化(仅火山引擎支持)""" + if not self._is_volcano: + raise NotImplementedError( + f"图片向量化仅支持火山引擎,当前 provider: {self._config.provider}" + ) + + result = self.embed_multimodal( + [{"type": "image_url", "image_url": {"url": image_url}}], + **kwargs + ) + return result[0] if result else [] + + def embed_video(self, video_url: str, **kwargs) -> List[float]: + """视频向量化(仅火山引擎支持)""" + if not self._is_volcano: + raise NotImplementedError( + f"视频向量化仅支持火山引擎,当前 provider: {self._config.provider}" + ) + + result = self.embed_multimodal( + [{"type": "video_url", "video_url": {"url": video_url}}], + **kwargs + ) + return result[0] if result else [] + + def embed_batch( + self, + items: List[Union[str, Dict[str, Any]]], + **kwargs + ) -> List[List[float]]: + """ + 批量向量化(支持混合类型) + + Args: + items: 可以是字符串列表或内容字典列表 + **kwargs: 其他参数 + + Returns: + 向量列表 + """ + # 如果全是字符串,使用标准方法 + if all(isinstance(item, str) for item in items): + return self.embed_documents(items) + + # 如果包含字典,需要多模态支持 + if not self._is_volcano: + raise NotImplementedError( + f"混合类型批量向量化仅支持火山引擎,当前 provider: {self._config.provider}" + ) + + # 标准化输入格式 + contents = [] + for item in items: + if isinstance(item, str): + contents.append({"type": "text", "text": item}) + elif isinstance(item, dict): + contents.append(item) + else: + raise ValueError(f"不支持的输入类型: {type(item)}") + + return self.embed_multimodal(contents, **kwargs) + + # ==================== 工具方法 ==================== + + def is_multimodal_supported(self) -> bool: + """检查是否支持多模态""" + return self._is_volcano + + def get_provider(self) -> str: + """获取 provider""" + return self._config.provider + + +# 保留 RedBearMultimodalEmbeddings 作为别名,向后兼容 +RedBearMultimodalEmbeddings = RedBearEmbeddings diff --git a/api/app/core/models/generation.py b/api/app/core/models/generation.py new file mode 100644 index 00000000..b6388d3f --- /dev/null +++ b/api/app/core/models/generation.py @@ -0,0 +1,344 @@ +""" +图片和视频生成模型封装 + +支持的 Provider: +- Volcano (火山引擎): 使用 volcenginesdkarkruntime +- OpenAI: 使用 openai SDK +""" +from typing import Any, Dict, Optional + +from volcenginesdkarkruntime import Ark +from volcenginesdkarkruntime.types.images.images import ( + SequentialImageGenerationOptions, + ContentGenerationTool, + OptimizePromptOptions +) + +from app.core.models.base import RedBearModelConfig +from app.core.exceptions import BusinessException +from app.core.error_codes import BizCode +from app.models.models_model import ModelProvider + + +class RedBearImageGenerator: + """图片生成模型封装""" + + def __init__(self, config: RedBearModelConfig): + self._config = config + self._client = self._create_client(config) + + def _create_client(self, config: RedBearModelConfig): + """根据 provider 创建客户端""" + provider = config.provider.lower() + + if provider == ModelProvider.VOLCANO: + return Ark(api_key=config.api_key, base_url=config.base_url) + # elif provider == ModelProvider.OPENAI: + # from openai import OpenAI + # return OpenAI(api_key=config.api_key, base_url=config.base_url) + else: + raise BusinessException( + f"不支持的图片生成提供商: {provider}", + code=BizCode.PROVIDER_NOT_SUPPORTED + ) + + def generate( + self, + prompt: str, + image: Optional[Any] = None, + size: Optional[str] = "2K", + output_format: str = "png", + response_format: str = "url", + watermark: bool = False, + sequential_image_generation: Optional[str] = None, + sequential_image_generation_options: Optional[Dict] = None, + tools: Optional[list] = None, + optimize_prompt_options: Optional[Dict] = None, + stream: bool = False, + **kwargs + ) -> Dict[str, Any]: + """ + 生成图片 + + Args: + prompt: 提示词 + image: 参考图片URL或URL列表(图文生图/多图融合) + size: 图片尺寸,支持 "2K", "2048x2048", "1920x1080" 等(至少3686400像素) + output_format: 输出格式,如 "png", "jpg" + response_format: 返回格式,"url" 或 "b64_json" + watermark: 是否添加水印 + sequential_image_generation: 组图生成模式,"auto" 或 "disabled" + sequential_image_generation_options: 组图生成选项,如 {"max_images": 4} + tools: 工具列表,如 [{"type": "web_search"}] 用于联网搜索生图 + optimize_prompt_options: 提示词优化选项,如 {"mode": "fast"} + stream: 是否使用流式生成 + **kwargs: 其他参数 + + Returns: + 生成结果 + """ + provider = self._config.provider.lower() + + if provider == ModelProvider.VOLCANO: + params = { + "model": self._config.model_name, + "prompt": prompt, + "size": size, + "output_format": output_format, + "response_format": response_format, + "watermark": watermark, + } + + if image is not None: + params["image"] = image + + if sequential_image_generation: + params["sequential_image_generation"] = sequential_image_generation + if sequential_image_generation_options: + params["sequential_image_generation_options"] = SequentialImageGenerationOptions( + **sequential_image_generation_options + ) + + if tools: + params["tools"] = [ContentGenerationTool(**tool) if isinstance(tool, dict) else tool for tool in tools] + + if optimize_prompt_options: + params["optimize_prompt_options"] = OptimizePromptOptions(**optimize_prompt_options) + + if stream: + params["stream"] = True + + params.update(kwargs) + response = self._client.images.generate(**params) + + # elif provider == ModelProvider.OPENAI: + # response = self._client.images.generate( + # model=self._config.model_name, + # prompt=prompt, + # size=size, + # n=n, + # **kwargs + # ) + else: + raise BusinessException( + f"不支持的提供商: {provider}", + code=BizCode.PROVIDER_NOT_SUPPORTED + ) + + return response.model_dump() if hasattr(response, 'model_dump') else response + + async def agenerate( + self, + prompt: str, + image: Optional[Any] = None, + size: Optional[str] = "2K", + output_format: str = "png", + response_format: str = "url", + watermark: bool = False, + **kwargs + ) -> Dict[str, Any]: + """异步生成图片""" + return self.generate(prompt, image, size, output_format, response_format, watermark, **kwargs) + + +class RedBearVideoGenerator: + """视频生成模型封装""" + + def __init__(self, config: RedBearModelConfig): + self._config = config + self._client = self._create_client(config) + + def _create_client(self, config: RedBearModelConfig): + """根据 provider 创建客户端""" + provider = config.provider.lower() + + if provider == ModelProvider.VOLCANO: + return Ark(api_key=config.api_key, base_url=config.base_url) + else: + raise BusinessException( + f"不支持的视频生成提供商: {provider}", + code=BizCode.PROVIDER_NOT_SUPPORTED + ) + + def generate( + self, + prompt: str, + image_url: Optional[str] = None, + first_frame_url: Optional[str] = None, + last_frame_url: Optional[str] = None, + reference_images: Optional[list] = None, + draft_task_id: Optional[str] = None, + duration: Optional[int] = None, + frames: Optional[int] = None, + ratio: Optional[str] = None, + resolution: Optional[str] = None, + generate_audio: bool = False, + watermark: bool = False, + camera_fixed: bool = False, + seed: Optional[int] = None, + return_last_frame: bool = False, + service_tier: str = "default", + execution_expires_after: Optional[int] = None, + draft: bool = False, + **kwargs + ) -> Dict[str, Any]: + """ + 生成视频 + + Args: + prompt: 提示词 + image_url: 首帧图片URL(图生视频-基于首帧) + first_frame_url: 首帧图片URL(图生视频-基于首尾帧) + last_frame_url: 尾帧图片URL(图生视频-基于首尾帧) + reference_images: 参考图片URL列表(图生视频-基于参考图) + draft_task_id: Draft任务ID(基于Draft生成正式视频) + duration: 视频时长(秒),与frames二选一 + frames: 视频帧数,与duration二选一 + ratio: 视频比例,如 "16:9", "9:16", "adaptive" + resolution: 视频分辨率,如 "720p", "1080p" + generate_audio: 是否生成音频 + watermark: 是否添加水印 + camera_fixed: 是否固定镜头 + seed: 随机种子 + return_last_frame: 是否返回最后一帧 + service_tier: 服务层级,"default" 或 "flex"(离线推理) + execution_expires_after: 任务过期时间(秒) + draft: 是否生成样片 + **kwargs: 其他参数 + + Returns: + 生成结果(包含任务ID,需要轮询获取结果) + """ + provider = self._config.provider.lower() + + if provider == ModelProvider.VOLCANO: + content = [{"type": "text", "text": prompt}] + + if draft_task_id: + content = [{"type": "draft_task", "draft_task": {"id": draft_task_id}}] + else: + if image_url: + content.append({"type": "image_url", "image_url": {"url": image_url}}) + + if first_frame_url: + content.append({"type": "image_url", "image_url": {"url": first_frame_url}, "role": "first_frame"}) + if last_frame_url: + content.append({"type": "image_url", "image_url": {"url": last_frame_url}, "role": "last_frame"}) + + if reference_images: + for ref_url in reference_images: + content.append({"type": "image_url", "image_url": {"url": ref_url}, "role": "reference_image"}) + + params = {"model": self._config.model_name, "content": content, "watermark": watermark} + + if duration: + params["duration"] = duration + if frames: + params["frames"] = frames + if ratio: + params["ratio"] = ratio + if resolution: + params["resolution"] = resolution + if generate_audio: + params["generate_audio"] = generate_audio + if camera_fixed: + params["camera_fixed"] = camera_fixed + if seed is not None: + params["seed"] = seed + if return_last_frame: + params["return_last_frame"] = return_last_frame + if service_tier != "default": + params["service_tier"] = service_tier + if execution_expires_after: + params["execution_expires_after"] = execution_expires_after + if draft: + params["draft"] = draft + + params.update(kwargs) + response = self._client.content_generation.tasks.create(**params) + else: + raise BusinessException( + f"不支持的提供商: {provider}", + code=BizCode.PROVIDER_NOT_SUPPORTED + ) + + return response.model_dump() if hasattr(response, 'model_dump') else response + + async def agenerate( + self, + prompt: str, + image_url: Optional[str] = None, + duration: Optional[int] = None, + **kwargs + ) -> Dict[str, Any]: + """异步生成视频""" + return self.generate(prompt, image_url=image_url, duration=duration, **kwargs) + + def get_task_status(self, task_id: str) -> Dict[str, Any]: + """ + 查询视频生成任务状态 + + Args: + task_id: 任务ID + + Returns: + 任务状态信息 + """ + provider = self._config.provider.lower() + + if provider == ModelProvider.VOLCANO: + response = self._client.content_generation.tasks.get(task_id=task_id) + return response.model_dump() if hasattr(response, 'model_dump') else response + else: + raise BusinessException( + f"不支持的提供商: {provider}", + code=BizCode.PROVIDER_NOT_SUPPORTED + ) + + async def aget_task_status(self, task_id: str) -> Dict[str, Any]: + """异步查询任务状态""" + return self.get_task_status(task_id) + + def list_tasks(self, page_size: int = 10, status: Optional[str] = None, **kwargs) -> Dict[str, Any]: + """ + 查询视频生成任务列表 + + Args: + page_size: 每页数量 + status: 任务状态筛选,如 "succeeded", "failed", "pending" + **kwargs: 其他参数 + + Returns: + 任务列表 + """ + provider = self._config.provider.lower() + + if provider == ModelProvider.VOLCANO: + params = {"page_size": page_size} + if status: + params["status"] = status + params.update(kwargs) + response = self._client.content_generation.tasks.list(**params) + return response.model_dump() if hasattr(response, 'model_dump') else response + else: + raise BusinessException( + f"不支持的提供商: {provider}", + code=BizCode.PROVIDER_NOT_SUPPORTED + ) + + def delete_task(self, task_id: str) -> None: + """ + 删除或取消视频生成任务 + + Args: + task_id: 任务ID + """ + provider = self._config.provider.lower() + + if provider == ModelProvider.VOLCANO: + self._client.content_generation.tasks.delete(task_id=task_id) + else: + raise BusinessException( + f"不支持的提供商: {provider}", + code=BizCode.PROVIDER_NOT_SUPPORTED + ) diff --git a/api/app/core/models/scripts/volcano_models.yaml b/api/app/core/models/scripts/volcano_models.yaml new file mode 100644 index 00000000..24609f5a --- /dev/null +++ b/api/app/core/models/scripts/volcano_models.yaml @@ -0,0 +1,334 @@ +provider: volcano +models: +# Doubao-Seed 2.0 系列 +- name: doubao-seed-2-0-pro-260215 + type: chat + provider: volcano + description: 旗舰级全能通用模型,面向 Agent 时代的复杂推理与长链路任务执行场景。强调多模态理解、长上下文推理、结构化生成与工具增强执行。复杂指令与多约束执行能力突出,可稳定应对多步复杂规划、复杂图文推理、视频内容理解与高难度分析等场景。侧重长链路推理能力与复杂任务稳定性,适配真实业务中的复杂场景。 + is_deprecated: false + is_official: true + capability: + - vision + - video + is_omni: false + tags: + - 大语言模型 + logo: volcano + +- name: doubao-seed-2-0-lite-260215 + type: chat + provider: volcano + description: 面向高频企业场景兼顾性能与成本的均衡型模型,综合能力超越上一代Doubao-Seed-1.8。胜任非结构化信息处理、内容创作、搜索推荐、数据分析等生产型工作,支持长上下文、多源信息融合、多步指令执行与高保真结构化输出。在保障稳定效果的同时显著优化成本。兼顾生成质量与响应速度,适合作为通用生产级模型。 + is_deprecated: false + is_official: true + capability: + - vision + - video + is_omni: false + tags: + - 大语言模型 + logo: volcano + +- name: doubao-seed-2-0-mini-260215 + type: chat + provider: volcano + description: 面向低时延、高并发与成本敏感场景,提供极致的模型推理速度。模型效果与Doubao-Seed-1.6相当。支持256k上下文、4档思考长度和多模态理解,适合成本和速度优先的轻量级任务。 + is_deprecated: false + is_official: true + capability: + - vision + - video + is_omni: false + tags: + - 大语言模型 + logo: volcano + +- name: doubao-seed-2-0-code-preview-260215 + type: chat + provider: volcano + description: 面向真实编程环境优化的 Coding 模型,能稳定调用 Claude Code 等常见 IDE 中的工具。模型特别优化了前端能力,在使用常见的前端框架时能有良好表现。模型支持使用 Skills,可以配合多种自定义技能使用。Seed 2.0 的编程加强版,更适合 Agentic Coding。 + is_deprecated: false + is_official: true + capability: + - vision + - video + is_omni: false + tags: + - 大语言模型 + - 代码模型 + logo: volcano + +# Doubao-Seed 1.x 系列 +- name: doubao-seed-1-8-251228 + type: chat + provider: volcano + description: Doubao-Seed-1.8 面向多模态 Agent 场景定向优化。Agent 能力上,Tool Use、复杂指令遵循等能力均大幅增强。多模态理解方面,视觉基础能力显著提升,可低帧率理解超长视频,视频运动理解、复杂空间理解及文档结构化解析能力也有所优化,还原生支持智能上下文管理,用户可配置上下文策略。 + is_deprecated: false + is_official: true + capability: + - vision + - video + is_omni: false + tags: + - 大语言模型 + logo: volcano + +- name: doubao-seed-1-6-251015 + type: chat + provider: volcano + description: Doubao-Seed-1.6全新多模态深度思考模型,同时支持minimal/low/medium/high 四种reasoning effort。 更强模型效果,服务复杂任务和有挑战场景。支持 256k 上下文窗口,输出长度支持最大 32k tokens。 + is_deprecated: false + is_official: true + capability: + - vision + - video + is_omni: false + tags: + - 大语言模型 + logo: volcano + +- name: doubao-seed-1-6-lite-251015 + type: chat + provider: volcano + description: 更高性价比,常见任务的最佳选择,支持minimal、low、medium、high 四种reasoning_effort思考深度 + is_deprecated: false + is_official: true + capability: + - vision + - video + is_omni: false + tags: + - 大语言模型 + logo: volcano + +- name: doubao-seed-1-6-flash-250828 + type: chat + provider: volcano + description: Doubao-Seed-1.6-flash推理速度极致的多模态深度思考模型,TPOT低至10ms; 同时支持文本和视觉理解,文本理解能力超过上一代lite,视觉理解比肩友商pro系列模型。支持 256k 上下文窗口,输出长度支持最大 16k tokens。 + is_deprecated: false + is_official: true + capability: + - vision + - video + is_omni: false + tags: + - 大语言模型 + logo: volcano + +- name: doubao-seed-code-preview-251028 + type: chat + provider: volcano + description: 面向Agentic编程任务进行了深度优化。 + is_deprecated: false + is_official: true + capability: + - vision + - video + is_omni: false + tags: + - 大语言模型 + - 代码模型 + logo: volcano + +- name: doubao-seed-1-6-vision-250815 + type: chat + provider: volcano + description: 全新Doubao-Seed-1.6系列视觉深度思考模型,视觉理解能力显著增强,并支持image_process视觉工具 + is_deprecated: false + is_official: true + capability: + - vision + - video + is_omni: false + tags: + - 大语言模型 + - 多模态模型 + logo: volcano + +# Doubao 1.5 系列 +- name: doubao-1-5-vision-pro-32k-250115 + type: chat + provider: volcano + description: 全新升级的多模态大模型,支持任意分辨率和极端长宽比图像识别,增强视觉推理、文档识别、细节信息理解和指令遵循能力。支持 32k 上下文窗口,输出长度支持最大 12k tokens。 + is_deprecated: false + is_official: true + capability: + - vision + is_omni: false + tags: + - 大语言模型 + - 多模态模型 + logo: volcano + +- name: doubao-1-5-pro-32k-250115 + type: chat + provider: volcano + description: 全新一代主力模型,性能全面升级,在知识、代码、推理等方面表现卓越。最大支持 128k 上下文窗口,输出长度支持最大 12k tokens。 + is_deprecated: false + is_official: true + capability: [] + is_omni: false + tags: + - 大语言模型 + logo: volcano + +- name: doubao-1-5-lite-32k-250115 + type: chat + provider: volcano + description: 全新一代轻量版模型,极致响应速度,效果与时延均达到全球一流水平。支持 32k 上下文窗口,输出长度支持最大 12k tokens。 + is_deprecated: false + is_official: true + capability: [] + is_omni: false + tags: + - 大语言模型 + logo: volcano + +# Doubao-Seedance 视频生成系列 +- name: doubao-seedance-1-5-pro-251215 + type: video + provider: volcano + description: 豆包视频生成模型Seedance 1.5 pro 作为全球领先的视频生成模型,可生成音画高精同步的视频内容。支持多人多语言对白,全面覆盖环境音、动作音、合成音、乐器音、背景音及人声,支持首尾帧,实现影视级叙事效果,满足影视、漫剧、电商及广告领域的高阶创作需求。 + is_deprecated: false + is_official: true + capability: + - vision + is_omni: false + tags: + - 视频生成 + logo: volcano + +- name: doubao-seedance-1-0-pro-250528 + type: video + provider: volcano + description: 一款支持多镜头叙事的视频生成基础模型,在各维度表现出色。它在语义理解与指令遵循能力上取得突破,能生成运动流畅、细节丰富、风格多样且具备影视级美感的 1080P 高清视频 + is_deprecated: false + is_official: true + capability: + - vision + is_omni: false + tags: + - 视频生成 + logo: volcano + +- name: doubao-seedance-1-0-pro-fast-251015 + type: video + provider: volcano + description: 一款价格触底、效能封顶的全面模型,在视频生成质量、速度、价格之间取得了卓越平衡。它继承了Seedance 1.0 pro 核心优势,同时生成速度提升、价格更具竞争力,为创作者带来效率与成本双重优化的体验。 + is_deprecated: false + is_official: true + capability: + - vision + is_omni: false + tags: + - 视频生成 + logo: volcano + +- name: doubao-seedance-1-0-lite-i2v-250428 + type: video + provider: volcano + description: 基于首帧图片、尾帧图片(可选)、参考图片(可选)和文本提示词(可选)相结合的方式生成视频 + is_deprecated: false + is_official: true + capability: + - vision + is_omni: false + tags: + - 视频生成 + - 图生视频 + logo: volcano + +- name: doubao-seedance-1-0-lite-t2v-250428 + type: video + provider: volcano + description: 基于文本提示词生成视频 + is_deprecated: false + is_official: true + capability: [] + is_omni: false + tags: + - 视频生成 + - 文生视频 + logo: volcano + +# Doubao-Seedream 图像生成系列 +- name: doubao-seedream-5-0-260128 + type: image + provider: volcano + description: 字节跳动发布的最新图像创作模型。该模型首次搭载联网检索功能,能融合实时网络信息,提升生图时效性。同时,模型的聪明度进一步升级,能够精准解析复杂指令和视觉内容。此外,模型在世界知识广度、参考一致性及专业场景生成质量上均有增强,可更好地满足企业级视觉创作需求。 + is_deprecated: false + is_official: true + capability: + - vision + is_omni: false + tags: + - 图像生成 + logo: volcano + +- name: doubao-seedream-4-5-251128 + type: image + provider: volcano + description: 字节跳动最新推出的图像多模态模型,整合了文生图、图生图、组图输出等能力,融合常识和推理能力。相比前代4.0模型生成效果大幅提升,具备更好的编辑一致性和多图融合效果,能更精准的控制画面细节,小字、小人脸生成更自然,图片排版、色彩更和谐,美感提升。 + is_deprecated: false + is_official: true + capability: + - vision + is_omni: false + tags: + - 图像生成 + logo: volcano + +- name: doubao-seedream-4-0-250828 + type: image + provider: volcano + description: 基于领先架构的SOTA级多模态图像创作模型,其生成美感、指令遵循、结构完整度、主体保持一致性处于世界头部水平。模型采用同一套架构实现文生图与编辑能力的统一,原生支持文本 、单图和多图输入,并能通过对提示词的深度推理,自动适配最优的图像比例尺寸与生成数量,可一次性连续输出最多 15 张内容关联的图像,支持 4K 超高清输出。 + is_deprecated: false + is_official: true + capability: + - vision + is_omni: false + tags: + - 图像生成 + logo: volcano + +- name: doubao-seedream-3-0-t2i-250415 + type: image + provider: volcano + description: 一款支持原生高分辨率的中英双语图像生成基础模型,综合能力媲美GPT-4o,处于世界第一梯队。支持原生 2K 分辨率输出;响应速度更快;小字生成更准确,文本排版效果增强;指令遵循能力强,美感&结构提升,保真度和细节表现较好。 + is_deprecated: false + is_official: true + capability: [] + is_omni: false + tags: + - 图像生成 + - 文生图 + logo: volcano + +# Doubao 翻译系列 +- name: doubao-seed-translation-250915 + type: chat + provider: volcano + description: 通用多语言翻译模型,支持30余种语言互译,支持 4K 上下文窗口,输出长度支持最大 3K tokens + is_deprecated: false + is_official: true + capability: [] + is_omni: false + tags: + - 翻译模型 + logo: volcano + +# Doubao Embedding 系列 +- name: doubao-embedding-vision-251215 + type: embedding + provider: volcano + description: 主要面向图文多模向量检索的使用场景,支持图片输入及中、英双语文本输入,最长 128K 上下文长度。 + is_deprecated: false + is_official: true + capability: + - vision + - video + is_omni: false + tags: + - 向量模型 + - 多模态模型 + logo: volcano diff --git a/api/app/core/rag/vdb/elasticsearch/elasticsearch_vector.py b/api/app/core/rag/vdb/elasticsearch/elasticsearch_vector.py index 198d1473..386920e0 100644 --- a/api/app/core/rag/vdb/elasticsearch/elasticsearch_vector.py +++ b/api/app/core/rag/vdb/elasticsearch/elasticsearch_vector.py @@ -61,24 +61,16 @@ class ElasticSearchConfig(BaseModel): class ElasticSearchVector(BaseVector): def __init__(self, index_name: str, config: ElasticSearchConfig, embedding_config: ModelApiKey, reranker_config: ModelApiKey): super().__init__(index_name.lower()) - # self.embeddings = XinferenceEmbeddings( - # server_url=os.getenv("XINFERENCE_URL", "http://127.0.0.1"), # Default Xinference port - # model_uid="bge-m3" # replace model_uid with the model UID return from launching the model - # ) - # Remove debug printing to avoid leaking sensitive information - # print("embedding:" + embedding_config.model_name + "|" + embedding_config.provider + "|" + embedding_config.api_key + "|" + embedding_config.api_base) + + # 初始化 Embedding 模型(自动支持火山引擎多模态) self.embeddings = RedBearEmbeddings(RedBearModelConfig( model_name=embedding_config.model_name, provider=embedding_config.provider, api_key=embedding_config.api_key, base_url=embedding_config.api_base )) - # self.reranker = XinferenceRerank( - # server_url=os.getenv("XINFERENCE_URL", "http://127.0.0.1"), - # model_uid="bge-reranker-large" - # ) - # Remove debug printing to avoid leaking sensitive information - # print("reranker:"+ reranker_config.model_name + "|" + reranker_config.provider + "|" + reranker_config.api_key + "|" + reranker_config.api_base) + self.is_multimodal_embedding = self.embeddings.is_multimodal_supported() + self.reranker = RedBearRerank(RedBearModelConfig( model_name=reranker_config.model_name, provider=reranker_config.provider, @@ -144,7 +136,11 @@ class ElasticSearchVector(BaseVector): def add_chunks(self, chunks: list[DocumentChunk], **kwargs): # 实现 Elasticsearch 保存向量 texts = [chunk.page_content for chunk in chunks] - embeddings = self.embeddings.embed_documents(list(texts)) + if self.is_multimodal_embedding: + # 火山引擎多模态 Embedding + embeddings = self.embeddings.embed_batch(texts) + else: + embeddings = self.embeddings.embed_documents(list(texts)) self.create(chunks, embeddings, **kwargs) def create(self, chunks: list[DocumentChunk], embeddings: list[list[float]], **kwargs): @@ -394,7 +390,11 @@ class ElasticSearchVector(BaseVector): updated count. """ indices = kwargs.get("indices", self._collection_name) # Default single index, multi-index available,etc "index1,index2,index3" - chunk.vector = self.embeddings.embed_query(chunk.page_content) + if self.is_multimodal_embedding: + # 火山引擎多模态 Embedding + chunk.vector = self.embeddings.embed_text(chunk.page_content) + else: + chunk.vector = self.embeddings.embed_query(chunk.page_content) body = { "script": { @@ -454,7 +454,11 @@ class ElasticSearchVector(BaseVector): def search_by_vector(self, query: str, **kwargs: Any) -> list[DocumentChunk]: """Search the nearest neighbors to a vector.""" - query_vector = self.embeddings.embed_query(query) + if self.is_multimodal_embedding: + # 火山引擎多模态 Embedding + query_vector = self.embeddings.embed_text(query) + else: + query_vector = self.embeddings.embed_query(query) top_k = kwargs.get("top_k", 1024) score_threshold = float(kwargs.get("score_threshold") or 0.3) indices = kwargs.get("indices", self._collection_name) # Default single index, multi-index available,etc "index1,index2,index3" diff --git a/api/app/core/storage/base.py b/api/app/core/storage/base.py index 8ab0fcde..09824c3f 100644 --- a/api/app/core/storage/base.py +++ b/api/app/core/storage/base.py @@ -109,17 +109,13 @@ class StorageBackend(ABC): pass @abstractmethod - async def get_url(self, file_key: str, expires: int = 3600) -> str: - """ - Get an access URL for the file. - - Args: - file_key: Unique identifier for the file in the storage system. - expires: URL validity period in seconds (default: 1 hour). - - Returns: - URL for accessing the file. - """ + async def get_url( + self, + file_key: str, + expires: int = 3600, + file_name: Optional[str] = None + ) -> str: + """Get an access URL for the file.""" pass async def get_permanent_url(self, file_key: str) -> Optional[str]: diff --git a/api/app/core/storage/local.py b/api/app/core/storage/local.py index 4b8ae829..13adfc20 100644 --- a/api/app/core/storage/local.py +++ b/api/app/core/storage/local.py @@ -210,7 +210,12 @@ class LocalStorage(StorageBackend): cause=e, ) - async def get_url(self, file_key: str, expires: int = 3600) -> str: + async def get_url( + self, + file_key: str, + expires: int = 3600, + file_name: Optional[str] = None + ) -> str: """ Get an access URL for the file. @@ -220,6 +225,7 @@ class LocalStorage(StorageBackend): Args: file_key: Unique identifier for the file in the storage system. expires: URL validity period in seconds (not used for local storage). + file_name: If set, adds Content-Disposition: attachment to force download. Returns: A relative URL path for accessing the file. diff --git a/api/app/core/storage/oss.py b/api/app/core/storage/oss.py index 27669ffa..c6c6ec48 100644 --- a/api/app/core/storage/oss.py +++ b/api/app/core/storage/oss.py @@ -7,6 +7,7 @@ Storage Service (OSS) using the oss2 SDK. import io import logging +import urllib.parse from typing import AsyncIterator, Optional import oss2 @@ -43,6 +44,8 @@ class OSSStorage(StorageBackend): access_key_id: str, access_key_secret: str, bucket_name: str, + connect_timeout: int = 30, + multipart_threshold: int = 10 * 1024 * 1024, # 10MB ): """ Initialize the OSSStorage backend. @@ -52,6 +55,8 @@ class OSSStorage(StorageBackend): access_key_id: The Aliyun access key ID. access_key_secret: The Aliyun access key secret. bucket_name: The name of the OSS bucket. + connect_timeout: Connection timeout in seconds (default: 30). + multipart_threshold: File size threshold for multipart upload (default: 10MB). Raises: StorageConfigError: If any required configuration is missing. @@ -68,10 +73,17 @@ class OSSStorage(StorageBackend): self.endpoint = endpoint self.bucket_name = bucket_name + self.multipart_threshold = multipart_threshold try: auth = oss2.Auth(access_key_id, access_key_secret) - self.bucket = oss2.Bucket(auth, endpoint, bucket_name) + # 设置超时和重试 + self.bucket = oss2.Bucket( + auth, + endpoint, + bucket_name, + connect_timeout=connect_timeout + ) logger.info( f"OSSStorage initialized with endpoint: {endpoint}, bucket: {bucket_name}" ) @@ -107,21 +119,38 @@ class OSSStorage(StorageBackend): if content_type: headers["Content-Type"] = content_type - self.bucket.put_object(file_key, content, headers=headers if headers else None) + # 大文件使用分片上传 + if len(content) > self.multipart_threshold: + logger.info(f"Using multipart upload for large file: {file_key} ({len(content)} bytes)") + upload_id = self.bucket.init_multipart_upload(file_key, headers=headers if headers else None).upload_id + parts = [] + part_size = 5 * 1024 * 1024 # 5MB per part + part_num = 1 + + for offset in range(0, len(content), part_size): + chunk = content[offset:offset + part_size] + result = self.bucket.upload_part(file_key, upload_id, part_num, chunk) + parts.append(oss2.models.PartInfo(part_num, result.etag)) + part_num += 1 + + self.bucket.complete_multipart_upload(file_key, upload_id, parts) + else: + self.bucket.put_object(file_key, content, headers=headers if headers else None) + logger.info(f"File uploaded to OSS successfully: {file_key}") return file_key except OssError as e: logger.error(f"OSS error uploading file {file_key}: {e}") raise StorageUploadError( - message=f"Failed to upload file to OSS: {e.message}", + message=f"Failed to upload file to OSS: {str(e)}", file_key=file_key, cause=e, ) except Exception as e: logger.error(f"Failed to upload file to OSS {file_key}: {e}") raise StorageUploadError( - message=f"Failed to upload file to OSS: {e}", + message=f"Failed to upload file to OSS: {str(e)}", file_key=file_key, cause=e, ) @@ -134,28 +163,73 @@ class OSSStorage(StorageBackend): ) -> int: """Upload from async stream to OSS. Returns total bytes written.""" buf = io.BytesIO() + headers = {"Content-Type": content_type} if content_type else None + upload_id = None + try: + # 收集流数据 + total_size = 0 async for chunk in stream: + if not chunk: + continue buf.write(chunk) + total_size += len(chunk) + content = buf.getvalue() - headers = {"Content-Type": content_type} if content_type else None - self.bucket.put_object(file_key, content, headers=headers) - logger.info(f"File stream uploaded to OSS successfully: {file_key}") - return len(content) + + if not content: + raise StorageUploadError( + message="Empty stream content", + file_key=file_key, + ) + + # 大文件使用分片上传 + if len(content) > self.multipart_threshold: + logger.info(f"Using multipart upload for stream: {file_key} ({len(content)} bytes)") + upload_id = self.bucket.init_multipart_upload(file_key, headers=headers).upload_id + parts = [] + part_size = 5 * 1024 * 1024 # 5MB + part_num = 1 + + for offset in range(0, len(content), part_size): + chunk = content[offset:offset + part_size] + result = self.bucket.upload_part(file_key, upload_id, part_num, chunk) + parts.append(oss2.models.PartInfo(part_num, result.etag)) + part_num += 1 + + self.bucket.complete_multipart_upload(file_key, upload_id, parts) + else: + self.bucket.put_object(file_key, content, headers=headers) + + logger.info(f"File stream uploaded to OSS successfully: {file_key} ({total_size} bytes)") + return total_size + except OssError as e: + if upload_id: + try: + self.bucket.abort_multipart_upload(file_key, upload_id) + except: + pass logger.error(f"OSS error stream uploading file {file_key}: {e}") raise StorageUploadError( - message=f"Failed to stream upload file to OSS: {e.message}", + message=f"Failed to stream upload file to OSS: {str(e)}", file_key=file_key, cause=e, ) except Exception as e: + if upload_id: + try: + self.bucket.abort_multipart_upload(file_key, upload_id) + except: + pass logger.error(f"Failed to stream upload file to OSS {file_key}: {e}") raise StorageUploadError( - message=f"Failed to stream upload file to OSS: {e}", + message=f"Failed to stream upload file to OSS: {str(e)}", file_key=file_key, cause=e, ) + finally: + buf.close() async def download(self, file_key: str) -> bytes: """ @@ -181,14 +255,14 @@ class OSSStorage(StorageBackend): except OssError as e: logger.error(f"OSS error downloading file {file_key}: {e}") raise StorageDownloadError( - message=f"Failed to download file from OSS: {e.message}", + message=f"Failed to download file from OSS: {str(e)}", file_key=file_key, cause=e, ) except Exception as e: logger.error(f"Failed to download file from OSS {file_key}: {e}") raise StorageDownloadError( - message=f"Failed to download file from OSS: {e}", + message=f"Failed to download file from OSS: {str(e)}", file_key=file_key, cause=e, ) @@ -214,14 +288,14 @@ class OSSStorage(StorageBackend): except OssError as e: logger.error(f"OSS error deleting file {file_key}: {e}") raise StorageDeleteError( - message=f"Failed to delete file from OSS: {e.message}", + message=f"Failed to delete file from OSS: {str(e)}", file_key=file_key, cause=e, ) except Exception as e: logger.error(f"Failed to delete file from OSS {file_key}: {e}") raise StorageDeleteError( - message=f"Failed to delete file from OSS: {e}", + message=f"Failed to delete file from OSS: {str(e)}", file_key=file_key, cause=e, ) @@ -242,24 +316,33 @@ class OSSStorage(StorageBackend): logger.error(f"Failed to check file existence in OSS {file_key}: {e}") return False - async def get_url(self, file_key: str, expires: int = 3600) -> str: + async def get_url( + self, + file_key: str, + expires: int = 3600, + file_name: Optional[str] = None, + ) -> str: """ Get a presigned URL for accessing the file. Args: file_key: Unique identifier for the file in the storage system. expires: URL validity period in seconds (default: 1 hour). + file_name: If set, adds Content-Disposition: attachment to force download. Returns: A presigned URL for accessing the file. """ try: - url = self.bucket.sign_url("GET", file_key, expires) + params = {} + if file_name: + filename_encoded = urllib.parse.quote(file_name.encode("utf-8")) + params["response-content-disposition"] = f"attachment; filename*=UTF-8''{filename_encoded}" + url = self.bucket.sign_url("GET", file_key, expires, params=params if params else None) logger.debug(f"Generated presigned URL for {file_key}, expires in {expires}s") return url except Exception as e: logger.error(f"Failed to generate presigned URL for {file_key}: {e}") - # Return a basic URL format as fallback return f"https://{self.bucket_name}.{self.endpoint.replace('https://', '').replace('http://', '')}/{file_key}" async def get_permanent_url(self, file_key: str) -> str: diff --git a/api/app/core/storage/s3.py b/api/app/core/storage/s3.py index c7b33ffe..f156f4a7 100644 --- a/api/app/core/storage/s3.py +++ b/api/app/core/storage/s3.py @@ -6,6 +6,7 @@ using the boto3 SDK. """ import io +import urllib.parse import logging from typing import AsyncIterator, Optional @@ -352,31 +353,37 @@ class S3Storage(StorageBackend): logger.error(f"Failed to check file existence in S3 {file_key}: {e}") return False - async def get_url(self, file_key: str, expires: int = 3600) -> str: + async def get_url( + self, + file_key: str, + expires: int = 3600, + file_name: Optional[str] = None, + ) -> str: """ Get a presigned URL for accessing the file. Args: file_key: Unique identifier for the file in the storage system. expires: URL validity period in seconds (default: 1 hour). + file_name: If set, adds Content-Disposition: attachment to force download. Returns: A presigned URL for accessing the file. """ try: + params = {"Bucket": self.bucket_name, "Key": file_key} + if file_name: + filename_encoded = urllib.parse.quote(file_name.encode("utf-8")) + params["ResponseContentDisposition"] = f"attachment; filename*=UTF-8''{filename_encoded}" url = self.client.generate_presigned_url( "get_object", - Params={ - "Bucket": self.bucket_name, - "Key": file_key, - }, + Params=params, ExpiresIn=expires, ) logger.debug(f"Generated presigned URL for {file_key}, expires in {expires}s") return url except Exception as e: logger.error(f"Failed to generate presigned URL for {file_key}: {e}") - # Return a basic URL format as fallback return f"https://{self.bucket_name}.s3.{self.region}.amazonaws.com/{file_key}" async def get_permanent_url(self, file_key: str) -> str: diff --git a/api/app/core/tools/mcp/client.py b/api/app/core/tools/mcp/client.py index 6df6df51..b437d021 100644 --- a/api/app/core/tools/mcp/client.py +++ b/api/app/core/tools/mcp/client.py @@ -99,7 +99,7 @@ class SimpleMCPClient: # 建立 SSE 连接 response = await self._session.get(self.server_url) - if response.status != 200: + if response.status not in (200, 202): error_text = await response.text() raise MCPConnectionError(f"SSE 连接失败 {response.status}: {error_text}") @@ -190,7 +190,9 @@ class SimpleMCPClient: try: async with self._session.post(self._endpoint_url, json=request) as response: - if response.status != 200: + # MCP SSE 协议:POST 请求返回 200 或 202 均为正常 + # 202 Accepted 表示请求已接受,结果通过 SSE 流异步返回 + if response.status not in (200, 202): error_text = await response.text() raise MCPConnectionError(f"请求失败 {response.status}: {error_text}") @@ -205,7 +207,7 @@ class SimpleMCPClient: raise MCPConnectionError("endpoint URL 未初始化") async with self._session.post(self._endpoint_url, json=notification) as response: - if response.status != 200: + if response.status not in (200, 202): logger.warning(f"通知发送失败: {response.status}") async def _initialize_modelscope_session(self): diff --git a/api/app/core/workflow/adapters/base_adapter.py b/api/app/core/workflow/adapters/base_adapter.py index 49321b89..2e24d085 100644 --- a/api/app/core/workflow/adapters/base_adapter.py +++ b/api/app/core/workflow/adapters/base_adapter.py @@ -9,7 +9,7 @@ from typing import Any from pydantic import BaseModel, Field -from app.core.workflow.adapters.errors import ExceptionDefineition +from app.core.workflow.adapters.errors import ExceptionDefinition from app.schemas.workflow_schema import ( EdgeDefinition, NodeDefinition, @@ -40,8 +40,8 @@ class WorkflowParserResult(BaseModel): edges: list[EdgeDefinition] = Field(default_factory=list) nodes: list[NodeDefinition] = Field(default_factory=list) variables: list[VariableDefinition] = Field(default_factory=list) - warnings: list[ExceptionDefineition] = Field(default_factory=list) - errors: list[ExceptionDefineition] = Field(default_factory=list) + warnings: list[ExceptionDefinition] = Field(default_factory=list) + errors: list[ExceptionDefinition] = Field(default_factory=list) class WorkflowImportResult(BaseModel): @@ -51,8 +51,8 @@ class WorkflowImportResult(BaseModel): edges: list[EdgeDefinition] = Field(default_factory=list) nodes: list[NodeDefinition] = Field(default_factory=list) variables: list[VariableDefinition] = Field(default_factory=list) - warnings: list[ExceptionDefineition] = Field(default_factory=list) - errors: list[ExceptionDefineition] = Field(default_factory=list) + warnings: list[ExceptionDefinition] = Field(default_factory=list) + errors: list[ExceptionDefinition] = Field(default_factory=list) class BasePlatformAdapter(ABC): diff --git a/api/app/core/workflow/adapters/dify/converter.py b/api/app/core/workflow/adapters/dify/converter.py index 467beb07..4fa9508b 100644 --- a/api/app/core/workflow/adapters/dify/converter.py +++ b/api/app/core/workflow/adapters/dify/converter.py @@ -9,9 +9,9 @@ from urllib.parse import quote from app.core.workflow.adapters.base_converter import BaseConverter from app.core.workflow.adapters.errors import ( - UnsupportVariableType, - UnknowModelWarning, - ExceptionDefineition, + UnsupportedVariableType, + UnknownModelWarning, + ExceptionDefinition, ExceptionType ) from app.core.workflow.nodes.assigner.config import AssignmentItem @@ -54,7 +54,7 @@ from app.core.workflow.nodes.http_request.config import ( HttpFormData, HttpTimeOutConfig, HttpRetryConfig, - HttpErrorDefaultTamplete, + HttpErrorDefaultTemplate, HttpErrorHandleConfig ) from app.core.workflow.nodes.if_else.config import ConditionDetail, ConditionBranchConfig @@ -108,7 +108,7 @@ class DifyConverter(BaseConverter): try: return config.model_validate(value) except Exception as e: - self.errors.append(ExceptionDefineition( + self.errors.append(ExceptionDefinition( type=ExceptionType.CONFIG, node_id=node_id, node_name=node_name, @@ -138,7 +138,7 @@ class DifyConverter(BaseConverter): var_selector = mapping.get(var_selector, var_selector) return var_selector - def _process_list_variable_litearl(self, variable_selector: list) -> str | None: + def _process_list_variable_literal(self, variable_selector: list) -> str | None: if not self.process_var_selector(".".join(variable_selector)): return None return "{{" + self.process_var_selector(".".join(variable_selector)) + "}}" @@ -269,7 +269,7 @@ class DifyConverter(BaseConverter): var_type = self.variable_type_map(var["type"]) if not var_type: self.errors.append( - UnsupportVariableType( + UnsupportedVariableType( scope=node["id"], name=var["variable"], var_type=var["type"], @@ -281,7 +281,7 @@ class DifyConverter(BaseConverter): if var_type in ["file", "array[file]"]: self.errors.append( - ExceptionDefineition( + ExceptionDefinition( type=ExceptionType.VARIABLE, node_id=node["id"], node_name=node_data["title"], @@ -311,7 +311,7 @@ class DifyConverter(BaseConverter): def convert_question_classifier_node_config(self, node: dict) -> dict: node_data = node["data"] self.warnings.append( - UnknowModelWarning( + UnknownModelWarning( node_id=node["id"], node_name=node_data["title"], model_name=node_data["model"].get("name") @@ -327,7 +327,7 @@ class DifyConverter(BaseConverter): ) result = QuestionClassifierNodeConfig.model_construct( - input_variable=self._process_list_variable_litearl(node_data.get("query_variable_selector")), + input_variable=self._process_list_variable_literal(node_data.get("query_variable_selector")), user_supplement_prompt=self.trans_variable_format(node_data.get("instructions", "")), categories=categories, ).model_dump() @@ -337,13 +337,13 @@ class DifyConverter(BaseConverter): def convert_llm_node_config(self, node: dict) -> dict: node_data = node["data"] self.warnings.append( - UnknowModelWarning( + UnknownModelWarning( node_id=node["id"], node_name=node_data["title"], model_name=node_data["model"].get("name") ) ) - context = self._process_list_variable_litearl(node_data["context"]["variable_selector"]) + context = self._process_list_variable_literal(node_data["context"]["variable_selector"]) memory = MemoryWindowSetting( enable=bool(node_data.get("memory")), enable_window=bool(node_data.get("memory", {}).get("window", {}).get("enabled", False)), @@ -367,7 +367,7 @@ class DifyConverter(BaseConverter): ) ) vision = node_data["vision"]["enabled"] - vision_input = self._process_list_variable_litearl( + vision_input = self._process_list_variable_literal( node_data["vision"]["configs"]["variable_selector"] ) if vision else None result = LLMNodeConfig.model_construct( @@ -433,7 +433,7 @@ class DifyConverter(BaseConverter): conditions.append( LoopConditionDetail.model_construct( operator=self.convert_compare_operator(condition["comparison_operator"]), - left=self._process_list_variable_litearl(condition["variable_selector"]), + left=self._process_list_variable_literal(condition["variable_selector"]), right=self.trans_variable_format( right_value ) if isinstance(right_value, str) and self.is_variable(right_value) else self.convert_variable_type( @@ -453,7 +453,7 @@ class DifyConverter(BaseConverter): right_input_type = variable["value_type"] right_value_type = self.variable_type_map(variable["var_type"]) if right_input_type == ValueInputType.VARIABLE: - right_value = self._process_list_variable_litearl(variable.get("value", "")) + right_value = self._process_list_variable_literal(variable.get("value", "")) else: right_value = self.convert_variable_type(right_value_type, variable.get("value", "")) loop_variables.append( @@ -475,10 +475,10 @@ class DifyConverter(BaseConverter): def convert_iteration_node_config(self, node: dict) -> dict: node_data = node["data"] result = IterationNodeConfig.model_construct( - input=self._process_list_variable_litearl(node_data["iterator_selector"]), + input=self._process_list_variable_literal(node_data["iterator_selector"]), parallel=node_data["is_parallel"], parallel_count=node_data["parallel_nums"], - output=self._process_list_variable_litearl(node_data["output_selector"]), + output=self._process_list_variable_literal(node_data["output_selector"]), output_type=self.variable_type_map(node_data.get("output_type")), flatten=node_data["flatten_output"], ).model_dump() @@ -494,8 +494,8 @@ class DifyConverter(BaseConverter): continue assignments.append( AssignmentItem( - variable_selector=self._process_list_variable_litearl(assignment["variable_selector"]), - value=self._process_list_variable_litearl( + variable_selector=self._process_list_variable_literal(assignment["variable_selector"]), + value=self._process_list_variable_literal( assignment["value"] ) if assignment["input_type"] == ValueInputType.VARIABLE else assignment["value"], operation=self.convert_assignment_operator(assignment["operation"]) @@ -514,7 +514,7 @@ class DifyConverter(BaseConverter): input_variables.append( InputVariable.model_construct( name=input_variable["variable"], - variable=self._process_list_variable_litearl(input_variable["value_selector"]), + variable=self._process_list_variable_literal(input_variable["value_selector"]), ) ) @@ -570,7 +570,7 @@ class DifyConverter(BaseConverter): else: if node_data["body"]["data"]: body_content = (node_data["body"]["data"][0].get("value") or - self._process_list_variable_litearl(node_data["body"]["data"][0].get("file"))) + self._process_list_variable_literal(node_data["body"]["data"][0].get("file"))) else: body_content = "" @@ -585,7 +585,7 @@ class DifyConverter(BaseConverter): self.trans_variable_format(key_value[0]) ] = self.trans_variable_format(key_value[1]) else: - self.warnings.append(ExceptionDefineition( + self.warnings.append(ExceptionDefinition( type=ExceptionType.CONFIG, node_id=node["id"], node_name=node_data["title"], @@ -603,7 +603,7 @@ class DifyConverter(BaseConverter): self.trans_variable_format(key_value[0]) ] = self.trans_variable_format(key_value[1]) else: - self.warnings.append(ExceptionDefineition( + self.warnings.append(ExceptionDefinition( type=ExceptionType.CONFIG, node_id=node["id"], node_name=node_data["title"], @@ -625,7 +625,7 @@ class DifyConverter(BaseConverter): default_header = var["value"] elif var["key"] == "status_code": default_status_code = var["value"] - default_value = HttpErrorDefaultTamplete( + default_value = HttpErrorDefaultTemplate( body=default_body, headers=default_header, status_code=default_status_code, @@ -668,7 +668,7 @@ class DifyConverter(BaseConverter): for variable in node_data["variables"]: mapping.append(VariablesMappingConfig.model_construct( name=variable["variable"], - value=self._process_list_variable_litearl(variable["value_selector"]) + value=self._process_list_variable_literal(variable["value_selector"]) )) result = JinjaRenderNodeConfig.model_construct( template=node_data["template"], @@ -679,14 +679,14 @@ class DifyConverter(BaseConverter): def convert_knowledge_node_config(self, node: dict) -> dict: node_data = node["data"] - self.warnings.append(ExceptionDefineition( + self.warnings.append(ExceptionDefinition( node_id=node["id"], node_name=node_data["title"], type=ExceptionType.CONFIG, detail=f"Please reconfigure the Knowledge Retrieval node.", )) result = KnowledgeRetrievalNodeConfig.model_construct( - query=self._process_list_variable_litearl(node_data["query_variable_selector"]), + query=self._process_list_variable_literal(node_data["query_variable_selector"]), ).model_dump() self.config_validate(node["id"], node["data"]["title"], KnowledgeRetrievalNodeConfig, result) @@ -695,7 +695,7 @@ class DifyConverter(BaseConverter): def convert_parameter_extractor_node_config(self, node: dict) -> dict: node_data = node["data"] self.warnings.append( - UnknowModelWarning( + UnknownModelWarning( node_id=node["id"], node_name=node_data["title"], model_name=node_data["model"].get("name") @@ -712,7 +712,7 @@ class DifyConverter(BaseConverter): ) ) result = ParameterExtractorNodeConfig.model_construct( - text=self._process_list_variable_litearl(node_data["query"]), + text=self._process_list_variable_literal(node_data["query"]), params=params, prompt=node_data.get("instruction") ).model_dump() @@ -727,14 +727,14 @@ class DifyConverter(BaseConverter): group_type = {} if not advanced_settings or not advanced_settings["group_enabled"]: group_variables = [ - self._process_list_variable_litearl(variable) + self._process_list_variable_literal(variable) for variable in node_data["variables"] ] group_type["output"] = node_data["output_type"] else: for group in advanced_settings["groups"]: group_variables[group["group_name"]] = [ - self._process_list_variable_litearl(variable) + self._process_list_variable_literal(variable) for variable in group["variables"] ] group_type[group["group_name"]] = group["output_type"] @@ -751,7 +751,7 @@ class DifyConverter(BaseConverter): def convert_tool_node_config(self, node: dict) -> dict: node_data = node["data"] - self.warnings.append(ExceptionDefineition( + self.warnings.append(ExceptionDefinition( node_id=node["id"], node_name=node_data["title"], type=ExceptionType.CONFIG, diff --git a/api/app/core/workflow/adapters/dify/dify_adapter.py b/api/app/core/workflow/adapters/dify/dify_adapter.py index 10397ad0..abd95408 100644 --- a/api/app/core/workflow/adapters/dify/dify_adapter.py +++ b/api/app/core/workflow/adapters/dify/dify_adapter.py @@ -12,7 +12,7 @@ from app.core.workflow.adapters.base_adapter import ( WorkflowParserResult ) from app.core.workflow.adapters.dify.converter import DifyConverter -from app.core.workflow.adapters.errors import ExceptionDefineition, ExceptionType +from app.core.workflow.adapters.errors import ExceptionDefinition, ExceptionType from app.core.workflow.nodes.enums import NodeType from app.schemas.workflow_schema import ( NodeDefinition, @@ -85,7 +85,7 @@ class DifyAdapter(BasePlatformAdapter, DifyConverter): if not all(field in self.config for field in require_fields): return False if self.config.get("app", {}).get("mode") == "workflow": - self.errors.append(ExceptionDefineition( + self.errors.append(ExceptionDefinition( type=ExceptionType.PLATFORM, detail="workflow mode is not supported" )) @@ -111,12 +111,12 @@ class DifyAdapter(BasePlatformAdapter, DifyConverter): edge = self._convert_edge(edge) if edge: self.edges.append(edge) - # + for variable in self.config.get("workflow").get("conversation_variables"): con_var = self._convert_variable(variable) if variable: self.conv_variables.append(con_var) - # + # for variables in config.get("workflow").get("environment_variables"): # variable = self._convert_variable(variables) # conv_variables.append(variable) @@ -152,7 +152,7 @@ class DifyAdapter(BasePlatformAdapter, DifyConverter): "y": node["position"]["y"] + position["y"] } self.errors.append( - ExceptionDefineition( + ExceptionDefinition( type=ExceptionType.NODE, node_id=node_id, detail="parent cycle node not found" @@ -189,7 +189,7 @@ class DifyAdapter(BasePlatformAdapter, DifyConverter): node_data = node["data"] converter = self.get_node_convert(node_type) if node_type == NodeType.UNKNOWN: - self.errors.append(ExceptionDefineition( + self.errors.append(ExceptionDefinition( type=ExceptionType.NODE, node_id=node["id"], node_name=node["data"]["title"], @@ -197,7 +197,7 @@ class DifyAdapter(BasePlatformAdapter, DifyConverter): )) return converter(node) except Exception as e: - self.errors.append(ExceptionDefineition( + self.errors.append(ExceptionDefinition( type=ExceptionType.NODE, node_id=node["id"], node_name=node["data"]["title"], @@ -207,7 +207,6 @@ class DifyAdapter(BasePlatformAdapter, DifyConverter): def _convert_edge(self, edge: dict[str, Any]) -> EdgeDefinition | None: try: - source = edge["source"] target = edge["target"] label = None @@ -230,7 +229,7 @@ class DifyAdapter(BasePlatformAdapter, DifyConverter): label=label, ) except Exception as e: - self.errors.append(ExceptionDefineition( + self.errors.append(ExceptionDefinition( type=ExceptionType.EDGE, detail=f"convert edge error - {e}", )) @@ -246,7 +245,7 @@ class DifyAdapter(BasePlatformAdapter, DifyConverter): description=variable.get("description") ) except Exception as e: - self.errors.append(ExceptionDefineition( + self.errors.append(ExceptionDefinition( type=ExceptionType.VARIABLE, name=variable.get("name"), detail=f"convert variable error - {e}", diff --git a/api/app/core/workflow/adapters/errors.py b/api/app/core/workflow/adapters/errors.py index c0340a5e..cb743c68 100644 --- a/api/app/core/workflow/adapters/errors.py +++ b/api/app/core/workflow/adapters/errors.py @@ -18,7 +18,7 @@ class ExceptionType(StrEnum): UNKNOWN = "unknown" -class ExceptionDefineition(BaseModel): +class ExceptionDefinition(BaseModel): type: ExceptionType detail: str @@ -29,7 +29,7 @@ class ExceptionDefineition(BaseModel): name: str | None = None -class UnknowModelWarning(ExceptionDefineition): +class UnknownModelWarning(ExceptionDefinition): type: ExceptionType = ExceptionType.NODE def __init__(self, node_id, node_name, model_name): @@ -40,36 +40,36 @@ class UnknowModelWarning(ExceptionDefineition): ) -class UnknowError(ExceptionDefineition): +class UnknownError(ExceptionDefinition): type: ExceptionType = ExceptionType.UNKNOWN def __init__(self, detail: str, **kwargs): super().__init__(detail=detail, **kwargs) -class UnsupportPlatform(ExceptionDefineition): +class UnsupportedPlatform(ExceptionDefinition): type: ExceptionType = ExceptionType.PLATFORM def __init__(self, platform: str): - super().__init__(detail=f"Unsupport platform {platform}") + super().__init__(detail=f"Unsupported platform {platform}") -class UnsupportVariableType(ExceptionDefineition): +class UnsupportedVariableType(ExceptionDefinition): type: ExceptionType = ExceptionType.VARIABLE def __init__(self, scope, name, var_type: str, **kwargs): - super().__init__(scope=scope, name=name, detail=f"Unsupport variable type:[{var_type}]", **kwargs) + super().__init__(scope=scope, name=name, detail=f"Unsupported variable type: [{var_type}]", **kwargs) -class InvalidConfiguration(ExceptionDefineition): +class InvalidConfiguration(ExceptionDefinition): type: ExceptionType = ExceptionType.CONFIG def __init__(self): super().__init__(detail="Invalid workflow configuration format") -class UnsupportNodeType(ExceptionDefineition): +class UnsupportedNodeType(ExceptionDefinition): type: ExceptionType = ExceptionType.NODE def __init__(self, node_id: str, node_type: str): - super().__init__(node_id=node_id, detail=f"Unsupport node Type {node_type}") + super().__init__(node_id=node_id, detail=f"Unsupported node type {node_type}") diff --git a/api/app/core/workflow/adapters/memory_bear/memory_bear_adapter.py b/api/app/core/workflow/adapters/memory_bear/memory_bear_adapter.py index 3516cb58..a2608a01 100644 --- a/api/app/core/workflow/adapters/memory_bear/memory_bear_adapter.py +++ b/api/app/core/workflow/adapters/memory_bear/memory_bear_adapter.py @@ -11,7 +11,7 @@ from app.core.workflow.adapters.base_adapter import ( BasePlatformAdapter, WorkflowParserResult ) -from app.core.workflow.adapters.errors import ExceptionDefineition, ExceptionType, UnsupportNodeType +from app.core.workflow.adapters.errors import ExceptionDefinition, ExceptionType, UnsupportedNodeType from app.core.workflow.adapters.memory_bear.memory_bear_converter import MemoryBearConverter from app.core.workflow.nodes.enums import NodeType from app.schemas.workflow_schema import ExecutionConfig, NodeDefinition, EdgeDefinition, VariableDefinition @@ -73,7 +73,7 @@ class MemoryBearAdapter(BasePlatformAdapter, MemoryBearConverter): try: node_type = self.map_node_type(node["type"]) if node_type == NodeType.UNKNOWN: - self.errors.append(UnsupportNodeType( + self.errors.append(UnsupportedNodeType( node_id=node_id, node_type=node["type"] )) @@ -85,7 +85,7 @@ class MemoryBearAdapter(BasePlatformAdapter, MemoryBearConverter): return NodeDefinition(**node) except Exception as e: - self.errors.append(ExceptionDefineition( + self.errors.append(ExceptionDefinition( type=ExceptionType.NODE, node_id=node_id, node_name=node_name, @@ -97,14 +97,14 @@ class MemoryBearAdapter(BasePlatformAdapter, MemoryBearConverter): def _convert_edge(self, edge: dict[str, Any], valid_node_ids: set) -> EdgeDefinition | None: try: if edge.get("source") not in valid_node_ids or edge.get("target") not in valid_node_ids: - self.warnings.append(ExceptionDefineition( + self.warnings.append(ExceptionDefinition( type=ExceptionType.EDGE, detail=f"edge {edge.get('id')} skipped: source or target node not found" )) return None return EdgeDefinition(**edge) except Exception as e: - self.errors.append(ExceptionDefineition( + self.errors.append(ExceptionDefinition( type=ExceptionType.EDGE, detail=f"convert edge error - {e}" )) @@ -115,7 +115,7 @@ class MemoryBearAdapter(BasePlatformAdapter, MemoryBearConverter): try: return VariableDefinition(**variable) except Exception as e: - self.warnings.append(ExceptionDefineition( + self.warnings.append(ExceptionDefinition( type=ExceptionType.VARIABLE, name=variable.get("name"), detail=f"convert variable error - {e}" diff --git a/api/app/core/workflow/adapters/memory_bear/memory_bear_converter.py b/api/app/core/workflow/adapters/memory_bear/memory_bear_converter.py index 031c7025..e96e0bf2 100644 --- a/api/app/core/workflow/adapters/memory_bear/memory_bear_converter.py +++ b/api/app/core/workflow/adapters/memory_bear/memory_bear_converter.py @@ -1,6 +1,6 @@ # -*- coding: UTF-8 -*- from app.core.workflow.adapters.base_converter import BaseConverter -from app.core.workflow.adapters.errors import ExceptionDefineition, ExceptionType +from app.core.workflow.adapters.errors import ExceptionDefinition, ExceptionType from app.core.workflow.nodes.base_config import BaseNodeConfig from app.core.workflow.nodes.configs import ( StartNodeConfig, @@ -65,7 +65,7 @@ class MemoryBearConverter(BaseConverter): try: return config_cls.model_validate(value) except Exception as e: - self.errors.append(ExceptionDefineition( + self.errors.append(ExceptionDefinition( type=ExceptionType.CONFIG, node_id=node_id, node_name=node_name, diff --git a/api/app/core/workflow/engine/graph_builder.py b/api/app/core/workflow/engine/graph_builder.py index 90668ad9..daef6e82 100644 --- a/api/app/core/workflow/engine/graph_builder.py +++ b/api/app/core/workflow/engine/graph_builder.py @@ -7,7 +7,7 @@ import re import uuid from collections import defaultdict from functools import lru_cache -from typing import Any, Iterable +from typing import Any, Iterable, Callable from langgraph.checkpoint.memory import InMemorySaver from langgraph.graph import START, END @@ -20,48 +20,52 @@ from app.core.workflow.engine.variable_pool import VariablePool from app.core.workflow.nodes import NodeFactory from app.core.workflow.nodes.enums import NodeType, BRANCH_NODES from app.core.workflow.utils.expression_evaluator import evaluate_condition +from app.core.workflow.validator import WorkflowValidator logger = logging.getLogger(__name__) +# Regex to split output into: +# - variable placeholders: {{ ... }} +# - normal literal text +# +# Example: +# "Hello {{user.name}}!" -> +# ["Hello ", "{{user.name}}", "!"] +_OUTPUT_PATTERN = re.compile(r'\{\{.*?}}|[^{}]+') +# Strict variable format: {{ node_id.field_name }} +_VARIABLE_PATTERN = re.compile(r'\{\{\s*[a-zA-Z0-9_]+\.[a-zA-Z0-9_]+\s*}}') + class GraphBuilder: def __init__( self, workflow_config: dict[str, Any], stream: bool = False, - subgraph: bool = False, + cycle: str = '', variable_pool: VariablePool | None = None ): self.workflow_config = workflow_config self.stream = stream - self.subgraph = subgraph + self.cycle = cycle - self.start_node_id = None - self.end_node_ids = [] - self.node_map = {node["id"]: node for node in self.nodes} + self.start_node_id: str | None = None + + self.node_map: dict[str, dict] = {} self.end_node_map: dict[str, StreamOutputConfig] = {} - self._find_upstream_branch_node = lru_cache( - maxsize=len(self.nodes) * 2 - )(self._find_upstream_branch_node) + self._find_upstream_activation_dep: Callable = self._find_upstream_activation_dep if variable_pool: self.variable_pool = variable_pool else: self.variable_pool = VariablePool() - self.graph = StateGraph(WorkflowState) - self.add_nodes() - self.add_edges() - self._analyze_end_node_output() - # EDGES MUST BE ADDED AFTER NODES ARE ADDED. - - @property - def nodes(self) -> list[dict[str, Any]]: - return self.workflow_config.get("nodes", []) - - @property - def edges(self) -> list[dict[str, Any]]: - return self.workflow_config.get("edges", []) + self.graph: StateGraph | None = None + self.nodes: list = [] + self.edges: list = [] + self.reachable_nodes: set[str] | None = None + self.end_nodes: list[dict] = [] + self._reverse_adj: dict[str, list[dict]] = defaultdict(list) + self._adj: dict[str, list[str]] = defaultdict(list) def get_node_type(self, node_id: str) -> str: """Retrieve the type of node given its ID. @@ -87,60 +91,51 @@ class GraphBuilder: result[node[0]].append(node[1]) return result - def _find_upstream_branch_node(self, target_node: str) -> tuple[bool, tuple[tuple[str, str]]]: - """ - Recursively find all upstream branch (control) nodes that influence the execution - of the given target node. + def _build_adj(self): + for edge in self.edges: + if edge["source"] not in self.reachable_nodes: + continue + self._reverse_adj[edge.get("target")].append({ + "id": edge["source"], "branch": edge.get("label") + }) + self._adj[edge.get("source")].append(edge["target"]) - This method walks upstream along the workflow graph starting from `target_node`. - It distinguishes between: - - branch nodes (node types listed in `BRANCH_NODES`) - - non-branch nodes (ordinary processing nodes) + def _find_upstream_activation_dep( + self, + target_node: str + ) -> tuple[tuple[tuple[str, str]], tuple[str]]: + """Find upstream dependencies that affect the activation of a target node. - Traversal rules: - 1. For each immediate upstream node: - - If it is a branch node, it is recorded as an affecting control node. - - If it is a non-branch node, the traversal continues recursively upstream. - 2. If ANY upstream path reaches a START / CYCLE_START node without encountering - a branch node, the traversal is considered invalid: - - `has_branch` will be False - - no branch nodes are returned. - 3. Only when ALL upstream non-branch paths eventually lead to at least one - branch node will `has_branch` be True. + Walks upstream along the workflow graph from the target node, collecting + two types of dependencies: + - Branch control nodes: upstream branch nodes (e.g. if-else) whose + routing outcome determines whether the target node executes. + - Output nodes: upstream END nodes that must complete their output + before the target node can activate. - Special case: - - If `target_node` has no upstream nodes AND its type is START or CYCLE_START, - it is considered directly reachable from the workflow entry, and therefore - has no controlling branch nodes. + The traversal terminates early and returns empty tuples if any upstream + path reaches START/CYCLE_START without encountering a branch or output + node, indicating the target node is directly reachable and should be + activated immediately. Args: - target_node (str): - The identifier of the node whose upstream control branches - are to be resolved. + target_node: The ID of the node whose upstream activation + dependencies are to be resolved. Returns: - tuple[bool, tuple[tuple[str, str]]]: - - has_branch (bool): - True if every upstream path from `target_node` encounters - at least one branch node. - False if any path reaches a start node without a branch. - - branch_nodes (tuple[tuple[str, str]]): - A deduplicated tuple of `(branch_node_id, branch_label)` pairs - representing all branch nodes that can influence `target_node`. - Returns an empty tuple if `has_branch` is False. + A tuple of two elements: + - A deduplicated tuple of (branch_node_id, branch_label) pairs + representing upstream branch control dependencies. Empty if + any clean path to START exists. + - A deduplicated tuple of upstream output node IDs that must + complete before this node activates. """ - source_nodes = [ - { - "id": edge.get("source"), - "branch": edge.get("label") - } - for edge in self.edges - if edge.get("target") == target_node - ] + source_nodes = self._reverse_adj[target_node] if not source_nodes and self.get_node_type(target_node) in [NodeType.START, NodeType.CYCLE_START]: - return False, tuple() + return tuple(), tuple() branch_nodes = [] + output_nodes = [] non_branch_nodes = [] for node_info in source_nodes: @@ -149,19 +144,23 @@ class GraphBuilder: (node_info["id"], node_info["branch"]) ) else: + if self.get_node_type(node_info["id"]) == NodeType.END: + output_nodes.append(node_info["id"]) non_branch_nodes.append(node_info["id"]) has_branch = True for node_id in non_branch_nodes: - node_has_branch, nodes = self._find_upstream_branch_node(node_id) - has_branch = has_branch and node_has_branch - if not has_branch: - break - branch_nodes.extend(nodes) - if not has_branch: - branch_nodes = [] + upstream_control_nodes, upstream_output_nodes = self._find_upstream_activation_dep(node_id) + if not upstream_control_nodes: + if not upstream_output_nodes and node_id not in output_nodes: + return tuple(), tuple() + branch_nodes = [] + has_branch = False + if has_branch: + branch_nodes.extend(upstream_control_nodes) + output_nodes.extend(upstream_output_nodes) - return has_branch, tuple(set(branch_nodes)) + return tuple(set(branch_nodes)), tuple(set(output_nodes)) def _analyze_end_node_output(self): """ @@ -182,11 +181,10 @@ class GraphBuilder: """ # Collect all End nodes in the workflow - end_nodes = [node for node in self.nodes if node.get("type") == "end"] - logger.info(f"[Prefix Analysis] Found {len(end_nodes)} End nodes") + logger.info(f"[Prefix Analysis] Found {len(self.end_nodes)} End nodes") # Iterate through each End node to analyze its output - for end_node in end_nodes: + for end_node in self.end_nodes: end_node_id = end_node.get("id") config = end_node.get("config", {}) output = config.get("output") @@ -195,42 +193,33 @@ class GraphBuilder: if not output: continue - # Regex to split output into: - # - variable placeholders: {{ ... }} - # - normal literal text - # - # Example: - # "Hello {{user.name}}!" -> - # ["Hello ", "{{user.name}}", "!"] - pattern = r'\{\{.*?\}\}|[^{}]+' - - # Strict variable format: {{ node_id.field_name }} - variable_pattern_string = r'\{\{\s*[a-zA-Z0-9_]+\.[a-zA-Z0-9_]+\s*\}\}' - variable_pattern = re.compile(variable_pattern_string) - # Split output into ordered segments - output_template = list(re.findall(pattern, output)) + output_template = list(_OUTPUT_PATTERN.findall(output)) # Determine whether each segment is literal text # True -> literal (can be directly output) # False -> variable placeholder (needs runtime value) output_flag = [ - not bool(variable_pattern.match(item)) + not bool(_VARIABLE_PATTERN.match(item)) for item in output_template ] # Stream mode: output activation depends on upstream branch nodes if self.stream: # Find upstream branch nodes that can control this End node - has_branch, control_nodes = self._find_upstream_branch_node(end_node_id) - + upstream_control_nodes, upstream_output_nodes = self._find_upstream_activation_dep(end_node_id) + activate = not bool(upstream_control_nodes) and not bool(upstream_output_nodes) # Build StreamOutputConfig for this End node self.end_node_map[end_node_id] = StreamOutputConfig( + id=end_node_id, # If there is no upstream branch, output is active immediately - activate=not has_branch, + activate=activate, # Branch nodes that control activation of this End node - control_nodes=self._merge_control_nodes(control_nodes), + control_nodes=self._merge_control_nodes(upstream_control_nodes), + upstream_output_nodes=list(upstream_output_nodes), + control_resolved=not bool(upstream_control_nodes), + output_resolved=not bool(upstream_output_nodes), # Convert output segments into OutputContent objects outputs=list( @@ -249,14 +238,16 @@ class GraphBuilder: cursor=0 ) logger.info(f"[Stream Analysis] end_id: {end_node_id}, " - f"activate: {not has_branch}, " - f"control_nodes: {control_nodes}," + f"activate: {activate}, " + f"control_nodes: {upstream_control_nodes}," + f"ref_outputs: {upstream_output_nodes}," f"output: {output_template}," f"output_activate: {output_flag}") # Non-stream mode: all outputs are activated by default else: self.end_node_map[end_node_id] = StreamOutputConfig( + id=end_node_id, activate=True, control_nodes={}, outputs=list( @@ -269,7 +260,10 @@ class GraphBuilder: for output_string, activate in zip(output_template, output_flag) ] ), - cursor=0 + cursor=0, + upstream_output_nodes=[], + control_resolved=True, + output_resolved=True, ) def add_nodes(self): @@ -292,24 +286,13 @@ class GraphBuilder: """ for node in self.nodes: node_type = node.get("type") - if node_type == NodeType.NOTES: - continue node_id = node.get("id") - cycle_node = node.get("cycle") - if cycle_node: - # Nodes within a loop subgraph are constructed by CycleGraphNode - if not self.subgraph: - continue - - # Record start and end node IDs - if node_type in [NodeType.START, NodeType.CYCLE_START]: - self.start_node_id = node_id - elif node_type == NodeType.END: - self.end_node_ids.append(node_id) + if node_id not in self.reachable_nodes: + continue # Create node instance (start and end nodes are also created) # NOTE:Loop node creation automatically removes the nodes and edges of the subgraph from the current graph - node_instance = NodeFactory.create_node(node, self.workflow_config) + node_instance = NodeFactory.create_node(node, self.workflow_config, self._adj[node_id]) if node_type in BRANCH_NODES: @@ -382,6 +365,8 @@ class GraphBuilder: for edge in self.edges: source = edge.get("source") target = edge.get("target") + if source not in self.reachable_nodes or target not in self.reachable_nodes: + continue condition = edge.get("condition") edge_type = edge.get("type") @@ -403,11 +388,12 @@ class GraphBuilder: # Add conditional edges for source_node, branches in conditional_edges.items(): def make_router(src, branch_list): - """reate a router function for each source node that routes to a NOP node for later merging.""" + """Create a router function for each source node that routes to a NOP node for later merging.""" def make_branch_node(node_name, targets): def node(s): - # NOTE: NOP NODE MUST NOT MODIFY STATE + # NOTE: NOP NODE USED FOR ROUTING ONLY. + # MUST NOT MUTATE STATE DIRECTLY; ONLY EMIT ACTIVATE SIGNALS. return { "activate": { node_id: s["activate"][node_name] @@ -448,7 +434,7 @@ class GraphBuilder: branch_activate = [] new_state = state.copy() new_state["activate"] = dict(state.get("activate", {})) # deep copy of activate - node_output = variable_pool.get_node_output(src, defalut=dict(), strict=False) + node_output = variable_pool.get_node_output(src, default=dict(), strict=False) for label, branch in unique_branch.items(): if node_output and evaluate_condition( branch["condition"], @@ -494,12 +480,52 @@ class GraphBuilder: logger.debug(f"Added waiting edge: {sources} -> {target}") # Connect End nodes to the global END node - for end_node_id in self.end_node_ids: - self.graph.add_edge(end_node_id, END) - logger.debug(f"Added edge: {end_node_id} -> END") + for node in self.reachable_nodes: + if not self._adj[node]: + self.graph.add_edge(node, END) return def build(self) -> CompiledStateGraph: + nodes = self.workflow_config.get("nodes", []) + edges = self.workflow_config.get("edges", []) + + for node in nodes: + if (node.get("cycle") or '') == self.cycle: + node_type = node.get("type") + if node_type in [NodeType.START, NodeType.CYCLE_START]: + self.start_node_id = node.get("id") + elif node_type == NodeType.NOTES: + continue + self.nodes.append(node) + self.node_map[node.get("id")] = node + + for edge in edges: + source_in = edge.get("source") in self.node_map + target_in = edge.get("target") in self.node_map + if source_in ^ target_in: + raise ValueError( + f"Cycle node is connected to external node, " + f"source: {edge.get('source')}, target: {edge.get('target')}" + ) + + if source_in and target_in: + self.edges.append(edge) + + self.reachable_nodes = WorkflowValidator.get_reachable_nodes(self.start_node_id, self.edges) + self.end_nodes = [ + node + for node in self.nodes + if node.get("type") == "end" and node.get("id") in self.reachable_nodes + ] + self._build_adj() + self._find_upstream_activation_dep: Callable = lru_cache( + maxsize=len(self.nodes)*2 + )(self._find_upstream_activation_dep) + + self.graph = StateGraph(WorkflowState) + self.add_nodes() + self.add_edges() + + self._analyze_end_node_output() checkpointer = InMemorySaver() - self.graph = self.graph.compile(checkpointer=checkpointer) - return self.graph + return self.graph.compile(checkpointer=checkpointer) diff --git a/api/app/core/workflow/engine/result_builder.py b/api/app/core/workflow/engine/result_builder.py index 31bccf57..be0c957a 100644 --- a/api/app/core/workflow/engine/result_builder.py +++ b/api/app/core/workflow/engine/result_builder.py @@ -2,6 +2,7 @@ # Author: Eternity # @Email: 1533512157@qq.com # @Time : 2026/2/10 13:33 +from app.core.workflow.engine.runtime_schema import ExecutionContext from app.core.workflow.engine.variable_pool import VariablePool @@ -9,9 +10,11 @@ class WorkflowResultBuilder: def build_final_output( self, result: dict, + execution_context: ExecutionContext, variable_pool: VariablePool, elapsed_time: float, final_output: str, + success: bool ): """Construct the final standardized output of the workflow execution. @@ -25,10 +28,13 @@ class WorkflowResultBuilder: - "node_outputs" (dict): Outputs of executed nodes. - "messages" (list): Conversation messages exchanged during execution. - "error" (str, optional): Error message if any node failed. + execution_context (ExecutionContext): The execution context containing metadata like + execution ID, workspace ID, and user ID.) variable_pool (VariablePool): Variable Pool elapsed_time (float): Total execution time in seconds. final_output (Any): The aggregated or final output content of the workflow (e.g., combined messages from all End nodes). + success (bool): Whether the execution was successful. Returns: dict: A dictionary containing the final workflow execution result with keys: @@ -46,18 +52,23 @@ class WorkflowResultBuilder: """ node_outputs = result.get("node_outputs", {}) token_usage = self.aggregate_token_usage(node_outputs) - conversation_id = variable_pool.get_value("sys.conversation_id") + conversation_vars = {} + sys_vars = {} + + if variable_pool: + conversation_vars = variable_pool.get_all_conversation_vars() + sys_vars = variable_pool.get_all_system_vars() return { - "status": "completed", + "status": "completed" if success else "failed", "output": final_output, "variables": { - "conv": variable_pool.get_all_conversation_vars(), - "sys": variable_pool.get_all_system_vars() + "conv": conversation_vars, + "sys": sys_vars }, "node_outputs": node_outputs, "messages": result.get("messages", []), - "conversation_id": conversation_id, + "conversation_id": execution_context.conversation_id, "elapsed_time": elapsed_time, "token_usage": token_usage, "error": result.get("error"), diff --git a/api/app/core/workflow/engine/runtime_schema.py b/api/app/core/workflow/engine/runtime_schema.py index e4bf65af..036ce0e8 100644 --- a/api/app/core/workflow/engine/runtime_schema.py +++ b/api/app/core/workflow/engine/runtime_schema.py @@ -12,14 +12,29 @@ class ExecutionContext(BaseModel): execution_id: str workspace_id: str user_id: str + conversation_id: str + memory_storage_type: str + user_rag_memory_id: str checkpoint_config: RunnableConfig @classmethod - def create(cls, execution_id: str, workspace_id: str, user_id: str): + def create( + cls, + execution_id: str, + workspace_id: str, + user_id: str, + conversation_id: str, + memory_storage_type: str, + user_rag_memory_id: str + ): return cls( execution_id=execution_id, workspace_id=workspace_id, user_id=user_id, + conversation_id=conversation_id, + memory_storage_type=memory_storage_type, + user_rag_memory_id=user_rag_memory_id, + checkpoint_config=RunnableConfig( configurable={ "thread_id": uuid.uuid4(), diff --git a/api/app/core/workflow/engine/state_manager.py b/api/app/core/workflow/engine/state_manager.py index 0a4a1463..2da0d3a8 100644 --- a/api/app/core/workflow/engine/state_manager.py +++ b/api/app/core/workflow/engine/state_manager.py @@ -33,6 +33,8 @@ class WorkflowState(dict): "workspace_id", "user_id", "activate", + "memory_storage_type", + "user_rag_memory_id" }) __optional_keys__ = frozenset({ "error", @@ -62,6 +64,9 @@ class WorkflowState(dict): # node activate status activate: Annotated[dict[str, bool], merge_activate_state] + memory_storage_type: str + user_rag_memory_id: str + class WorkflowStateManager: def create_initial_state( @@ -85,7 +90,9 @@ class WorkflowStateManager: looping=0, activate={ start_node_id: True - } + }, + memory_storage_type=execution_context.memory_storage_type, + user_rag_memory_id=execution_context.user_rag_memory_id ) @staticmethod diff --git a/api/app/core/workflow/engine/stream_output_coordinator.py b/api/app/core/workflow/engine/stream_output_coordinator.py index ddee9adc..dcc92fdb 100644 --- a/api/app/core/workflow/engine/stream_output_coordinator.py +++ b/api/app/core/workflow/engine/stream_output_coordinator.py @@ -3,6 +3,7 @@ # @Email: 1533512157@qq.com # @Time : 2026/2/9 15:11 import re +from collections import deque from typing import AsyncGenerator from pydantic import BaseModel, Field, PrivateAttr @@ -37,8 +38,8 @@ class OutputContent(BaseModel): activate: bool = Field( ..., description=( - "Whether this output segment is currently active.\n" - "- True: allowed to be emitted/output\n" + "Whether this output segment is currently active." + "- True: allowed to be emitted/output" "- False: blocked until activated by branch control" ) ) @@ -46,8 +47,8 @@ class OutputContent(BaseModel): is_variable: bool = Field( ..., description=( - "Whether this segment represents a variable placeholder.\n" - "True -> variable (e.g. {{ node.field }})\n" + "Whether this segment represents a variable placeholder." + "True -> variable (e.g. {{ node.field }})" "False -> literal text" ) ) @@ -86,12 +87,16 @@ class StreamOutputConfig(BaseModel): - which upstream branch/control nodes gate the activation - how each parsed output segment is streamed and activated """ + id: str = Field( + ..., + description="ID of the End node this configuration belongs to." + ) activate: bool = Field( ..., description=( - "Global activation flag for the End node output.\n" - "When False, output segments should not be emitted even if available.\n" + "Global activation flag for the End node output." + "When False, output segments should not be emitted even if available." "This flag typically becomes True once required control branch conditions " "are satisfied." ) @@ -100,17 +105,46 @@ class StreamOutputConfig(BaseModel): control_nodes: dict[str, list[str]] = Field( ..., description=( - "Control branch conditions for this End node output.\n" - "Mapping of `branch_node_id -> expected_branch_label`.\n" + "Control branch conditions for this End node output." + "Mapping of `branch_node_id -> expected_branch_label`." "The End node output becomes globally active when a controlling branch node " "reports a matching completion status." ) ) + upstream_output_nodes: list[str] = Field( + ..., + description=( + "Upstream output node dependencies (data flow)." + "Represents END/output nodes that this output depends on." + "These nodes provide data sources required before this output can be activated " + "or streamed." + "Used to ensure correct ordering and dependency resolution in streaming mode." + ) + ) + + control_resolved: bool = Field( + ..., + description=( + "Whether all upstream branch control dependencies have been satisfied." + "True if no upstream branch nodes exist or the required branch " + "conditions have been met." + ) + ) + + output_resolved: bool = Field( + ..., + description=( + "Whether all upstream output node dependencies have been completed." + "True if no upstream output nodes exist or all upstream output " + "nodes have finished their output." + ) + ) + outputs: list[OutputContent] = Field( ..., description=( - "Ordered list of output segments parsed from the output template.\n" + "Ordered list of output segments parsed from the output template." "Each segment represents either a literal text block or a variable placeholder " "that may be activated independently." ) @@ -119,49 +153,97 @@ class StreamOutputConfig(BaseModel): cursor: int = Field( ..., description=( - "Streaming cursor index.\n" - "Indicates the next output segment index to be emitted.\n" + "Streaming cursor index." + "Indicates the next output segment index to be emitted." "Segments with index < cursor are considered already streamed." ) ) + force: bool = Field( + default=False, + description=( + "Force flag for output emission." + "When True, all output segments are emitted regardless of activation state." + "Triggered when this output node has finished execution." + ) + ) + def update_activate(self, scope: str, status=None): """ - Update streaming activation state based on an upstream node or special variable. + Update streaming activation state based on upstream events. Args: scope (str): Identifier of the completed upstream entity. - If a control branch node, it should match a key in `control_nodes`. - - If a variable placeholder (e.g., "sys.xxx"), it may appear in output segments. + - If an upstream output node, it should match an entry in `upstream_output_nodes`. + - If a variable placeholder (e.g., "sys.xxx" or "node_id.field"), + it may appear in output segments. + status (optional): Completion status of the control branch node. Required when `scope` refers to a control node. Behavior: - 1. Control branch nodes: - - If `scope` matches a key in `control_nodes` and `status` matches the expected - branch label, the End node output becomes globally active (`activate = True`). + 1. Force activation: + - If `self.force` is True, the method returns immediately. + - If `scope == self.id`, the node marks itself as completed: + - `activate = True` + - `force = True` + This is typically used for final flushing when the node finishes execution. - 2. Variable output segments: - - For each segment that is a variable (`is_variable=True`): - - If the segment literal references `scope`, mark the segment as active. - - This applies both to regular node variables (e.g., "node_id.field") - and special system variables (e.g., "sys.xxx"). + 2. Control dependency resolution: + - If `scope` matches a key in `control_nodes`: + - `status` must be provided. + - If `status` matches expected branch labels, mark control as resolved + (`control_resolved = True`). + + 3. Upstream output dependency resolution: + - If `scope` is in `upstream_output_nodes`, + mark data dependency as resolved (`output_resolved = True`). + + 4. Global activation condition: + - The node becomes active when BOTH conditions are satisfied: + - control_resolved == True + - output_resolved == True + - Once activated, `activate` remains True. + + 5. Variable segment activation: + - For each output segment that is a variable (`is_variable=True`): + - If the segment depends on the given `scope`, + mark the segment as active. + - This applies to both node variables (e.g., "node_id.field") + and system variables (e.g., "sys.xxx"). Notes: - - This method does not emit output or advance the streaming cursor. - - It only updates activation flags based on upstream events or special variables. + - This method does NOT emit output or advance the streaming cursor. + - It only updates activation and dependency resolution states. + - Activation is driven by both control flow (branch nodes) and + data flow (upstream output nodes). """ + if self.force: + return - # Case 1: resolve control branch dependency + if scope == self.id: + self.activate = True + self.force = True + return + + # resolve control branch dependency if scope in self.control_nodes: if status is None: raise RuntimeError("[Stream Output] Control node activation status not provided") if status in self.control_nodes[scope]: - self.activate = True + self.control_resolved = True - # Case 2: activate variable segments related to this node + if scope in self.upstream_output_nodes: + self.upstream_output_nodes.remove(scope) + if not self.upstream_output_nodes: + self.output_resolved = True + + self.activate = self.activate or (self.control_resolved and self.output_resolved) + + # activate variable segments related to this node for i in range(len(self.outputs)): if ( self.outputs[i].is_variable @@ -174,12 +256,17 @@ class StreamOutputCoordinator: def __init__(self): self.end_outputs: dict[str, StreamOutputConfig] = {} self.activate_end: str | None = None + self.output_queue: deque[str] = deque() + self.processed_outputs = [] def initialize_end_outputs( self, end_node_map: dict[str, StreamOutputConfig] ): self.end_outputs = end_node_map + self.processed_outputs = [] + self.activate_end = None + self.output_queue = deque() @property def current_activate_end_info(self): @@ -209,10 +296,13 @@ class StreamOutputCoordinator: scope (str): The node ID or scope that has completed execution. status (str | None): Optional status of the node (used for branch/control nodes). """ - for node in self.end_outputs.keys(): + for node in self.end_outputs: self.end_outputs[node].update_activate(scope, status) - if self.end_outputs[node].activate and self.activate_end is None: - self.activate_end = node + if self.end_outputs[node].activate and node not in self.processed_outputs: + self.output_queue.append(node) + self.processed_outputs.append(node) + if self.activate_end is None and self.output_queue: + self.activate_end = self.output_queue.popleft() async def emit_activate_chunk( self, @@ -256,7 +346,7 @@ class StreamOutputCoordinator: final_chunk = '' current_segment = end_info.outputs[end_info.cursor] - if not current_segment.activate and not force: + if not current_segment.activate and not force and not end_info.force: # Stop processing until this segment becomes active break @@ -273,7 +363,7 @@ class StreamOutputCoordinator: logger.warning(f"[STREAM] Failed to evaluate segment: {current_segment.literal}, error: {e}") if final_chunk: - logger.info(f"[STREAM] StreamOutput Node:{self.activate_end}, chunk:{final_chunk}") + logger.info(f"[STREAM] StreamOutput Node:{self.activate_end}, chunk_length:{len(final_chunk)}") yield { "event": "message", "data": { @@ -285,8 +375,7 @@ class StreamOutputCoordinator: end_info.cursor += 1 if end_info.cursor >= len(end_info.outputs): - self.end_outputs.pop(self.activate_end) - self.activate_end = None + self.pop_current_activate_end() async def flush_remaining_chunk( self, @@ -325,6 +414,8 @@ class StreamOutputCoordinator: async for msg_event in self.emit_activate_chunk(variable_pool, force=True): yield msg_event + if self.output_queue: + self.activate_end = self.output_queue.popleft() # Move to next active End node if current one is done if not self.activate_end and self.end_outputs: self.activate_end = list(self.end_outputs.keys())[0] diff --git a/api/app/core/workflow/engine/variable_pool.py b/api/app/core/workflow/engine/variable_pool.py index bc88df19..60f1257e 100644 --- a/api/app/core/workflow/engine/variable_pool.py +++ b/api/app/core/workflow/engine/variable_pool.py @@ -13,7 +13,7 @@ from pydantic import BaseModel from app.core.workflow.engine.runtime_schema import ExecutionContext from app.core.workflow.variable.base_variable import VariableType, DEFAULT_VALUE -from app.core.workflow.variable.variable_objects import T, create_variable_instance +from app.core.workflow.variable.variable_objects import T, create_variable_instance, ArrayVariable, FileVariable logger = logging.getLogger(__name__) @@ -351,12 +351,12 @@ class VariablePool: } return runtime_vars - def get_node_output(self, node_id: str, defalut: Any = None, strict: bool = True) -> dict[str, Any] | None: + def get_node_output(self, node_id: str, default: Any = None, strict: bool = True) -> dict[str, Any] | None: """获取指定节点的输出(运行时变量) Args: node_id: 节点 ID - defalut: 默认值 + default: 默认值 strict: 是否严格模式 Returns: @@ -368,11 +368,21 @@ class VariablePool: if strict: raise KeyError(f"node {node_id} output not exist") else: - return defalut + return default def copy(self, pool: 'VariablePool'): self.variables = deepcopy(pool.variables) + def is_file_variable(self, selector): + variable_struct = self.get_instance(selector, default=None, strict=False) + if variable_struct is None: + return False + if isinstance(variable_struct, FileVariable): + return True + elif isinstance(variable_struct, ArrayVariable) and variable_struct.child_type == FileVariable: + return True + return False + def to_dict(self) -> dict[str, Any]: """导出为字典 diff --git a/api/app/core/workflow/executor.py b/api/app/core/workflow/executor.py index ff979f2b..0a820826 100644 --- a/api/app/core/workflow/executor.py +++ b/api/app/core/workflow/executor.py @@ -3,6 +3,7 @@ # @Email: 1533512157@qq.com # @Time : 2026/2/9 13:51 import datetime +import time import logging from typing import Any @@ -82,13 +83,15 @@ class WorkflowExecutor: CompiledStateGraph: The compiled and ready-to-run state graph. """ logger.info(f"Starting workflow graph build: execution_id={self.execution_context.execution_id}") + start_time = time.time() builder = GraphBuilder( self.workflow_config, stream=stream, ) + + self.graph = builder.build() self.start_node_id = builder.start_node_id self.variable_pool = builder.variable_pool - self.graph = builder.build() self.stream_coordinator.initialize_end_outputs(builder.end_node_map) self.event_handler = EventStreamHandler( @@ -96,7 +99,8 @@ class WorkflowExecutor: variable_pool=self.variable_pool, execution_id=self.execution_context.execution_id ) - logger.info(f"Workflow graph build completed: execution_id={self.execution_context.execution_id}") + logger.info(f"Workflow graph build completed: execution_id={self.execution_context.execution_id}, " + f"cost: {time.time() - start_time:.4f}s") return self.graph @@ -128,89 +132,18 @@ class WorkflowExecutor: - token_usage: aggregated token usage if available - error: error message if any """ - logger.info(f"Starting workflow execution: execution_id={self.execution_context.execution_id}") - - start_time = datetime.datetime.now() - - # Execute the workflow - try: - # Build the workflow graph - graph = self.build_graph() - - # Initialize the variable pool with input data - await self.variable_initializer.initialize( - variable_pool=self.variable_pool, - input_data=input_data, - execution_context=self.execution_context - ) - initial_state = self.state_manager.create_initial_state( - workflow_config=self.workflow_config, - input_data=input_data, - execution_context=self.execution_context, - start_node_id=self.start_node_id - ) - - result = await graph.ainvoke(initial_state, config=self.execution_context.checkpoint_config) - - # Aggregate output from all End nodes - full_content = '' - for end_id in self.stream_coordinator.end_outputs.keys(): - full_content += self.variable_pool.get_value(f"{end_id}.output", default="", strict=False) - - # Append messages for user and assistant - if input_data.get("files"): - result["messages"].extend( - [ - { - "role": "user", - "content": input_data.get("message", '') - }, - { - "role": "user", - "content": input_data.get("files") - }, - { - "role": "assistant", - "content": full_content - } - ] - ) - else: - result["messages"].extend( - [ - { - "role": "user", - "content": input_data.get("message", '') - }, - { - "role": "assistant", - "content": full_content - } - ] - ) - # Calculate elapsed time - end_time = datetime.datetime.now() - elapsed_time = (end_time - start_time).total_seconds() - - logger.info( - f"Workflow execution completed: execution_id={self.execution_context.execution_id}, elapsed_time={elapsed_time:.2f}ms") - - return self.result_builder.build_final_output(result, self.variable_pool, elapsed_time, full_content) - - except Exception as e: - end_time = datetime.datetime.now() - elapsed_time = (end_time - start_time).total_seconds() - - logger.error(f"Workflow execution failed: execution_id={self.execution_context.execution_id}, error={e}", - exc_info=True) - return { - "status": "failed", - "error": str(e), - "output": None, - "node_outputs": {}, - "elapsed_time": elapsed_time, - "token_usage": None - } + start = datetime.datetime.now() + async for event in self.execute_stream(input_data): + if event.get("event") == "workflow_end": + return event.get("data") + return self.result_builder.build_final_output( + {"error": "Workflow execution did not end as expected"}, + self.execution_context, + self.variable_pool, + (datetime.datetime.now() - start).total_seconds(), + "", + success=False + ) async def execute_stream( self, @@ -244,11 +177,12 @@ class WorkflowExecutor: "data": { "execution_id": self.execution_context.execution_id, "workspace_id": self.execution_context.workspace_id, - "conversation_id": input_data.get("conversation_id"), + "conversation_id": self.execution_context.conversation_id, "timestamp": int(start_time.timestamp() * 1000) } } - + result = None + full_content = '' try: # Build the workflow graph in streaming mode graph = self.build_graph(stream=True) @@ -266,7 +200,6 @@ class WorkflowExecutor: start_node_id=self.start_node_id ) - full_content = '' self.stream_coordinator.update_scope_activation("sys") # Execute the workflow with streaming @@ -363,7 +296,13 @@ class WorkflowExecutor: yield { "event": "workflow_end", - "data": self.result_builder.build_final_output(result, self.variable_pool, elapsed_time, full_content) + "data": self.result_builder.build_final_output( + result, + self.execution_context, + self.variable_pool, + elapsed_time, + full_content, + success=True) } except Exception as e: @@ -372,16 +311,20 @@ class WorkflowExecutor: logger.error(f"Workflow execution failed: execution_id={self.execution_context.execution_id}, error={e}", exc_info=True) - + if result is None: + result = {"error": str(e)} + else: + result["error"] = str(e) yield { "event": "workflow_end", - "data": { - "execution_id": self.execution_context.execution_id, - "status": "failed", - "error": str(e), - "elapsed_time": elapsed_time, - "timestamp": end_time.isoformat() - } + "data": self.result_builder.build_final_output( + result, + self.execution_context, + self.variable_pool, + elapsed_time, + full_content, + success=False + ) } @@ -390,7 +333,9 @@ async def execute_workflow( input_data: dict[str, Any], execution_id: str, workspace_id: str, - user_id: str + user_id: str, + memory_storage_type: str, + user_rag_memory_id: str ) -> dict[str, Any]: """ Execute a workflow (convenience function, non-streaming). @@ -401,6 +346,8 @@ async def execute_workflow( execution_id (str): Execution ID. workspace_id (str): Workspace ID. user_id (str): User ID. + user_rag_memory_id: rag knowledge db id + memory_storage_type: neo4j / rag Returns: dict: Workflow execution result. @@ -408,7 +355,10 @@ async def execute_workflow( execution_context = ExecutionContext.create( execution_id=execution_id, workspace_id=workspace_id, - user_id=user_id + user_id=user_id, + conversation_id=input_data.get("conversation_id"), + memory_storage_type=memory_storage_type, + user_rag_memory_id=user_rag_memory_id ) executor = WorkflowExecutor( workflow_config=workflow_config, @@ -422,7 +372,9 @@ async def execute_workflow_stream( input_data: dict[str, Any], execution_id: str, workspace_id: str, - user_id: str + user_id: str, + memory_storage_type: str, + user_rag_memory_id: str ): """ Execute a workflow in streaming mode (convenience function). @@ -433,6 +385,8 @@ async def execute_workflow_stream( execution_id (str): Execution ID. workspace_id (str): Workspace ID. user_id (str): User ID. + user_rag_memory_id: rag knowledge db id + memory_storage_type: neo4j / rag Yields: dict: Streaming workflow events, e.g. node start, node end, chunk messages, workflow end. @@ -440,7 +394,10 @@ async def execute_workflow_stream( execution_context = ExecutionContext.create( execution_id=execution_id, workspace_id=workspace_id, - user_id=user_id + user_id=user_id, + memory_storage_type=memory_storage_type, + conversation_id=input_data.get("conversation_id"), + user_rag_memory_id=user_rag_memory_id ) executor = WorkflowExecutor( workflow_config=workflow_config, diff --git a/api/app/core/workflow/nodes/agent/node.py b/api/app/core/workflow/nodes/agent/node.py index 8959e27c..7b146a9c 100644 --- a/api/app/core/workflow/nodes/agent/node.py +++ b/api/app/core/workflow/nodes/agent/node.py @@ -64,9 +64,7 @@ class AgentNode(BaseNode): if not release: raise ValueError(f"Agent 不存在: {agent_id}") - - return release, message async def execute(self, state: WorkflowState, variable_pool: VariablePool) -> dict[str, Any]: diff --git a/api/app/core/workflow/nodes/assigner/node.py b/api/app/core/workflow/nodes/assigner/node.py index 4c897d5a..f5bdf000 100644 --- a/api/app/core/workflow/nodes/assigner/node.py +++ b/api/app/core/workflow/nodes/assigner/node.py @@ -14,8 +14,8 @@ logger = logging.getLogger(__name__) class AssignerNode(BaseNode): - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.variable_updater = True self.typed_config: AssignerNodeConfig | None = None diff --git a/api/app/core/workflow/nodes/base_node.py b/api/app/core/workflow/nodes/base_node.py index 0e3fecee..8567ebbe 100644 --- a/api/app/core/workflow/nodes/base_node.py +++ b/api/app/core/workflow/nodes/base_node.py @@ -28,7 +28,7 @@ class BaseNode(ABC): All node types should inherit from this class and implement the `execute` method. """ - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): """Initialize the node. Args: @@ -41,6 +41,7 @@ class BaseNode(ABC): self.node_type = node_config["type"] self.cycle = node_config.get("cycle") self.node_name = node_config.get("name", self.node_id) + self.down_stream_nodes = down_stream_nodes # 使用 or 运算符处理 None 值 self.config = node_config.get("config") or {} self.error_handling = node_config.get("error_handling") or {} @@ -93,18 +94,16 @@ class BaseNode(ABC): dict: A dict with a single key 'activate', mapping node IDs to their activation status (True/False). """ - edges = self.workflow_config.get("edges") - under_stream_nodes = [ - edge.get("target") - for edge in edges - if edge.get("source") == self.node_id and self.node_type not in BRANCH_NODES - ] - return { - "activate": { - node_id: self.check_activate(state) - for node_id in under_stream_nodes - } | {self.node_id: self.check_activate(state)} - } + activate_flag = self.check_activate(state) + + if self.node_type not in BRANCH_NODES: + activate = {node_id: activate_flag for node_id in self.down_stream_nodes} + else: + activate = {} + + activate[self.node_id] = activate_flag + + return {"activate": activate} @abstractmethod async def execute(self, state: WorkflowState, variable_pool: VariablePool) -> Any: @@ -315,8 +314,8 @@ class BaseNode(ABC): elapsed_time = (time.time() - start_time) * 1000 - logger.info(f"Node {self.node_id} streaming execution finished, " - f"time elapsed: {elapsed_time:.2f}ms, chunks: {chunk_count}") + logger.debug(f"Node {self.node_id} streaming execution finished, " + f"time elapsed: {elapsed_time:.2f}ms, chunks: {chunk_count}") # Extract processed output (call subclass's _extract_output) extracted_output = self._extract_output(final_result) @@ -428,8 +427,8 @@ class BaseNode(ABC): when an error edge exists. If no error edge exists, this method raises an exception to stop the workflow. """ - # Check if the node has an error edge defined - error_edge = self._find_error_edge() + # # Check if the node has an error edge defined + # error_edge = self._find_error_edge() # Extract input data (for logging or audit purposes) input_data = self._extract_input(state, variable_pool) @@ -447,27 +446,26 @@ class BaseNode(ABC): "error": error_message } - if error_edge: - # If an error edge exists, log a warning and continue to error node - logger.warning( - f"Node {self.node_id} execution failed, redirecting to error node: {error_edge['target']}" - ) - return { - "node_outputs": { - self.node_id: node_output - }, - "error": error_message, - "error_node": self.node_id - } - else: - # If no error edge, send the error via stream writer and stop the workflow - writer = get_stream_writer() - writer({ - "type": "node_error", - **node_output - }) - logger.error(f"Node {self.node_id} execution failed, stopping workflow: {error_message}") - raise Exception(f"Node {self.node_id} execution failed: {error_message}") + # if error_edge: + # # If an error edge exists, log a warning and continue to error node + # logger.warning( + # f"Node {self.node_id} execution failed, redirecting to error node: {error_edge['target']}" + # ) + # return { + # "node_outputs": { + # self.node_id: node_output + # }, + # "error": error_message, + # "error_node": self.node_id + # } + # else: + writer = get_stream_writer() + writer({ + "type": "node_error", + **node_output + }) + logger.error(f"Node {self.node_id} execution failed, stopping workflow: {error_message}") + raise Exception(f"Node {self.node_id} execution failed: {error_message}") def _extract_input(self, state: WorkflowState, variable_pool: VariablePool) -> dict[str, Any]: """Extracts the input data for this node (used for logging or audit). @@ -623,7 +621,6 @@ class BaseNode(ABC): async def process_message( api_config: ModelInfo, content: str | dict | FileObject, - end_user_id: str, enable_file=False ) -> list | str | None: provider = api_config.provider @@ -642,10 +639,10 @@ class BaseNode(ABC): return content elif isinstance(content, FileObject): - if content.content_cache.get(provider): - return content.content_cache[provider] + if content.content_cache.get(f"{provider}_{api_config.is_omni}"): + return content.content_cache[f"{provider}_{api_config.is_omni}"] with get_db_read() as db: - multimodel_service = MultimodalService(db, api_config=api_config) + multimodal_service = MultimodalService(db, api_config=api_config) file_obj = FileInput( type=content.type, url=content.url, @@ -654,16 +651,15 @@ class BaseNode(ABC): upload_file_id=uuid.UUID(content.file_id) if content.file_id else None, ) file_obj.set_content(content.get_content()) - message = await multimodel_service.process_files( - end_user_id, + message = await multimodal_service.process_files( [file_obj], ) content.set_content(file_obj.get_content()) if message: - content.content_cache[provider] = message + content.content_cache[f"{provider}_{api_config.is_omni}"] = message return message return None - raise TypeError(f'Unexpect input value type - {type(content)}') + raise TypeError(f'Unexpected input value type - {type(content)}') @staticmethod def process_model_output(content) -> str: diff --git a/api/app/core/workflow/nodes/code/node.py b/api/app/core/workflow/nodes/code/node.py index 9303302d..d89b208b 100644 --- a/api/app/core/workflow/nodes/code/node.py +++ b/api/app/core/workflow/nodes/code/node.py @@ -51,8 +51,8 @@ console.log(result) class CodeNode(BaseNode): - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: CodeNodeConfig | None = None def _output_types(self) -> dict[str, VariableType]: @@ -128,7 +128,7 @@ class CodeNode(BaseNode): else: raise ValueError(f"Unsupported language: {self.typed_config.language}") - async with httpx.AsyncClient() as client: + async with httpx.AsyncClient(timeout=60) as client: response = await client.post( "http://sandbox:8194/v1/sandbox/run", headers={ diff --git a/api/app/core/workflow/nodes/cycle_graph/config.py b/api/app/core/workflow/nodes/cycle_graph/config.py index 52aca1d9..75358c47 100644 --- a/api/app/core/workflow/nodes/cycle_graph/config.py +++ b/api/app/core/workflow/nodes/cycle_graph/config.py @@ -51,7 +51,7 @@ class ConditionDetail(BaseModel): ) right: Any = Field( - ..., + default=None, description="Right-hand operand of the comparison expression" ) diff --git a/api/app/core/workflow/nodes/cycle_graph/loop.py b/api/app/core/workflow/nodes/cycle_graph/loop.py index d3ada1ec..84901bad 100644 --- a/api/app/core/workflow/nodes/cycle_graph/loop.py +++ b/api/app/core/workflow/nodes/cycle_graph/loop.py @@ -158,7 +158,7 @@ class LoopRuntime: self.variable_pool.variables["conv"].update( self.child_variable_pool.variables["conv"] ) - loop_vars = self.child_variable_pool.get_node_output(self.node_id, defalut={}, strict=False) + loop_vars = self.child_variable_pool.get_node_output(self.node_id, default={}, strict=False) loopstate["node_outputs"][self.node_id] = loop_vars def evaluate_conditional(self) -> bool: @@ -261,4 +261,4 @@ class LoopRuntime: idx += 1 logger.info(f"loop node {self.node_id}: execution completed") - return self.child_variable_pool.get_node_output(self.node_id) | {"__child_state": child_state} + return self.child_variable_pool.get_node_output(self.node_id, default={}, strict=False) | {"__child_state": child_state} diff --git a/api/app/core/workflow/nodes/cycle_graph/node.py b/api/app/core/workflow/nodes/cycle_graph/node.py index 71e0dbdb..fc80939f 100644 --- a/api/app/core/workflow/nodes/cycle_graph/node.py +++ b/api/app/core/workflow/nodes/cycle_graph/node.py @@ -30,17 +30,13 @@ class CycleGraphNode(BaseNode): It acts as a container and execution controller for a subgraph. """ - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) - - self.cycle_nodes = list() # Nodes belonging to this cycle - self.cycle_edges = list() # Edges connecting nodes within the cycle + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) + self.cycle_nodes, self.cycle_edges = self.pure_cycle_graph() self.start_node_id = None # ID of the start node within the cycle self.graph: StateGraph | CompiledStateGraph | None = None self.child_variable_pool: VariablePool | None = None - self.build_graph() - self.iteration_flag = True def _output_types(self) -> dict[str, VariableType]: outputs = {"__child_state": VariableType.ARRAY_OBJECT} @@ -119,11 +115,11 @@ class CycleGraphNode(BaseNode): else: remain_edges.append(edge) - # Update workflow_config by removing cycle nodes and internal edges - self.workflow_config["nodes"] = [ - node for node in nodes if node.get("cycle") != self.node_id - ] - self.workflow_config["edges"] = remain_edges + # # Update workflow_config by removing cycle nodes and internal edges + # self.workflow_config["nodes"] = [ + # node for node in nodes if node.get("cycle") != self.node_id + # ] + # self.workflow_config["edges"] = remain_edges return cycle_nodes, cycle_edges @@ -137,18 +133,18 @@ class CycleGraphNode(BaseNode): 3. Compile the graph for runtime execution """ from app.core.workflow.engine.graph_builder import GraphBuilder - self.cycle_nodes, self.cycle_edges = self.pure_cycle_graph() + self.child_variable_pool = VariablePool() builder = GraphBuilder( { "nodes": self.cycle_nodes, "edges": self.cycle_edges, }, - subgraph=True, - variable_pool=self.child_variable_pool + variable_pool=self.child_variable_pool, + cycle=self.node_id ) - self.start_node_id = builder.start_node_id self.graph = builder.build() + self.start_node_id = builder.start_node_id self.child_variable_pool = builder.variable_pool async def execute(self, state: WorkflowState, variable_pool: VariablePool) -> Any: @@ -169,6 +165,7 @@ class CycleGraphNode(BaseNode): Raises: RuntimeError: If the node type is unsupported. """ + self.build_graph() if self.node_type == NodeType.LOOP: return await LoopRuntime( start_id=self.start_node_id, @@ -194,6 +191,7 @@ class CycleGraphNode(BaseNode): raise RuntimeError("Unknown cycle node type") async def execute_stream(self, state: WorkflowState, variable_pool: VariablePool): + self.build_graph() if self.node_type == NodeType.LOOP: yield { "__final__": True, diff --git a/api/app/core/workflow/nodes/document_extractor/__init__.py b/api/app/core/workflow/nodes/document_extractor/__init__.py new file mode 100644 index 00000000..c51bc2c0 --- /dev/null +++ b/api/app/core/workflow/nodes/document_extractor/__init__.py @@ -0,0 +1,4 @@ +from .config import DocExtractorNodeConfig +from .node import DocExtractorNode + +__all__ = ["DocExtractorNode", "DocExtractorNodeConfig"] diff --git a/api/app/core/workflow/nodes/document_extractor/config.py b/api/app/core/workflow/nodes/document_extractor/config.py new file mode 100644 index 00000000..69f7f76d --- /dev/null +++ b/api/app/core/workflow/nodes/document_extractor/config.py @@ -0,0 +1,18 @@ +from pydantic import Field +from app.core.workflow.nodes.base_config import BaseNodeConfig + + +class DocExtractorNodeConfig(BaseNodeConfig): + file_selector: str = Field( + ..., + description="File variable selector, e.g. {{ sys.files }} or {{ node_id.file }}" + ) + + class Config: + json_schema_extra = { + "examples": [ + { + "file_selector": "{{ sys.files }}" + } + ] + } diff --git a/api/app/core/workflow/nodes/document_extractor/node.py b/api/app/core/workflow/nodes/document_extractor/node.py new file mode 100644 index 00000000..40641f3c --- /dev/null +++ b/api/app/core/workflow/nodes/document_extractor/node.py @@ -0,0 +1,103 @@ +import logging +from typing import Any + +from app.core.workflow.engine.state_manager import WorkflowState +from app.core.workflow.engine.variable_pool import VariablePool +from app.core.workflow.nodes.base_node import BaseNode +from app.core.workflow.nodes.document_extractor.config import DocExtractorNodeConfig +from app.core.workflow.variable.base_variable import VariableType, FileObject +from app.db import get_db_read +from app.schemas.app_schema import FileInput, FileType, TransferMethod + +logger = logging.getLogger(__name__) + + +def _file_object_to_file_input(f: FileObject) -> FileInput: + """Convert workflow FileObject to multimodal FileInput.""" + return FileInput( + type=FileType.DOCUMENT, + transfer_method=TransferMethod(f.transfer_method), + url=f.url or None, + upload_file_id=f.file_id or None, + file_type=f.origin_file_type or "", + ) + + +def _normalise_files(val: Any) -> list[FileObject]: + if isinstance(val, FileObject): + return [val] + if isinstance(val, dict) and val.get("is_file"): + return [FileObject(**val)] + if isinstance(val, list): + result: list[FileObject] = [] + for item in val: + if isinstance(item, FileObject): + result.append(item) + elif isinstance(item, dict) and item.get("is_file"): + result.append(FileObject(**item)) + else: + logger.warning("Ignoring non-file entry in file list for document extractor: %r", item) + return result + return [] + + +class DocExtractorNode(BaseNode): + """Document Extractor Node. + + Reads one or more file variables and extracts their text content + by delegating to MultimodalService._extract_document_text. + + Outputs: + text (string) – full concatenated text of all input files + chunks (array[string]) – per-file extracted text + """ + + def _output_types(self) -> dict[str, VariableType]: + return { + "text": VariableType.STRING, + "chunks": VariableType.ARRAY_STRING, + } + + def _extract_output(self, business_result: Any) -> Any: + return business_result + + def _extract_input(self, state: WorkflowState, variable_pool: VariablePool) -> dict[str, Any]: + return {"file_selector": self.config.get("file_selector")} + + async def execute(self, state: WorkflowState, variable_pool: VariablePool) -> Any: + config = DocExtractorNodeConfig(**self.config) + + raw_val = self.get_variable(config.file_selector, variable_pool, strict=False) + if raw_val is None: + logger.warning(f"Node {self.node_id}: file variable '{config.file_selector}' is empty") + return {"text": "", "chunks": []} + + files = _normalise_files(raw_val) + if not files: + return {"text": "", "chunks": []} + + chunks: list[str] = [] + with get_db_read() as db: + from app.services.multimodal_service import MultimodalService + svc = MultimodalService(db) + for f in files: + try: + file_input = _file_object_to_file_input(f) + # Ensure URL is populated for local files + if not file_input.url: + file_input.url = await svc.get_file_url(file_input) + # Reuse cached bytes if already fetched + if f.get_content(): + file_input.set_content(f.get_content()) + text = await svc._extract_document_text(file_input) + chunks.append(text) + except Exception as e: + logger.error( + f"Node {self.node_id}: failed to extract file url={f.url} file_id={f.file_id}: {e}", + exc_info=True, + ) + chunks.append("") + + full_text = "\n\n".join(c for c in chunks if c) + logger.info(f"Node {self.node_id}: extracted {len(files)} file(s), total chars={len(full_text)}") + return {"text": full_text, "chunks": chunks} diff --git a/api/app/core/workflow/nodes/end/config.py b/api/app/core/workflow/nodes/end/config.py index 5c2a6c2a..02df5091 100644 --- a/api/app/core/workflow/nodes/end/config.py +++ b/api/app/core/workflow/nodes/end/config.py @@ -1,9 +1,7 @@ """End 节点配置""" - from pydantic import Field -from app.core.workflow.nodes.base_config import BaseNodeConfig, VariableDefinition -from app.core.workflow.variable.base_variable import VariableType +from app.core.workflow.nodes.base_config import BaseNodeConfig class EndNodeConfig(BaseNodeConfig): diff --git a/api/app/core/workflow/nodes/end/node.py b/api/app/core/workflow/nodes/end/node.py index 2799316a..770cf328 100644 --- a/api/app/core/workflow/nodes/end/node.py +++ b/api/app/core/workflow/nodes/end/node.py @@ -36,8 +36,6 @@ class EndNode(BaseNode): Returns: 最终输出字符串 """ - logger.info(f"节点 {self.node_id} (End) 开始执行") - # 获取配置的输出模板 output_template = self.config.get("output") @@ -46,11 +44,4 @@ class EndNode(BaseNode): output = self._render_template(output_template, variable_pool, strict=False) else: output = "" - - # 统计信息(用于日志) - node_outputs = state.get("node_outputs", {}) - total_nodes = len(node_outputs) - - logger.info(f"节点 {self.node_id} (End) 执行完成,共执行 {total_nodes} 个节点") - return output diff --git a/api/app/core/workflow/nodes/enums.py b/api/app/core/workflow/nodes/enums.py index 43ab593b..529cd0b3 100644 --- a/api/app/core/workflow/nodes/enums.py +++ b/api/app/core/workflow/nodes/enums.py @@ -23,12 +23,13 @@ class NodeType(StrEnum): BREAK = "break" MEMORY_READ = "memory-read" MEMORY_WRITE = "memory-write" + DOCUMENT_EXTRACTOR = "document-extractor" UNKNOWN = "unknown" NOTES = "notes" -BRANCH_NODES = [NodeType.IF_ELSE, NodeType.HTTP_REQUEST, NodeType.QUESTION_CLASSIFIER] +BRANCH_NODES = frozenset({NodeType.IF_ELSE, NodeType.HTTP_REQUEST, NodeType.QUESTION_CLASSIFIER}) class ComparisonOperator(StrEnum): diff --git a/api/app/core/workflow/nodes/http_request/config.py b/api/app/core/workflow/nodes/http_request/config.py index fe38fafb..e1b84f0c 100644 --- a/api/app/core/workflow/nodes/http_request/config.py +++ b/api/app/core/workflow/nodes/http_request/config.py @@ -115,7 +115,7 @@ class HttpRetryConfig(BaseModel): ) -class HttpErrorDefaultTamplete(BaseModel): +class HttpErrorDefaultTemplate(BaseModel): body: str = Field( default="", description="Default body returned on HTTP error", @@ -143,7 +143,7 @@ class HttpErrorHandleConfig(BaseModel): description="Error handling strategy: 'none', 'default', or 'branch'", ) - default: HttpErrorDefaultTamplete | None = Field( + default: HttpErrorDefaultTemplate | None = Field( default=None, description="Default response template for error handling", ) diff --git a/api/app/core/workflow/nodes/http_request/node.py b/api/app/core/workflow/nodes/http_request/node.py index 23378c83..086bee4a 100644 --- a/api/app/core/workflow/nodes/http_request/node.py +++ b/api/app/core/workflow/nodes/http_request/node.py @@ -16,7 +16,7 @@ from app.core.workflow.engine.variable_pool import VariablePool from app.core.workflow.nodes.base_node import BaseNode from app.core.workflow.nodes.enums import HttpRequestMethod, HttpErrorHandle, HttpAuthType, HttpContentType from app.core.workflow.nodes.http_request.config import HttpRequestNodeConfig, HttpRequestNodeOutput -from app.core.workflow.utils.file_processer import mime_to_file_type +from app.core.workflow.utils.file_processor import mime_to_file_type from app.core.workflow.variable.base_variable import VariableType, FileObject from app.core.workflow.variable.variable_objects import FileVariable, ArrayVariable from app.schemas import FileType, TransferMethod @@ -157,8 +157,8 @@ class HttpRequestNode(BaseNode): or a branch identifier string when error branching is enabled. """ - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: HttpRequestNodeConfig | None = None def _output_types(self) -> dict[str, VariableType]: diff --git a/api/app/core/workflow/nodes/if_else/config.py b/api/app/core/workflow/nodes/if_else/config.py index 894898f0..638e4b2d 100644 --- a/api/app/core/workflow/nodes/if_else/config.py +++ b/api/app/core/workflow/nodes/if_else/config.py @@ -18,7 +18,7 @@ class ConditionDetail(BaseModel): ) right: Any = Field( - ..., + default=None, description="Value to compare with" ) diff --git a/api/app/core/workflow/nodes/if_else/node.py b/api/app/core/workflow/nodes/if_else/node.py index 7e98efab..ec46b20b 100644 --- a/api/app/core/workflow/nodes/if_else/node.py +++ b/api/app/core/workflow/nodes/if_else/node.py @@ -14,8 +14,8 @@ logger = logging.getLogger(__name__) class IfElseNode(BaseNode): - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: IfElseNodeConfig | None = None def _output_types(self) -> dict[str, VariableType]: @@ -31,13 +31,13 @@ class IfElseNode(BaseNode): expressions.append({ "left": self.get_variable(expression.left, variable_pool, strict=False), "right": expression.right - if expression.input_type == ValueInputType.CONSTANT + if expression.input_type == ValueInputType.CONSTANT or expression.right is None else self.get_variable(expression.right, variable_pool, strict=False), - "operator": expression.operator, + "operator": str(expression.operator), }) result.append({ "expressions": expressions, - "logical_operator": case.logical_operator, + "logical_operator": str(case.logical_operator), }) return { "cases": result diff --git a/api/app/core/workflow/nodes/jinja_render/node.py b/api/app/core/workflow/nodes/jinja_render/node.py index e13709d4..abf21524 100644 --- a/api/app/core/workflow/nodes/jinja_render/node.py +++ b/api/app/core/workflow/nodes/jinja_render/node.py @@ -12,8 +12,8 @@ logger = logging.getLogger(__name__) class JinjaRenderNode(BaseNode): - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: JinjaRenderNodeConfig | None = None def _output_types(self) -> dict[str, VariableType]: diff --git a/api/app/core/workflow/nodes/knowledge/node.py b/api/app/core/workflow/nodes/knowledge/node.py index d3e9efd9..92699cb4 100644 --- a/api/app/core/workflow/nodes/knowledge/node.py +++ b/api/app/core/workflow/nodes/knowledge/node.py @@ -21,8 +21,8 @@ logger = logging.getLogger(__name__) class KnowledgeRetrievalNode(BaseNode): - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: KnowledgeRetrievalNodeConfig | None = None self.vector_service: ElasticSearchVector | None = None diff --git a/api/app/core/workflow/nodes/llm/node.py b/api/app/core/workflow/nodes/llm/node.py index b293d1f4..a691001f 100644 --- a/api/app/core/workflow/nodes/llm/node.py +++ b/api/app/core/workflow/nodes/llm/node.py @@ -70,8 +70,8 @@ class LLMNode(BaseNode): - ai/assistant: AI 消息(AIMessage) """ - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: LLMNodeConfig | None = None self.messages = [] @@ -144,7 +144,6 @@ class LLMNode(BaseNode): f"创建 LLM 实例: provider={model_info.provider}, model={model_info.model_name}, streaming={stream}") messages_config = self.typed_config.messages - if messages_config: # 使用 LangChain 消息格式 messages = [] @@ -153,7 +152,6 @@ class LLMNode(BaseNode): content_template = msg_config.content content_template = self._render_context(content_template, variable_pool) content = self._render_template(content_template, variable_pool) - user_id = self.get_variable("sys.user_id", variable_pool) # 根据角色创建对应的消息对象 if role == "system": messages.append({ @@ -161,32 +159,31 @@ class LLMNode(BaseNode): "content": await self.process_message( model_info, content, - user_id, self.typed_config.vision, ) }) elif role in ["user", "human"]: messages.append({ "role": "user", - "content": await self.process_message(model_info, content, user_id, self.typed_config.vision) + "content": await self.process_message(model_info, content, self.typed_config.vision) }) elif role in ["ai", "assistant"]: messages.append({ "role": "assistant", - "content": await self.process_message(model_info, content, user_id, self.typed_config.vision) + "content": await self.process_message(model_info, content, self.typed_config.vision) }) else: logger.warning(f"未知的消息角色: {role},默认使用 user") messages.append({ "role": "user", - "content": await self.process_message(model_info, content, user_id, self.typed_config.vision) + "content": await self.process_message(model_info, content, self.typed_config.vision) }) if self.typed_config.vision_input and self.typed_config.vision: file_content = [] files = variable_pool.get_instance(self.typed_config.vision_input) for file in files.value: - content = await self.process_message(model_info, file.value, user_id, self.typed_config.vision) + content = await self.process_message(model_info, file.value, self.typed_config.vision) if content: file_content.extend(content) if messages and messages[-1]["role"] == 'user': @@ -200,7 +197,7 @@ class LLMNode(BaseNode): if isinstance(message["content"], list): file_content = [] for file in message["content"]: - content = await self.process_message(model_info, file, user_id, self.typed_config.vision) + content = await self.process_message(model_info, file, self.typed_config.vision) if content: file_content.extend(content) history_message.append( @@ -210,7 +207,6 @@ class LLMNode(BaseNode): message["content"] = await self.process_message( model_info, message["content"], - user_id, self.typed_config.vision ) history_message.append(message) diff --git a/api/app/core/workflow/nodes/memory/node.py b/api/app/core/workflow/nodes/memory/node.py index 1d42e82e..73c52b79 100644 --- a/api/app/core/workflow/nodes/memory/node.py +++ b/api/app/core/workflow/nodes/memory/node.py @@ -1,3 +1,4 @@ +import re from typing import Any from app.core.workflow.engine.state_manager import WorkflowState @@ -5,14 +6,16 @@ from app.core.workflow.engine.variable_pool import VariablePool from app.core.workflow.nodes.base_node import BaseNode from app.core.workflow.nodes.memory.config import MemoryReadNodeConfig, MemoryWriteNodeConfig from app.core.workflow.variable.base_variable import VariableType +from app.core.workflow.variable.variable_objects import FileVariable, ArrayVariable from app.db import get_db_read +from app.schemas import FileInput from app.services.memory_agent_service import MemoryAgentService from app.tasks import write_message_task class MemoryReadNode(BaseNode): - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: MemoryReadNodeConfig | None = None def _output_types(self) -> dict[str, VariableType]: @@ -36,19 +39,32 @@ class MemoryReadNode(BaseNode): search_switch=self.typed_config.search_switch, history=[], db=db, - storage_type="neo4j", - user_rag_memory_id="" + storage_type=state["memory_storage_type"], + user_rag_memory_id=state["user_rag_memory_id"] ) class MemoryWriteNode(BaseNode): - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: MemoryWriteNodeConfig | None = None def _output_types(self) -> dict[str, VariableType]: return {"output": VariableType.STRING} + @staticmethod + def _extract_multimodal_memory_variables(content: str, variable_pool: VariablePool) -> tuple[list[str], str]: + variable_pattern_string = r'\{\{\s*[a-zA-Z0-9_]+\.[a-zA-Z0-9_]+\s*\}\}' + variable_pattern = re.compile(variable_pattern_string) + variables = variable_pattern.findall(content) + file_variables = [] + for variable in variables: + if variable_pool.is_file_variable(variable): + file_variables.append(variable) + for var in file_variables: + content = content.replace(var, "") + return file_variables, content + async def execute(self, state: WorkflowState, variable_pool: VariablePool) -> Any: self.typed_config = MemoryWriteNodeConfig(**self.config) end_user_id = self.get_variable("sys.user_id", variable_pool) @@ -63,17 +79,42 @@ class MemoryWriteNode(BaseNode): }) for message in self.typed_config.messages: + file_variables, content = self._extract_multimodal_memory_variables( + message.content, + variable_pool + ) + file_info = [] + for var in file_variables: + instence: FileVariable | ArrayVariable[FileVariable] = variable_pool.get_instance(var) + if isinstance(instence, FileVariable): + file_info.append(FileInput( + type=instence.value.type, + transfer_method=instence.value.transfer_method, + upload_file_id=instence.value.file_id, + url=instence.value.url, + file_type=instence.value.origin_file_type + ).model_dump()) + elif isinstance(instence, ArrayVariable) and instence.child_type == FileVariable: + for file_instence in instence.value: + file_info.append(FileInput( + type=file_instence.value.type, + transfer_method=file_instence.value.transfer_method, + upload_file_id=file_instence.value.file_id, + url=file_instence.value.url, + file_type=file_instence.value.origin_file_type + ).model_dump()) messages.append({ "role": message.role, - "content": self._render_template(message.content, variable_pool) + "content": self._render_template(content, variable_pool), + "files": file_info }) write_message_task.delay( - end_user_id, - messages, - str(self.typed_config.config_id), - "neo4j", - "" + end_user_id=end_user_id, + message=messages, + config_id=str(self.typed_config.config_id), + storage_type=state["memory_storage_type"], + user_rag_memory_id=state["user_rag_memory_id"] ) return "success" diff --git a/api/app/core/workflow/nodes/node_factory.py b/api/app/core/workflow/nodes/node_factory.py index 864e3251..49add867 100644 --- a/api/app/core/workflow/nodes/node_factory.py +++ b/api/app/core/workflow/nodes/node_factory.py @@ -26,6 +26,7 @@ from app.core.workflow.nodes.variable_aggregator import VariableAggregatorNode from app.core.workflow.nodes.question_classifier import QuestionClassifierNode from app.core.workflow.nodes.breaker import BreakNode from app.core.workflow.nodes.tool import ToolNode +from app.core.workflow.nodes.document_extractor import DocExtractorNode logger = logging.getLogger(__name__) @@ -49,7 +50,8 @@ WorkflowNode = Union[ ToolNode, MemoryReadNode, MemoryWriteNode, - CodeNode + CodeNode, + DocExtractorNode ] @@ -81,6 +83,7 @@ class NodeFactory: NodeType.MEMORY_READ: MemoryReadNode, NodeType.MEMORY_WRITE: MemoryWriteNode, NodeType.CODE: CodeNode, + NodeType.DOCUMENT_EXTRACTOR: DocExtractorNode } @classmethod @@ -104,13 +107,15 @@ class NodeFactory: def create_node( cls, node_config: dict[str, Any], - workflow_config: dict[str, Any] + workflow_config: dict[str, Any], + down_stream_nodes: list[str] ) -> WorkflowNode | None: """创建节点实例 Args: node_config: 节点配置 workflow_config: 工作流配置 + down_stream_nodes: 下游节点 Returns: 节点实例或 None(对于不支持的节点类型) @@ -127,7 +132,7 @@ class NodeFactory: # 创建节点实例 logger.debug(f"create node instance: {node_config.get('id')} (type={node_type})") - return node_class(node_config, workflow_config) + return node_class(node_config, workflow_config, down_stream_nodes) @classmethod def get_supported_types(cls) -> list[str]: diff --git a/api/app/core/workflow/nodes/operators.py b/api/app/core/workflow/nodes/operators.py index be33d35a..14fc9d9f 100644 --- a/api/app/core/workflow/nodes/operators.py +++ b/api/app/core/workflow/nodes/operators.py @@ -250,6 +250,8 @@ class ConditionBase(ABC): self.type_limit = getattr(self, "type_limit", None) def resolve_right_literal_value(self): + if self.right_selector is None: + return None if self.input_type == ValueInputType.VARIABLE: pattern = r"\{\{\s*(.*?)\s*\}\}" right_expression = re.sub(pattern, r"\1", self.right_selector).strip() diff --git a/api/app/core/workflow/nodes/parameter_extractor/node.py b/api/app/core/workflow/nodes/parameter_extractor/node.py index acac09e4..3dc5fcc3 100644 --- a/api/app/core/workflow/nodes/parameter_extractor/node.py +++ b/api/app/core/workflow/nodes/parameter_extractor/node.py @@ -21,8 +21,8 @@ logger = logging.getLogger(__name__) class ParameterExtractorNode(BaseNode): - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: ParameterExtractorNodeConfig | None = None self.response_metadata = {} diff --git a/api/app/core/workflow/nodes/question_classifier/node.py b/api/app/core/workflow/nodes/question_classifier/node.py index 5cebd886..31fadaf6 100644 --- a/api/app/core/workflow/nodes/question_classifier/node.py +++ b/api/app/core/workflow/nodes/question_classifier/node.py @@ -22,8 +22,8 @@ DEFAULT_EMPTY_QUESTION_CASE = f"{DEFAULT_CASE_PREFIX}1" class QuestionClassifierNode(BaseNode): """问题分类器节点""" - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: QuestionClassifierNodeConfig | None = None self.category_to_case_map = {} self.response_metadata = {} diff --git a/api/app/core/workflow/nodes/start/node.py b/api/app/core/workflow/nodes/start/node.py index a9618f7b..7a324cc4 100644 --- a/api/app/core/workflow/nodes/start/node.py +++ b/api/app/core/workflow/nodes/start/node.py @@ -27,14 +27,8 @@ class StartNode(BaseNode): 注意:变量的验证和默认值处理由 Executor 在初始化时完成。 """ - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - """初始化 Start 节点 - - Args: - node_config: 节点配置 - workflow_config: 工作流配置 - """ - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) # 解析并验证配置 self.typed_config: StartNodeConfig | None = None @@ -62,7 +56,6 @@ class StartNode(BaseNode): 包含系统参数、会话变量和自定义变量的字典 """ self.typed_config = StartNodeConfig(**self.config) - logger.info(f"节点 {self.node_id} (Start) 开始执行") # 处理自定义变量(传入 pool 避免重复创建) custom_vars = self._process_custom_variables(variable_pool) @@ -77,9 +70,9 @@ class StartNode(BaseNode): **custom_vars # 自定义变量作为节点输出的一部分 } - logger.info( - f"节点 {self.node_id} (Start) 执行完成," - f"输出了 {len(custom_vars)} 个自定义变量" + logger.debug( + f"Node {self.node_id} (Start) execution completed, " + f"outputting {len(custom_vars)} custom variables" ) return result diff --git a/api/app/core/workflow/nodes/tool/node.py b/api/app/core/workflow/nodes/tool/node.py index 0e9d3c62..72c5c6a8 100644 --- a/api/app/core/workflow/nodes/tool/node.py +++ b/api/app/core/workflow/nodes/tool/node.py @@ -20,8 +20,8 @@ TEMPLATE_PATTERN = re.compile(r"\{\{.*?}}") class ToolNode(BaseNode): """工具节点""" - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: ToolNodeConfig | None = None def _output_types(self) -> dict[str, VariableType]: diff --git a/api/app/core/workflow/nodes/variable_aggregator/node.py b/api/app/core/workflow/nodes/variable_aggregator/node.py index de82f8ff..9a9c5566 100644 --- a/api/app/core/workflow/nodes/variable_aggregator/node.py +++ b/api/app/core/workflow/nodes/variable_aggregator/node.py @@ -12,8 +12,8 @@ logger = logging.getLogger(__name__) class VariableAggregatorNode(BaseNode): - def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any]): - super().__init__(node_config, workflow_config) + def __init__(self, node_config: dict[str, Any], workflow_config: dict[str, Any], down_stream_nodes: list[str]): + super().__init__(node_config, workflow_config, down_stream_nodes) self.typed_config: VariableAggregatorNodeConfig | None = None def _output_types(self) -> dict[str, VariableType]: diff --git a/api/app/core/workflow/utils/file_processer.py b/api/app/core/workflow/utils/file_processor.py similarity index 100% rename from api/app/core/workflow/utils/file_processer.py rename to api/app/core/workflow/utils/file_processor.py diff --git a/api/app/core/workflow/utils/template_renderer.py b/api/app/core/workflow/utils/template_renderer.py index 424fdf20..6a73efc4 100644 --- a/api/app/core/workflow/utils/template_renderer.py +++ b/api/app/core/workflow/utils/template_renderer.py @@ -153,7 +153,8 @@ class TemplateRenderer: # 全局渲染器实例(严格模式) -_default_renderer = TemplateRenderer(strict=True) +_strict_renderer = TemplateRenderer(strict=True) +_lenient_renderer = TemplateRenderer(strict=False) def render_template( @@ -184,7 +185,7 @@ def render_template( ... ) '请分析: 这是一段文本' """ - renderer = TemplateRenderer(strict=strict) + renderer = _strict_renderer if strict else _lenient_renderer return renderer.render(template, conv_vars, node_outputs, system_vars) @@ -197,4 +198,4 @@ def validate_template(template: str) -> list[str]: Returns: 错误列表 """ - return _default_renderer.validate(template) + return _strict_renderer.validate(template) diff --git a/api/app/core/workflow/validator.py b/api/app/core/workflow/validator.py index 3b6e9036..0ad74865 100644 --- a/api/app/core/workflow/validator.py +++ b/api/app/core/workflow/validator.py @@ -6,6 +6,7 @@ import copy import logging +from collections import defaultdict, deque from typing import Any, Union, TYPE_CHECKING from app.core.workflow.nodes.enums import NodeType @@ -119,7 +120,6 @@ class WorkflowValidator: errors = [] graphs = cls.get_subgraph(workflow_config) - logger.info(graphs) for index, graph in enumerate(graphs): nodes = graph.get("nodes", []) edges = graph.get("edges", []) @@ -170,7 +170,7 @@ class WorkflowValidator: # 仅在发布时验证所有节点可达 # 6. 验证所有节点可达(从 start 节点出发) if start_nodes and not errors: # 只有在前面验证通过时才检查可达性 - reachable = WorkflowValidator._get_reachable_nodes( + reachable = WorkflowValidator.get_reachable_nodes( start_nodes[0]["id"], edges ) @@ -183,7 +183,7 @@ class WorkflowValidator: has_cycle, cycle_path = WorkflowValidator._has_cycle(nodes, edges) if has_cycle: errors.append( - f"工作流存在循环依赖(请使用 loop 节点实现循环): {' -> '.join(cycle_path)}" + f"工作流存在循环依赖(请使用 loop/iteration 节点实现循环): {' -> '.join(cycle_path)}" ) # 8. 验证变量名 @@ -194,7 +194,7 @@ class WorkflowValidator: return len(errors) == 0, errors @staticmethod - def _get_reachable_nodes(start_id: str, edges: list[dict]) -> set[str]: + def get_reachable_nodes(start_id: str, edges: list[dict]) -> set[str]: """获取从 start 节点可达的所有节点 Args: @@ -204,18 +204,18 @@ class WorkflowValidator: Returns: 可达节点 ID 集合 """ + adj = defaultdict(list) + for edge in edges: + adj[edge["source"]].append(edge["target"]) + reachable = {start_id} - queue = [start_id] - + queue = deque([start_id]) while queue: - current = queue.pop(0) - for edge in edges: - if edge.get("source") == current: - target = edge.get("target") - if target and target not in reachable: - reachable.add(target) - queue.append(target) - + current = queue.popleft() + for target in adj[current]: + if target not in reachable: + reachable.add(target) + queue.append(target) return reachable @staticmethod @@ -229,10 +229,6 @@ class WorkflowValidator: Returns: (has_cycle, cycle_path): 是否有循环和循环路径 """ - # 排除 loop 类型的节点 - loop_nodes = {n["id"] for n in nodes if n.get("type") == "loop"} - - # 构建邻接表(排除 loop 节点的边和错误边) graph: dict[str, list[str]] = {} for edge in edges: source = edge.get("source") @@ -243,10 +239,6 @@ class WorkflowValidator: if edge_type == "error": continue - # 如果涉及 loop 节点,跳过 - if source in loop_nodes or target in loop_nodes: - continue - if source and target: if source not in graph: graph[source] = [] diff --git a/api/app/core/workflow/variable/base_variable.py b/api/app/core/workflow/variable/base_variable.py index aea40cf6..f5d8ff8f 100644 --- a/api/app/core/workflow/variable/base_variable.py +++ b/api/app/core/workflow/variable/base_variable.py @@ -2,7 +2,7 @@ from enum import StrEnum from abc import abstractmethod, ABC from typing import Any -from pydantic import BaseModel, Field +from pydantic import BaseModel, Field, PrivateAttr from app.schemas import FileType @@ -41,10 +41,10 @@ class VariableType(StrEnum): """ if isinstance(var, str): return cls.STRING - elif isinstance(var, (int, float)): - return cls.NUMBER elif isinstance(var, bool): return cls.BOOLEAN + elif isinstance(var, (int, float)): + return cls.NUMBER elif isinstance(var, FileObject) or (isinstance(var, dict) and var.get('is_file')): return cls.FILE elif isinstance(var, dict): @@ -116,7 +116,7 @@ class FileObject(BaseModel): content_cache: dict = Field(default_factory=dict) is_file: bool - _byte_content: bytes | None = None + _byte_content: bytes | None = PrivateAttr(default=None) def get_content(self): return self._byte_content diff --git a/api/app/core/workflow/variable/variable_objects.py b/api/app/core/workflow/variable/variable_objects.py index 63437fd9..79e023c1 100644 --- a/api/app/core/workflow/variable/variable_objects.py +++ b/api/app/core/workflow/variable/variable_objects.py @@ -10,6 +10,7 @@ T = TypeVar("T", bound=BaseVariable) class StringVariable(BaseVariable): + value: str type = 'str' def valid_value(self, value) -> str: @@ -22,6 +23,7 @@ class StringVariable(BaseVariable): class NumberVariable(BaseVariable): + value: int | float type = 'number' def valid_value(self, value) -> int | float: @@ -34,6 +36,7 @@ class NumberVariable(BaseVariable): class BooleanVariable(BaseVariable): + value: bool type = 'boolean' def valid_value(self, value) -> bool: @@ -46,11 +49,12 @@ class BooleanVariable(BaseVariable): class DictVariable(BaseVariable): + value: dict type = 'object' def valid_value(self, value) -> dict: if not isinstance(value, dict): - raise TypeError(f"Value must be a dict - {type(value)}:{value}") + raise TypeError(f"Value must be a dict - {type(value)}:{value}") return value def to_literal(self) -> str: @@ -58,6 +62,7 @@ class DictVariable(BaseVariable): class FileVariable(BaseVariable): + value: FileObject type = 'file' def valid_value(self, value) -> FileObject: @@ -102,6 +107,7 @@ class FileVariable(BaseVariable): class ArrayVariable(BaseVariable, Generic[T]): + value: list[T] type = 'array' def __init__(self, child_type: Type[T], value: list[Any]): @@ -129,6 +135,7 @@ class ArrayVariable(BaseVariable, Generic[T]): class NestedArrayVariable(BaseVariable): + value: list[ArrayVariable] type = 'array_nest' def valid_value(self, value: list[T]) -> list[T]: @@ -153,6 +160,7 @@ class NestedArrayVariable(BaseVariable): category=RuntimeWarning ) class AnyVariable(BaseVariable): + value: Any type = 'any' def valid_value(self, value: Any) -> Any: diff --git a/api/app/db.py b/api/app/db.py index 80ab2756..32261c46 100644 --- a/api/app/db.py +++ b/api/app/db.py @@ -65,6 +65,7 @@ def get_db_read() -> Generator[Session, None, None]: yield db finally: db.rollback() # 只读任务无需 commit + db.close() def get_pool_status(): diff --git a/api/app/main.py b/api/app/main.py index f4c23ca8..9e501f11 100644 --- a/api/app/main.py +++ b/api/app/main.py @@ -1,5 +1,6 @@ import os import subprocess +from app.repositories.neo4j.create_indexes import create_all_indexes from contextlib import asynccontextmanager from fastapi import FastAPI, APIRouter @@ -60,8 +61,10 @@ async def lifespan(app: FastAPI): logger.warning(f"加载预定义模型时出错: {str(e)}") else: logger.info("预定义模型加载已禁用 (LOAD_MODEL=false)") - + await create_all_indexes() logger.info("应用程序启动完成") + + yield # 应用关闭事件 logger.info("应用程序正在关闭") diff --git a/api/app/models/__init__.py b/api/app/models/__init__.py index c6098a6d..e889504a 100644 --- a/api/app/models/__init__.py +++ b/api/app/models/__init__.py @@ -16,6 +16,7 @@ from .agent_app_config_model import AgentConfig from .app_release_model import AppRelease from .memory_increment_model import MemoryIncrement from .end_user_model import EndUser +from .end_user_info_model import EndUserInfo from .appshare_model import AppShare from .release_share_model import ReleaseShare from .conversation_model import Conversation, Message @@ -60,6 +61,7 @@ __all__ = [ "AppRelease", "MemoryIncrement", "EndUser", + "EndUserInfo", "AppShare", "ReleaseShare", "Conversation", diff --git a/api/app/models/end_user_info_model.py b/api/app/models/end_user_info_model.py new file mode 100644 index 00000000..c02f254c --- /dev/null +++ b/api/app/models/end_user_info_model.py @@ -0,0 +1,24 @@ +import datetime +import uuid + +from sqlalchemy import Column, DateTime, ForeignKey, String, Text, ARRAY +from sqlalchemy.dialects.postgresql import UUID, JSONB +from sqlalchemy.orm import relationship + +from app.db import Base + + +class EndUserInfo(Base): + """终端用户信息表 - 存储用户的别名和扩展信息""" + __tablename__ = "end_user_info" + + id = Column(UUID(as_uuid=True), primary_key=True, default=uuid.uuid4, nullable=False, index=True) + end_user_id = Column(UUID(as_uuid=True), ForeignKey("end_users.id"), nullable=False, index=True, comment="关联的终端用户ID") + other_name = Column(String, nullable=False, comment="关联的用户名称") + aliases = Column(ARRAY(String), nullable=True, comment="用户别名列表(字符串数组)") + meta_data = Column(JSONB, nullable=True, comment="用户相关的扩展信息(JSON格式)") + created_at = Column(DateTime, default=datetime.datetime.now, comment="创建时间") + updated_at = Column(DateTime, default=datetime.datetime.now, onupdate=datetime.datetime.now, comment="更新时间") + + # 与 EndUser 的关系 + end_user = relationship("EndUser", back_populates="info") diff --git a/api/app/models/end_user_model.py b/api/app/models/end_user_model.py index 60600fcf..ff46786a 100644 --- a/api/app/models/end_user_model.py +++ b/api/app/models/end_user_model.py @@ -22,6 +22,14 @@ class EndUser(Base): created_at = Column(DateTime, default=datetime.datetime.now) updated_at = Column(DateTime, default=datetime.datetime.now, onupdate=datetime.datetime.now) + # 用户档案字段 - User Profile Fields + position = Column(String, nullable=True, comment="职位") + department = Column(String, nullable=True, comment="部门") + contact = Column(String, nullable=True, comment="联系方式") + phone = Column(String, nullable=True, comment="电话") + hire_date = Column(DateTime, nullable=True, comment="入职日期") + updatetime_profile = Column(DateTime, nullable=True, comment="核心档案信息最后更新时间") + memory_config_id = Column( UUID(as_uuid=True), ForeignKey("memory_config.config_id"), @@ -30,14 +38,6 @@ class EndUser(Base): comment="关联的记忆配置ID" ) - # 用户基本信息字段 - position = Column(String, nullable=True, comment="职位") - department = Column(String, nullable=True, comment="部门") - contact = Column(String, nullable=True, comment="联系方式") - phone = Column(String, nullable=True, comment="电话") - hire_date = Column(DateTime, nullable=True, comment="入职日期") - updatetime_profile = Column(DateTime, nullable=True, comment="核心档案信息最后更新时间") - # 用户摘要四个维度 - User Summary Four Dimensions user_summary = Column(Text, nullable=True, comment="缓存的用户摘要(基本介绍)") personality_traits = Column(Text, nullable=True, comment="性格特点") @@ -65,4 +65,7 @@ class EndUser(Base): ) # 与 WorkSpace 的反向关系 - workspace = relationship("Workspace", back_populates="end_users") \ No newline at end of file + workspace = relationship("Workspace", back_populates="end_users") + + # 与 EndUserInfo 的反向关系 + info = relationship("EndUserInfo", back_populates="end_user", cascade="all, delete-orphan") \ No newline at end of file diff --git a/api/app/models/memory_config_model.py b/api/app/models/memory_config_model.py index 1095a386..616f7f3a 100644 --- a/api/app/models/memory_config_model.py +++ b/api/app/models/memory_config_model.py @@ -30,6 +30,9 @@ class MemoryConfig(Base): llm_id = Column(String, nullable=True, comment="LLM模型配置ID") embedding_id = Column(String, nullable=True, comment="嵌入模型配置ID") rerank_id = Column(String, nullable=True, comment="重排序模型配置ID") + vision_id = Column(String, nullable=True, comment="视觉模型配置ID") + audio_id = Column(String, nullable=True, comment="语音模型配置ID") + video_id = Column(String, nullable=True, comment="视频模型配置ID") # 记忆萃取引擎配置 enable_llm_dedup_blockwise = Column(Boolean, default=True, comment="启用LLM决策去重") diff --git a/api/app/models/memory_perceptual_model.py b/api/app/models/memory_perceptual_model.py index 9fed7c5d..ae8cc1bd 100644 --- a/api/app/models/memory_perceptual_model.py +++ b/api/app/models/memory_perceptual_model.py @@ -9,7 +9,6 @@ from sqlalchemy.dialects.postgresql import JSONB from app.db import Base from app.schemas import FileType - class PerceptualType(IntEnum): VISION = 1 AUDIO = 2 diff --git a/api/app/models/models_model.py b/api/app/models/models_model.py index 23fafcef..69bedc3d 100644 --- a/api/app/models/models_model.py +++ b/api/app/models/models_model.py @@ -2,10 +2,11 @@ import datetime import uuid from enum import StrEnum -from sqlalchemy import Column, String, Boolean, DateTime, Text, ForeignKey, Enum as SQLEnum, UniqueConstraint, Integer, ARRAY, Table, text -from sqlalchemy.dialects.postgresql import UUID, JSON +from sqlalchemy import Column, String, Boolean, DateTime, Text, ForeignKey, UniqueConstraint, Integer, Table, text +from sqlalchemy.dialects.postgresql import UUID, JSON, ARRAY from sqlalchemy.orm import relationship from sqlalchemy.sql import func + from app.db import Base @@ -26,9 +27,9 @@ class ModelType(StrEnum): RERANK = "rerank" # TTS = "tts" # SPEECH2TEXT = "speech2text" - # IMAGE = "image" + IMAGE = "image" # AUDIO = "audio" - # VISION = "vision" + VIDEO = "video" class ModelProvider(StrEnum): @@ -45,6 +46,7 @@ class ModelProvider(StrEnum): XINFERENCE = "xinference" GPUSTACK = "gpustack" BEDROCK = "bedrock" + VOLCANO = "volcano" COMPOSITE = "composite" diff --git a/api/app/models/tenant_model.py b/api/app/models/tenant_model.py index 044857d2..a92b5629 100644 --- a/api/app/models/tenant_model.py +++ b/api/app/models/tenant_model.py @@ -23,6 +23,17 @@ class Tenants(Base): # 国际化语言配置字段 default_language = Column(String(10), nullable=False, default='zh', server_default='zh', index=True) # 租户默认语言 supported_languages = Column(ARRAY(String(10)), nullable=False, default=lambda: ['zh', 'en'], server_default=text("'{zh,en}'")) # 租户支持的语言列表 + + # 租户联系信息 + contact_name = Column(String(100), nullable=True) # 联系人姓名 + contact_email = Column(String(255), nullable=True) # 联系人邮箱 + contact_phone = Column(String(50), nullable=True) # 联系人电话 + + # 租户套餐信息 + plan = Column(String(50), nullable=True) # 套餐类型 + plan_expired_at = Column(DateTime, nullable=True) # 套餐到期时间 + api_ops_rate_limit = Column(String(100), nullable=True) # API 调用频率限制 + status = Column(String(50), nullable=True, default='active') # 租户状态 # Relationship to users - one tenant has many users users = relationship("User", back_populates="tenant") diff --git a/api/app/models/user_model.py b/api/app/models/user_model.py index b6de28ec..c0b17d14 100644 --- a/api/app/models/user_model.py +++ b/api/app/models/user_model.py @@ -9,7 +9,7 @@ class User(Base): __tablename__ = "users" id = Column(UUID(as_uuid=True), primary_key=True, default=uuid.uuid4, index=True) - username = Column(String, unique=True, index=True, nullable=False) + username = Column(String, index=True, nullable=False) # 社区版:用户名不唯一,仅邮箱唯一 email = Column(String, unique=True, index=True, nullable=False) hashed_password = Column(String, nullable=False) is_active = Column(Boolean, default=True, nullable=False) @@ -19,9 +19,12 @@ class User(Base): last_login_at = Column(DateTime, nullable=True) # 最后登录时间,可为空 # SSO 外部关联字段 - external_id = Column(String(100), nullable=True) # 外部用户ID + external_id = Column(String(100), nullable=True) # 外部用户 ID external_source = Column(String(50), nullable=True) # 来源系统 + # 用户联系方式 + phone = Column(String(50), nullable=True) # 用户电话 + # 用户语言偏好 preferred_language = Column(String(10), server_default=text("'zh'"), default='zh', nullable=False, index=True) # 用户偏好语言,默认中文 diff --git a/api/app/repositories/conversation_repository.py b/api/app/repositories/conversation_repository.py index 90f2d6ec..0676a255 100644 --- a/api/app/repositories/conversation_repository.py +++ b/api/app/repositories/conversation_repository.py @@ -199,6 +199,96 @@ class ConversationRepository: ) return conversations, total + def list_app_conversations( + self, + app_id: uuid.UUID, + workspace_id: uuid.UUID, + is_draft: Optional[bool] = None, + page: int = 1, + pagesize: int = 20 + ) -> tuple[list[Conversation], int]: + """ + 查询应用日志会话列表(带分页和过滤) + + Args: + app_id: 应用 ID + workspace_id: 工作空间 ID + is_draft: 是否草稿会话(None 表示不过滤) + page: 页码(从 1 开始) + pagesize: 每页数量 + + Returns: + Tuple[List[Conversation], int]: (会话列表,总数) + """ + stmt = select(Conversation).where( + Conversation.app_id == app_id, + Conversation.workspace_id == workspace_id, + Conversation.is_active.is_(True) + ) + + if is_draft is not None: + stmt = stmt.where(Conversation.is_draft == is_draft) + + # Calculate total number of records + total = int(self.db.execute( + select(func.count()).select_from(stmt.subquery()) + ).scalar_one()) + + # Apply pagination + stmt = stmt.order_by(desc(Conversation.updated_at)) + stmt = stmt.offset((page - 1) * pagesize).limit(pagesize) + + conversations = list(self.db.scalars(stmt).all()) + + logger.info( + "Listed app conversations successfully", + extra={ + "app_id": str(app_id), + "workspace_id": str(workspace_id), + "returned": len(conversations), + "total": total + } + ) + return conversations, total + + def get_conversation_for_app_log( + self, + conversation_id: uuid.UUID, + app_id: uuid.UUID, + workspace_id: uuid.UUID + ) -> Conversation: + """ + 查询应用日志的会话详情 + + Args: + conversation_id: 会话 ID + app_id: 应用 ID + workspace_id: 工作空间 ID + + Returns: + Conversation: 会话对象 + + Raises: + ResourceNotFoundException: 当会话不存在时 + """ + logger.info(f"Fetching conversation for app log: {conversation_id}") + + stmt = select(Conversation).where( + Conversation.id == conversation_id, + Conversation.app_id == app_id, + Conversation.workspace_id == workspace_id, + Conversation.is_active.is_(True) + ) + + conversation = self.db.scalars(stmt).first() + + if not conversation: + logger.warning(f"Conversation not found: {conversation_id}") + raise ResourceNotFoundException("会话", str(conversation_id)) + + logger.info(f"Conversation fetched successfully: {conversation_id}") + return conversation + def soft_delete_conversation_by_conversation_id( self, conversation_id: uuid.UUID, @@ -290,6 +380,34 @@ class MessageRepository: self.db.add(message) return message + def get_messages_by_conversation( + self, + conversation_id: uuid.UUID + ) -> list[Message]: + """ + 查询会话的所有消息(按时间正序) + + Args: + conversation_id: 会话 ID + + Returns: + List[Message]: 消息列表 + """ + stmt = select(Message).where( + Message.conversation_id == conversation_id + ).order_by(Message.created_at) + + messages = list(self.db.scalars(stmt).all()) + + logger.info( + "Fetched messages for conversation", + extra={ + "conversation_id": str(conversation_id), + "message_count": len(messages) + } + ) + return messages + def get_message_by_conversation_id( self, conversation_id: uuid.UUID, diff --git a/api/app/repositories/end_user_info_repository.py b/api/app/repositories/end_user_info_repository.py new file mode 100644 index 00000000..f627b46f --- /dev/null +++ b/api/app/repositories/end_user_info_repository.py @@ -0,0 +1,71 @@ +""" +终端用户信息仓储层 +""" +import uuid +from typing import List, Optional +from sqlalchemy.orm import Session + +from app.models.end_user_info_model import EndUserInfo +from app.core.logging_config import get_logger + +logger = get_logger(__name__) + + +class EndUserInfoRepository: + """终端用户信息仓储类""" + + def __init__(self, db: Session): + self.db = db + + def create(self, end_user_id: uuid.UUID, other_name: str, aliases: List[str] = None, meta_data: dict = None) -> EndUserInfo: + """创建终端用户信息""" + end_user_info = EndUserInfo( + end_user_id=end_user_id, + other_name=other_name, + aliases=aliases or [], + meta_data=meta_data + ) + self.db.add(end_user_info) + self.db.commit() + self.db.refresh(end_user_info) + logger.info(f"创建终端用户信息: end_user_id={end_user_id}, aliases={aliases}") + return end_user_info + + def get_by_id(self, info_id: uuid.UUID) -> Optional[EndUserInfo]: + """根据ID获取用户信息""" + return self.db.query(EndUserInfo).filter(EndUserInfo.id == info_id).first() + + + def get_by_end_user_id(self, end_user_id: uuid.UUID) -> Optional[EndUserInfo]: + """获取用户的信息记录""" + return self.db.query(EndUserInfo).filter(EndUserInfo.end_user_id == end_user_id).first() + + def update(self, info_id: uuid.UUID, aliases: List[str] = None, meta_data: dict = None) -> Optional[EndUserInfo]: + """更新用户信息""" + end_user_info = self.get_by_id(info_id) + if end_user_info: + if aliases is not None: + end_user_info.aliases = aliases + if meta_data is not None: + end_user_info.meta_data = meta_data + self.db.commit() + self.db.refresh(end_user_info) + logger.info(f"更新终端用户信息: info_id={info_id}") + return end_user_info + + def delete(self, info_id: uuid.UUID) -> bool: + """删除用户信息""" + end_user_info = self.get_by_id(info_id) + if end_user_info: + self.db.delete(end_user_info) + self.db.commit() + logger.info(f"删除终端用户信息: info_id={info_id}") + return True + return False + + def delete_by_end_user_id(self, end_user_id: uuid.UUID) -> int: + """删除用户的所有信息记录""" + count = self.db.query(EndUserInfo).filter(EndUserInfo.end_user_id == end_user_id).delete() + self.db.commit() + logger.info(f"删除用户所有信息记录: end_user_id={end_user_id}, count={count}") + return count diff --git a/api/app/repositories/end_user_repository.py b/api/app/repositories/end_user_repository.py index 71c93634..aad80707 100644 --- a/api/app/repositories/end_user_repository.py +++ b/api/app/repositories/end_user_repository.py @@ -7,6 +7,7 @@ from sqlalchemy.orm import Session from app.core.logging_config import get_db_logger from app.models.app_model import App from app.models.end_user_model import EndUser +from app.models.end_user_info_model import EndUserInfo from app.models.workspace_model import Workspace # 获取数据库专用日志器 @@ -70,7 +71,8 @@ class EndUserRepository: app_id: uuid.UUID, workspace_id: uuid.UUID, other_id: str, - original_user_id: Optional[str] = None + original_user_id: Optional[str] = None, + other_name: Optional[str] = None ) -> EndUser: """获取或创建终端用户 @@ -79,6 +81,7 @@ class EndUserRepository: workspace_id: 工作空间ID other_id: 第三方ID original_user_id: 原始用户ID (存储到 other_id) + other_name: 用户名称(用于创建 EndUserInfo) """ try: # 尝试查找现有用户 @@ -106,10 +109,22 @@ class EndUserRepository: other_id=other_id ) self.db.add(end_user) + self.db.flush() # 刷新以获取 end_user.id,但不提交事务 + + # 创建对应的 EndUserInfo 记录 + end_user_info = EndUserInfo( + end_user_id=end_user.id, + other_name=other_name or "", # 如果没有提供 other_name,使用空字符串 + aliases=[], + meta_data={} + ) + self.db.add(end_user_info) + + # 一起提交 self.db.commit() self.db.refresh(end_user) - db_logger.info(f"创建新终端用户: (other_id: {other_id}) for workspace {workspace_id}") + db_logger.info(f"创建新终端用户及其信息: (other_id: {other_id}) for workspace {workspace_id}") return end_user except Exception as e: @@ -117,6 +132,82 @@ class EndUserRepository: db_logger.error(f"获取或创建终端用户时出错: {str(e)}") raise + def get_or_create_end_user_with_config( + self, + app_id: Optional[uuid.UUID], + workspace_id: uuid.UUID, + other_id: str, + memory_config_id: Optional[uuid.UUID] = None, + other_name: Optional[str] = None + ) -> EndUser: + """获取或创建终端用户,并在单次事务中关联记忆配置。 + + 与 get_or_create_end_user 类似,但额外支持在创建/获取时 + 一并设置 memory_config_id,避免多次提交。 + + Args: + app_id: 应用ID(可为 None) + workspace_id: 工作空间ID + other_id: 第三方ID + memory_config_id: 记忆配置ID(可选,仅在用户尚无配置时设置) + other_name: 用户名称(用于创建 EndUserInfo) + + Returns: + EndUser: 终端用户对象(已关联记忆配置) + """ + try: + end_user = ( + self.db.query(EndUser) + .filter( + EndUser.workspace_id == workspace_id, + EndUser.other_id == other_id + ) + .order_by(EndUser.created_at.asc()) + .first() + ) + + if end_user: + db_logger.debug(f"找到现有终端用户: workspace_id={workspace_id}, other_id={other_id}") + if app_id is not None: + end_user.app_id = app_id + if memory_config_id and not end_user.memory_config_id: + end_user.memory_config_id = memory_config_id + self.db.commit() + self.db.refresh(end_user) + return end_user + + # 创建新用户 + end_user = EndUser( + app_id=app_id, + workspace_id=workspace_id, + other_id=other_id, + memory_config_id=memory_config_id, + ) + self.db.add(end_user) + self.db.flush() + + end_user_info = EndUserInfo( + end_user_id=end_user.id, + other_name=other_name or "", + aliases=[], + meta_data={} + ) + self.db.add(end_user_info) + + self.db.commit() + self.db.refresh(end_user) + + db_logger.info( + f"创建新终端用户及其信息: (other_id: {other_id}) for workspace {workspace_id}, " + f"memory_config_id={memory_config_id}" + ) + return end_user + + except Exception as e: + self.db.rollback() + db_logger.error(f"获取或创建终端用户(含配置)时出错: {str(e)}") + raise + def get_by_id(self, end_user_id: uuid.UUID) -> Optional[EndUser]: """根据ID获取终端用户(用于缓存操作) @@ -500,6 +591,51 @@ class EndUserRepository: ) raise + def batch_update_memory_config_id_by_app( + self, + app_id: uuid.UUID, + memory_config_id: uuid.UUID + ) -> int: + """批量更新应用下所有终端用户的 memory_config_id + + Args: + app_id: 应用ID + memory_config_id: 新的记忆配置ID + + Returns: + int: 更新的终端用户数量 + + Raises: + Exception: 数据库操作失败时抛出 + """ + try: + from sqlalchemy import update + + stmt = ( + update(EndUser) + .where(EndUser.app_id == app_id) + .values(memory_config_id=memory_config_id) + ) + + result = self.db.execute(stmt) + self.db.commit() + + updated_count = result.rowcount + + db_logger.info( + f"批量更新终端用户记忆配置: app_id={app_id}, " + f"memory_config_id={memory_config_id}, updated_count={updated_count}" + ) + + return updated_count + except Exception as e: + self.db.rollback() + db_logger.error( + f"批量更新终端用户记忆配置时出错: app_id={app_id}, " + f"memory_config_id={memory_config_id}, error={str(e)}" + ) + raise + def count_by_memory_config_id( self, memory_config_id: uuid.UUID diff --git a/api/app/repositories/home_page_repository.py b/api/app/repositories/home_page_repository.py index bcb3b622..6d74bcaf 100644 --- a/api/app/repositories/home_page_repository.py +++ b/api/app/repositories/home_page_repository.py @@ -2,7 +2,7 @@ from datetime import datetime, timedelta from sqlalchemy.orm import Session from sqlalchemy import func from uuid import UUID -from typing import Dict +from typing import Dict, Optional, Any from app.models.end_user_model import EndUser from app.models.user_model import User @@ -190,4 +190,63 @@ class HomePageRepository: user_count_dict = {workspace_id: count for workspace_id, count in user_counts} - return workspaces, app_count_dict, user_count_dict \ No newline at end of file + return workspaces, app_count_dict, user_count_dict + + @staticmethod + def get_version_introduction(db: Session, version: str) -> Optional[Dict[str, Any]]: + """ + 从数据库获取版本说明(优先读取已发布的版本) + 使用反射方式读取表结构,不依赖 premium 模型类 + + Args: + db: 数据库会话 + version: 版本号,如 "v0.2.7" + + Returns: + 版本说明字典,格式与 version_info.json 一致 + 如果数据库中没有该版本,返回 None + """ + try: + from sqlalchemy import Table, MetaData + + metadata = MetaData() + version_notes = Table('version_notes', metadata, autoload_with=db.engine) + version_note_items = Table('version_note_items', metadata, autoload_with=db.engine) + + note = db.query(version_notes).filter( + version_notes.c.version == version, + version_notes.c.is_published == True + ).first() + + if not note: + return None + + items = db.query(version_note_items).filter( + version_note_items.c.note_id == note.id + ).order_by(version_note_items.c.sort_order).all() + + core_upgrades = [] + for item in items: + title = item.title + content = item.content + if content: + core_upgrades.append(f"{title}
{content}") + else: + core_upgrades.append(title) + + return { + "introduction": { + "codeName": "", + "releaseDate": note.release_date.isoformat() if note.release_date else "", + "upgradePosition": "", + "coreUpgrades": core_upgrades + }, + "introduction_en": { + "codeName": "", + "releaseDate": note.release_date.isoformat() if note.release_date else "", + "upgradePosition": "", + "coreUpgrades": core_upgrades + } + } + except Exception: + return None \ No newline at end of file diff --git a/api/app/repositories/memory_config_repository.py b/api/app/repositories/memory_config_repository.py index 22f13449..e64d19a3 100644 --- a/api/app/repositories/memory_config_repository.py +++ b/api/app/repositories/memory_config_repository.py @@ -9,21 +9,22 @@ Classes: """ import uuid -from uuid import UUID from typing import Dict, List, Optional, Tuple +from uuid import UUID + +from sqlalchemy import desc, select +from sqlalchemy.orm import Session + from app.core.exceptions import BusinessException from app.core.logging_config import get_config_logger, get_db_logger from app.models.memory_config_model import MemoryConfig +from app.models.workspace_model import Workspace from app.schemas.memory_storage_schema import ( - ConfigKey, ConfigParamsCreate, ConfigUpdate, ConfigUpdateExtracted, ConfigUpdateForget, ) -from sqlalchemy import desc, select -from sqlalchemy.orm import Session - from app.utils.config_utils import resolve_config_id # 获取数据库专用日志器 @@ -157,7 +158,7 @@ class MemoryConfigRepository: return memory_config_obj @staticmethod - def query_reflection_config_by_id(db: Session, config_id: uuid.UUID|int|str) -> MemoryConfig: + def query_reflection_config_by_id(db: Session, config_id: uuid.UUID | int | str) -> MemoryConfig: """构建反思配置查询语句,通过config_id查询反思配置(SQLAlchemy text() 命名参数) Args: @@ -309,57 +310,21 @@ class MemoryConfigRepository: Returns: Optional[MemoryConfig]: 更新后的配置对象,不存在则返回None - - Raises: - ValueError: 没有字段需要更新时抛出 """ db_logger.debug(f"更新萃取配置: config_id={update.config_id}") try: - db_config = db.query(MemoryConfig).filter(MemoryConfig.config_id == update.config_id).first() + stmt = select(MemoryConfig).where(MemoryConfig.config_id == update.config_id) + db_config = db.execute(stmt).scalar_one_or_none() if not db_config: db_logger.warning(f"记忆配置不存在: config_id={update.config_id}") return None - # 更新字段映射 - field_mapping = { - # 模型选择 - "llm_id": "llm_id", - "embedding_id": "embedding_id", - "rerank_id": "rerank_id", - # 记忆萃取引擎 - "enable_llm_dedup_blockwise": "enable_llm_dedup_blockwise", - "enable_llm_disambiguation": "enable_llm_disambiguation", - "deep_retrieval": "deep_retrieval", - "t_type_strict": "t_type_strict", - "t_name_strict": "t_name_strict", - "t_overall": "t_overall", - "state": "state", - "chunker_strategy": "chunker_strategy", - # 句子提取 - "statement_granularity": "statement_granularity", - "include_dialogue_context": "include_dialogue_context", - "max_context": "max_context", - # 剪枝配置 - "pruning_enabled": "pruning_enabled", - "pruning_scene": "pruning_scene", - "pruning_threshold": "pruning_threshold", - # 自我反思配置 - "enable_self_reflexion": "enable_self_reflexion", - "iteration_period": "iteration_period", - "reflexion_range": "reflexion_range", - "baseline": "baseline", - } + update_data = update.model_dump(exclude_unset=True) + update_data.pop("config_id", None) - has_update = False - for api_field, db_field in field_mapping.items(): - value = getattr(update, api_field, None) - if value is not None: - setattr(db_config, db_field, value) - has_update = True - - if not has_update: - raise ValueError("No fields to update") + for field, value in update_data.items(): + setattr(db_config, field, value) db.commit() db.refresh(db_config) @@ -443,6 +408,9 @@ class MemoryConfigRepository: "llm_id": db_config.llm_id, "embedding_id": db_config.embedding_id, "rerank_id": db_config.rerank_id, + "vision_id": db_config.vision_id, + "audio_id": db_config.audio_id, + "video_id": db_config.video_id, "enable_llm_dedup_blockwise": db_config.enable_llm_dedup_blockwise, "enable_llm_disambiguation": db_config.enable_llm_disambiguation, "deep_retrieval": db_config.deep_retrieval, @@ -527,7 +495,10 @@ class MemoryConfigRepository: raise @staticmethod - def get_config_with_workspace(db: Session, config_id: uuid.UUID | int | str) -> Optional[tuple]: + def get_config_with_workspace( + db: Session, + config_id: uuid.UUID | int | str + ) -> Optional[tuple[MemoryConfig, Workspace]]: """Get memory config and its associated workspace information Args: @@ -542,8 +513,6 @@ class MemoryConfigRepository: """ import time - from app.models.workspace_model import Workspace - start_time = time.time() config_id = resolve_config_id(config_id, db) @@ -630,7 +599,7 @@ class MemoryConfigRepository: db_logger.debug( f"Memory config and workspace query successful: config={config.config_name}, workspace={workspace.name}") - return (config, workspace) + return config, workspace except ValueError: # Re-raise known business exceptions @@ -666,7 +635,7 @@ class MemoryConfigRepository: List[Tuple[MemoryConfig, Optional[str]]]: 配置列表,每项为 (配置对象, 场景名称) """ from app.models.ontology_scene import OntologyScene - + db_logger.debug(f"查询所有配置: workspace_id={workspace_id}") try: @@ -730,7 +699,7 @@ class MemoryConfigRepository: Optional[MemoryConfig]: 默认配置对象,不存在则返回None """ db_logger.debug(f"查询工作空间默认配置: workspace_id={workspace_id}") - + try: # 优先查找显式标记为默认的配置 stmt = ( @@ -742,13 +711,13 @@ class MemoryConfigRepository: ) .limit(1) ) - + config = db.scalars(stmt).first() - + if config: db_logger.debug(f"找到默认配置: config_id={config.config_id}") return config - + # 回退:获取最早创建的活跃配置 stmt = ( select(MemoryConfig) @@ -759,25 +728,25 @@ class MemoryConfigRepository: .order_by(MemoryConfig.created_at.asc()) .limit(1) ) - + config = db.scalars(stmt).first() - + if config: db_logger.debug(f"使用最早创建的配置作为默认: config_id={config.config_id}") else: db_logger.warning(f"工作空间没有活跃的记忆配置: workspace_id={workspace_id}") - + return config - + except Exception as e: db_logger.error(f"查询工作空间默认配置失败: workspace_id={workspace_id} - {str(e)}") raise @staticmethod def get_with_fallback( - db: Session, - config_id: Optional[uuid.UUID], - workspace_id: uuid.UUID + db: Session, + config_id: Optional[uuid.UUID], + workspace_id: uuid.UUID ) -> Optional[MemoryConfig]: """获取记忆配置,支持回退到工作空间默认配置 @@ -792,19 +761,18 @@ class MemoryConfigRepository: Optional[MemoryConfig]: 配置对象,如果都不存在则返回None """ db_logger.debug(f"查询配置(支持回退): config_id={config_id}, workspace_id={workspace_id}") - + if not config_id: db_logger.debug("config_id 为空,使用工作空间默认配置") return MemoryConfigRepository.get_workspace_default(db, workspace_id) - + config = db.get(MemoryConfig, config_id) - + if config: return config - + db_logger.warning( f"配置不存在,回退到工作空间默认配置: missing_config_id={config_id}, workspace_id={workspace_id}" ) - - return MemoryConfigRepository.get_workspace_default(db, workspace_id) + return MemoryConfigRepository.get_workspace_default(db, workspace_id) diff --git a/api/app/repositories/model_repository.py b/api/app/repositories/model_repository.py index f49227d3..8c477d39 100644 --- a/api/app/repositories/model_repository.py +++ b/api/app/repositories/model_repository.py @@ -1,14 +1,15 @@ -from sqlalchemy.orm import Session, joinedload, selectinload -from sqlalchemy import and_, or_, func, desc, select -from typing import List, Optional, Dict, Any, Tuple import uuid +from typing import List, Optional, Dict, Any, Tuple +from sqlalchemy import and_, or_, func, desc +from sqlalchemy.orm import Session, joinedload + +from app.core.logging_config import get_db_logger from app.models.models_model import ModelConfig, ModelApiKey, ModelType, ModelBase, model_config_api_key_association from app.schemas.model_schema import ( ModelConfigUpdate, ModelApiKeyCreate, ModelApiKeyUpdate, ModelConfigQuery, ModelConfigQueryNew ) -from app.core.logging_config import get_db_logger # 获取数据库专用日志器 db_logger = get_db_logger() @@ -137,6 +138,9 @@ class ModelConfigRepository: type_values.append(ModelType.LLM) filters.append(ModelConfig.type.in_(type_values)) + if query.capability: + filters.append(ModelConfig.capability.contains(query.capability)) + if query.is_active is not None: filters.append(ModelConfig.is_active == query.is_active) @@ -435,7 +439,6 @@ class ModelConfigRepository: ModelConfig.is_public ), ModelConfig.provider == provider, - ModelConfig.is_active, ~ModelConfig.is_composite ) ).all() diff --git a/api/app/repositories/neo4j/add_nodes.py b/api/app/repositories/neo4j/add_nodes.py index 42c178b3..1939a062 100644 --- a/api/app/repositories/neo4j/add_nodes.py +++ b/api/app/repositories/neo4j/add_nodes.py @@ -1,17 +1,22 @@ +import logging from typing import List, Optional -from app.repositories.neo4j.cypher_queries import DIALOGUE_NODE_SAVE, STATEMENT_NODE_SAVE, CHUNK_NODE_SAVE,MEMORY_SUMMARY_NODE_SAVE from app.core.memory.models.graph_models import DialogueNode, StatementNode, ChunkNode, MemorySummaryNode +from app.repositories.neo4j.cypher_queries import DIALOGUE_NODE_SAVE, STATEMENT_NODE_SAVE, CHUNK_NODE_SAVE, \ + MEMORY_SUMMARY_NODE_SAVE # 使用新的仓储层 from app.repositories.neo4j.neo4j_connector import Neo4jConnector +logger = logging.getLogger(__name__) + async def delete_all_nodes(end_user_id: str, connector: Neo4jConnector): """Delete all nodes in the database.""" result = await connector.execute_query(f"MATCH (n {{end_user_id: '{end_user_id}'}}) DETACH DELETE n") - print(f"All end_user_id: {end_user_id} node and edge deleted successfully") + logger.warning(f"All end_user_id: {end_user_id} node and edge deleted successfully") return result + async def add_dialogue_nodes(dialogues: List[DialogueNode], connector: Neo4jConnector) -> Optional[List[str]]: """Add dialogue nodes to Neo4j database. @@ -23,7 +28,7 @@ async def add_dialogue_nodes(dialogues: List[DialogueNode], connector: Neo4jConn List of created node UUIDs or None if failed """ if not dialogues: - print("No dialogues to save") + logger.info("No dialogues to save") return [] try: @@ -48,11 +53,11 @@ async def add_dialogue_nodes(dialogues: List[DialogueNode], connector: Neo4jConn ) created_uuids = [record["uuid"] for record in result] - print(f"Successfully created {len(created_uuids)} dialogue nodes: {created_uuids}") + logger.info(f"Successfully created {len(created_uuids)} dialogue nodes: {created_uuids}") return created_uuids except Exception as e: - print(f"Error creating dialogue nodes: {e}") + logger.error(f"Error creating dialogue nodes: {e}") return None @@ -67,7 +72,7 @@ async def add_statement_nodes(statements: List[StatementNode], connector: Neo4jC List of created node UUIDs or None if failed """ if not statements: - print("No statements to save") + logger.info("No statements to save") return [] try: @@ -120,13 +125,14 @@ async def add_statement_nodes(statements: List[StatementNode], connector: Neo4jC ) created_uuids = [record["uuid"] for record in result] - print(f"Successfully created {len(created_uuids)} statement nodes") + logger.info(f"Successfully created {len(created_uuids)} statement nodes") return created_uuids except Exception as e: - print(f"Error creating statement nodes: {e}") + logger.error(f"Error creating statement nodes: {e}") return None + async def add_chunk_nodes(chunks: List[ChunkNode], connector: Neo4jConnector) -> Optional[List[str]]: """Add chunk nodes to Neo4j in batch. @@ -138,7 +144,7 @@ async def add_chunk_nodes(chunks: List[ChunkNode], connector: Neo4jConnector) -> List of created chunk UUIDs or None if failed """ if not chunks: - print("No chunk nodes to add") + logger.info("No chunk nodes to add") return [] try: @@ -171,16 +177,18 @@ async def add_chunk_nodes(chunks: List[ChunkNode], connector: Neo4jConnector) -> ) created_uuids = [record["uuid"] for record in result] - print(f"Successfully created {len(created_uuids)} chunk nodes") + logger.info(f"Successfully created {len(created_uuids)} chunk nodes") return created_uuids except Exception as e: - print(f"Error creating chunk nodes: {e}") + logger.error(f"Error creating chunk nodes: {e}") return None - -async def add_memory_summary_nodes(summaries: List[MemorySummaryNode], connector: Neo4jConnector) -> Optional[List[str]]: +async def add_memory_summary_nodes( + summaries: List[MemorySummaryNode], + connector: Neo4jConnector +) -> Optional[List[str]]: """Add memory summary nodes to Neo4j in batch. Args: @@ -191,7 +199,7 @@ async def add_memory_summary_nodes(summaries: List[MemorySummaryNode], connector List of created summary node ids or None if failed """ if not summaries: - print("No memory summary nodes to add") + logger.info("No memory summary nodes to add") return [] try: @@ -211,16 +219,14 @@ async def add_memory_summary_nodes(summaries: List[MemorySummaryNode], connector "summary_embedding": s.summary_embedding if s.summary_embedding else None, "config_id": s.config_id, # 添加 config_id }) - + result = await connector.execute_query( MEMORY_SUMMARY_NODE_SAVE, summaries=flattened ) created_ids = [record.get("uuid") for record in result] - print(f"Successfully saved {len(created_ids)} MemorySummary nodes to Neo4j") + logger.info(f"Successfully saved {len(created_ids)} MemorySummary nodes to Neo4j") return created_ids except Exception as e: - print(f"Failed to save MemorySummary nodes to Neo4j: {e}") + logger.error(f"Failed to save MemorySummary nodes to Neo4j: {e}") return None - - diff --git a/api/app/repositories/neo4j/community_repository.py b/api/app/repositories/neo4j/community_repository.py index e89ee451..bd448c99 100644 --- a/api/app/repositories/neo4j/community_repository.py +++ b/api/app/repositories/neo4j/community_repository.py @@ -13,9 +13,14 @@ from app.repositories.neo4j.cypher_queries import ( ENTITY_LEAVE_ALL_COMMUNITIES, GET_ENTITY_NEIGHBORS, GET_ALL_ENTITIES_FOR_USER, + GET_ENTITY_COUNT_FOR_USER, + GET_ALL_ENTITY_IDS_FOR_USER, + GET_ENTITIES_PAGE, GET_COMMUNITY_MEMBERS, + GET_COMMUNITY_RELATIONSHIPS, GET_ALL_COMMUNITY_MEMBERS_BATCH, GET_ALL_ENTITY_NEIGHBORS_BATCH, + GET_ENTITY_NEIGHBORS_BATCH_FOR_IDS, CHECK_USER_HAS_COMMUNITIES, UPDATE_COMMUNITY_MEMBER_COUNT, UPDATE_COMMUNITY_METADATA, @@ -23,6 +28,7 @@ from app.repositories.neo4j.cypher_queries import ( GET_INCOMPLETE_COMMUNITIES_WITH_EMBEDDING, CHECK_COMMUNITY_IS_COMPLETE, CHECK_COMMUNITY_IS_COMPLETE_WITH_EMBEDDING, + BATCH_UPDATE_COMMUNITY_METADATA, ) logger = logging.getLogger(__name__) @@ -114,10 +120,69 @@ class CommunityRepository: logger.error(f"get_all_entities failed: {e}") return [] + async def get_entity_count(self, end_user_id: str) -> int: + """仅返回用户实体总数,不加载实体数据。""" + try: + result = await self.connector.execute_query( + GET_ENTITY_COUNT_FOR_USER, + end_user_id=end_user_id, + ) + return result[0]["entity_count"] if result else 0 + except Exception as e: + logger.error(f"get_entity_count failed: {e}") + return 0 + + async def get_all_entity_ids(self, end_user_id: str) -> List[str]: + """仅返回用户所有实体 ID 列表,不加载 embedding 等大字段。""" + try: + result = await self.connector.execute_query( + GET_ALL_ENTITY_IDS_FOR_USER, + end_user_id=end_user_id, + ) + return [r["id"] for r in result] + except Exception as e: + logger.error(f"get_all_entity_ids failed: {e}") + return [] + + async def get_entities_page( + self, end_user_id: str, skip: int, limit: int + ) -> List[Dict]: + """分页拉取实体,用于全量聚类分批处理。""" + try: + return await self.connector.execute_query( + GET_ENTITIES_PAGE, + end_user_id=end_user_id, + skip=skip, + limit=limit, + ) + except Exception as e: + logger.error(f"get_entities_page failed: {e}") + return [] + + async def get_entity_neighbors_for_ids( + self, entity_ids: List[str], end_user_id: str + ) -> Dict[str, List[Dict]]: + """批量拉取指定实体列表的邻居,返回 {entity_id: [neighbors]}。""" + try: + rows = await self.connector.execute_query( + GET_ENTITY_NEIGHBORS_BATCH_FOR_IDS, + entity_ids=entity_ids, + end_user_id=end_user_id, + ) + result: Dict[str, List[Dict]] = {} + for row in rows: + eid = row["entity_id"] + neighbor = {k: v for k, v in row.items() if k != "entity_id"} + result.setdefault(eid, []).append(neighbor) + return result + except Exception as e: + logger.error(f"get_entity_neighbors_for_ids failed: {e}") + return {} + async def get_community_members( self, community_id: str, end_user_id: str ) -> List[Dict]: - """查询社区成员列表。""" + """查询社区成员列表(含 example 字段)。""" try: return await self.connector.execute_query( GET_COMMUNITY_MEMBERS, @@ -128,6 +193,20 @@ class CommunityRepository: logger.error(f"get_community_members failed: {e}") return [] + async def get_community_relationships( + self, community_id: str, end_user_id: str + ) -> List[Dict]: + """查询社区内实体间的关系三元组(subject, predicate, object)。""" + try: + return await self.connector.execute_query( + GET_COMMUNITY_RELATIONSHIPS, + community_id=community_id, + end_user_id=end_user_id, + ) + except Exception as e: + logger.error(f"get_community_relationships failed: {e}") + return [] + async def get_all_community_members_batch( self, community_ids: List[str], end_user_id: str ) -> Dict[str, List[Dict]]: @@ -221,5 +300,27 @@ class CommunityRepository: ) return bool(result) except Exception as e: - logger.error(f"update_community_metadata failed: {e}") + logger.error(f"update_community_metadata failed: {e}", exc_info=True) + return False + + async def batch_update_community_metadata( + self, + communities: List[Dict], + ) -> bool: + """批量更新多个社区的元数据。 + + Args: + communities: 每项包含 community_id, end_user_id, name, summary, + core_entities, summary_embedding + """ + if not communities: + return True + try: + await self.connector.execute_query( + BATCH_UPDATE_COMMUNITY_METADATA, + communities=communities, + ) + return True + except Exception as e: + logger.error(f"batch_update_community_metadata failed: {e}") return False diff --git a/api/app/repositories/neo4j/create_indexes.py b/api/app/repositories/neo4j/create_indexes.py index 55dead1b..5132aa09 100644 --- a/api/app/repositories/neo4j/create_indexes.py +++ b/api/app/repositories/neo4j/create_indexes.py @@ -1,55 +1,47 @@ +import asyncio from app.repositories.neo4j.neo4j_connector import Neo4jConnector - - async def create_fulltext_indexes(): """Create full-text indexes for keyword search with BM25 scoring.""" connector = Neo4jConnector() try: - print("\n" + "=" * 70) - print("Creating Full-Text Indexes (for keyword search)") - print("=" * 70) + # 创建 Statements 索引 await connector.execute_query(""" CREATE FULLTEXT INDEX statementsFulltext IF NOT EXISTS FOR (s:Statement) ON EACH [s.statement] OPTIONS { indexConfig: { `fulltext.analyzer`: 'cjk' } } - """) - print("✓ Created: statementsFulltext") + """) # # 创建 Dialogues 索引 # await connector.execute_query(""" # CREATE FULLTEXT INDEX dialoguesFulltext IF NOT EXISTS FOR (d:Dialogue) ON EACH [d.content] # OPTIONS { indexConfig: { `fulltext.analyzer`: 'cjk' } } # """) - # 创建 Entities 索引 await connector.execute_query(""" CREATE FULLTEXT INDEX entitiesFulltext IF NOT EXISTS FOR (e:ExtractedEntity) ON EACH [e.name] OPTIONS { indexConfig: { `fulltext.analyzer`: 'cjk' } } - """) - print("✓ Created: entitiesFulltext") + """) # 创建 Chunks 索引 await connector.execute_query(""" CREATE FULLTEXT INDEX chunksFulltext IF NOT EXISTS FOR (c:Chunk) ON EACH [c.content] OPTIONS { indexConfig: { `fulltext.analyzer`: 'cjk' } } - """) - print("✓ Created: chunksFulltext") + """) # 创建 MemorySummary 索引 await connector.execute_query(""" CREATE FULLTEXT INDEX summariesFulltext IF NOT EXISTS FOR (m:MemorySummary) ON EACH [m.content] OPTIONS { indexConfig: { `fulltext.analyzer`: 'cjk' } } + """) + # 创建 Community 索引 + await connector.execute_query(""" + CREATE FULLTEXT INDEX communitiesFulltext IF NOT EXISTS FOR (c:Community) ON EACH [c.name, c.summary] + OPTIONS { indexConfig: { `fulltext.analyzer`: 'cjk' } } """) - print("✓ Created: summariesFulltext") - print("\nFull-text indexes created successfully with BM25 support.") - except Exception as e: - print(f"✗ Error creating full-text indexes: {e}") finally: await connector.close() - - async def create_vector_indexes(): """Create vector indexes for fast embedding similarity search. @@ -58,12 +50,7 @@ async def create_vector_indexes(): """ connector = Neo4jConnector() try: - print("\n" + "=" * 70) - print("Creating Vector Indexes (for embedding search)") - print("=" * 70) - print("Note: Adjust vector.dimensions if using different embedding model") - print(" Current setting: 1024 dimensions (for bge-m3)") - print() + # Statement embedding index await connector.execute_query(""" @@ -75,7 +62,7 @@ async def create_vector_indexes(): `vector.similarity_function`: 'cosine' }} """) - print("✓ Created: statement_embedding_index") + # Chunk embedding index await connector.execute_query(""" @@ -87,7 +74,7 @@ async def create_vector_indexes(): `vector.similarity_function`: 'cosine' }} """) - print("✓ Created: chunk_embedding_index") + # Entity name embedding index await connector.execute_query(""" @@ -99,7 +86,7 @@ async def create_vector_indexes(): `vector.similarity_function`: 'cosine' }} """) - print("✓ Created: entity_embedding_index") + # Memory summary embedding index await connector.execute_query(""" @@ -111,7 +98,17 @@ async def create_vector_indexes(): `vector.similarity_function`: 'cosine' }} """) - print("✓ Created: summary_embedding_index") + + # Community summary embedding index + await connector.execute_query(""" + CREATE VECTOR INDEX community_summary_embedding_index IF NOT EXISTS + FOR (c:Community) + ON c.summary_embedding + OPTIONS {indexConfig: { + `vector.dimensions`: 1024, + `vector.similarity_function`: 'cosine' + }} + """) # Dialogue embedding index (optional) await connector.execute_query(""" @@ -123,79 +120,15 @@ async def create_vector_indexes(): `vector.similarity_function`: 'cosine' }} """) - print("✓ Created: dialogue_embedding_index") - print("\nVector indexes created successfully!") - print("\nExpected performance improvement:") - print(" Before: ~1.4s for embedding search") - print(" After: ~0.05-0.2s for embedding search (10-30x faster!)") - - except Exception as e: - print(f"✗ Error creating vector indexes: {e}") finally: await connector.close() - - -async def create_config_id_indexes(): - """Create indexes on config_id fields for improved query performance. - - These indexes enable fast filtering of nodes by configuration ID, - which is essential for configuration isolation and multi-tenant scenarios. - """ - connector = Neo4jConnector() - try: - print("\n" + "=" * 70) - print("Creating Config ID Indexes") - print("=" * 70) - - # Dialogue.config_id index - await connector.execute_query(""" - CREATE INDEX dialogue_config_id_index IF NOT EXISTS - FOR (d:Dialogue) ON (d.config_id) - """) - print("✓ Created: dialogue_config_id_index") - - # Statement.config_id index - await connector.execute_query(""" - CREATE INDEX statement_config_id_index IF NOT EXISTS - FOR (s:Statement) ON (s.config_id) - """) - print("✓ Created: statement_config_id_index") - - # ExtractedEntity.config_id index - await connector.execute_query(""" - CREATE INDEX entity_config_id_index IF NOT EXISTS - FOR (e:ExtractedEntity) ON (e.config_id) - """) - print("✓ Created: entity_config_id_index") - - # MemorySummary.config_id index - await connector.execute_query(""" - CREATE INDEX summary_config_id_index IF NOT EXISTS - FOR (m:MemorySummary) ON (m.config_id) - """) - print("✓ Created: summary_config_id_index") - - print("\nConfig ID indexes created successfully!") - print("These indexes enable fast filtering by configuration ID.") - - except Exception as e: - print(f"✗ Error creating config_id indexes: {e}") - finally: - await connector.close() - - async def create_unique_constraints(): """Create uniqueness constraints for core node identifiers. - Ensures concurrent MERGE operations remain safe and prevents duplicates. """ connector = Neo4jConnector() - try: - print("\n" + "=" * 70) - print("Creating Unique Constraints") - print("=" * 70) - + try: # Dialogue.id unique await connector.execute_query( """ @@ -203,8 +136,7 @@ async def create_unique_constraints(): FOR (d:Dialogue) REQUIRE d.id IS UNIQUE """ ) - print("✓ Created: dialog_id_unique") - + # Statement.id unique await connector.execute_query( """ @@ -212,8 +144,7 @@ async def create_unique_constraints(): FOR (s:Statement) REQUIRE s.id IS UNIQUE """ ) - print("✓ Created: statement_id_unique") - + # Chunk.id unique await connector.execute_query( """ @@ -221,112 +152,13 @@ async def create_unique_constraints(): FOR (c:Chunk) REQUIRE c.id IS UNIQUE """ ) - print("✓ Created: chunk_id_unique") - - print("\nUnique constraints ensured for Dialogue, Statement, and Chunk.") - except Exception as e: - print(f"✗ Error creating unique constraints: {e}") + finally: await connector.close() - - async def create_all_indexes(): """Create all indexes and constraints in one go.""" - print("\n" + "=" * 70) - print("Neo4j Index & Constraint Setup") - print("=" * 70) - print("This will create:") - print(" 1. Full-text indexes (for keyword/BM25 search)") - print(" 2. Vector indexes (for embedding similarity search)") - print(" 3. Config ID indexes (for configuration isolation)") - print(" 4. Unique constraints (for data integrity)") - print("=" * 70) - await create_fulltext_indexes() await create_vector_indexes() - await create_config_id_indexes() await create_unique_constraints() - - print("\n" + "=" * 70) print("✓ All indexes and constraints created successfully!") - print("=" * 70) - print("\nTo verify, run in Neo4j Browser:") - print(" SHOW INDEXES") - print(" SHOW CONSTRAINTS") - print() - - -async def check_indexes(): - """Check what indexes currently exist.""" - connector = Neo4jConnector() - - try: - print("\n" + "=" * 70) - print("Checking Existing Indexes") - print("=" * 70) - query = "SHOW INDEXES" - result = await connector.execute_query(query) - - fulltext_indexes = [idx for idx in result if idx.get('type') == 'FULLTEXT'] - vector_indexes = [idx for idx in result if idx.get('type') == 'VECTOR'] - range_indexes = [idx for idx in result if idx.get('type') == 'RANGE'] - - print(f"\nFull-text indexes: {len(fulltext_indexes)}") - for idx in fulltext_indexes: - print(f" ✓ {idx.get('name')}") - - print(f"\nVector indexes: {len(vector_indexes)}") - for idx in vector_indexes: - print(f" ✓ {idx.get('name')}") - - print(f"\nRange indexes (including config_id): {len(range_indexes)}") - for idx in range_indexes: - print(f" ✓ {idx.get('name')}") - - if not vector_indexes: - print("\n⚠️ WARNING: No vector indexes found!") - print(" Embedding search will be VERY SLOW (~1.4s)") - print(" Run: python create_indexes.py") - - # Check for config_id indexes - config_id_indexes = [idx for idx in range_indexes if 'config_id' in idx.get('name', '')] - if len(config_id_indexes) < 4: - print("\n⚠️ WARNING: Not all config_id indexes found!") - print(f" Expected 4, found {len(config_id_indexes)}") - print(" Run: python create_indexes.py config_id") - - print("=" * 70) - - finally: - await connector.close() - - -if __name__ == "__main__": - import asyncio - import sys - - if len(sys.argv) > 1: - command = sys.argv[1] - if command == "check": - asyncio.run(check_indexes()) - elif command == "fulltext": - asyncio.run(create_fulltext_indexes()) - elif command == "vector": - asyncio.run(create_vector_indexes()) - elif command == "config_id": - asyncio.run(create_config_id_indexes()) - elif command == "constraints": - asyncio.run(create_unique_constraints()) - else: - print(f"Unknown command: {command}") - print("\nUsage:") - print(" python create_indexes.py # Create all indexes") - print(" python create_indexes.py check # Check existing indexes") - print(" python create_indexes.py fulltext # Create only full-text indexes") - print(" python create_indexes.py vector # Create only vector indexes") - print(" python create_indexes.py config_id # Create only config_id indexes") - print(" python create_indexes.py constraints # Create only constraints") - else: - asyncio.run(create_all_indexes()) - diff --git a/api/app/repositories/neo4j/cypher_queries.py b/api/app/repositories/neo4j/cypher_queries.py index 66d24fab..26ffe350 100644 --- a/api/app/repositories/neo4j/cypher_queries.py +++ b/api/app/repositories/neo4j/cypher_queries.py @@ -336,6 +336,53 @@ ORDER BY score DESC LIMIT $limit """ +SEARCH_ENTITIES_BY_NAME_OR_ALIAS = """ +CALL db.index.fulltext.queryNodes("entitiesFulltext", $q) YIELD node AS e, score +WHERE ($end_user_id IS NULL OR e.end_user_id = $end_user_id) +WITH e, score +WITH collect({entity: e, score: score}) AS fulltextResults + +OPTIONAL MATCH (ae:ExtractedEntity) +WHERE ($end_user_id IS NULL OR ae.end_user_id = $end_user_id) + AND ae.aliases IS NOT NULL + AND ANY(alias IN ae.aliases WHERE toLower(alias) CONTAINS toLower($q)) +WITH fulltextResults, collect(ae) AS aliasEntities + +UNWIND (fulltextResults + [x IN aliasEntities | {entity: x, score: + CASE + WHEN ANY(alias IN x.aliases WHERE toLower(alias) = toLower($q)) THEN 1.0 + WHEN ANY(alias IN x.aliases WHERE toLower(alias) STARTS WITH toLower($q)) THEN 0.9 + ELSE 0.8 + END +}]) AS row +WITH row.entity AS e, row.score AS score +WITH DISTINCT e, MAX(score) AS score +OPTIONAL MATCH (s:Statement)-[:REFERENCES_ENTITY]->(e) +OPTIONAL MATCH (c:Chunk)-[:CONTAINS]->(s) +RETURN e.id AS id, + e.name AS name, + e.end_user_id AS end_user_id, + e.entity_type AS entity_type, + e.created_at AS created_at, + e.expired_at AS expired_at, + e.entity_idx AS entity_idx, + e.statement_id AS statement_id, + e.description AS description, + e.aliases AS aliases, + e.name_embedding AS name_embedding, + e.connect_strength AS connect_strength, + collect(DISTINCT s.id) AS statement_ids, + collect(DISTINCT c.id) AS chunk_ids, + COALESCE(e.activation_value, e.importance_score, 0.5) AS activation_value, + COALESCE(e.importance_score, 0.5) AS importance_score, + e.last_access_time AS last_access_time, + COALESCE(e.access_count, 0) AS access_count, + score +ORDER BY score DESC +LIMIT $limit +""" + + SEARCH_CHUNKS_BY_CONTENT = """ CALL db.index.fulltext.queryNodes("chunksFulltext", $q) YIELD node AS c, score WHERE ($end_user_id IS NULL OR c.end_user_id = $end_user_id) @@ -709,7 +756,6 @@ SET r.end_user_id = e.end_user_id, RETURN elementId(r) AS uuid """ - # Entity Merge Query MERGE_ENTITIES = """ MATCH (canonical:ExtractedEntity {id: $canonical_id}) @@ -829,9 +875,8 @@ neo4j_query_all = """ other as entity2 """ - '''针对当前节点下扩长的句子,实体和总结''' -Memory_Timeline_ExtractedEntity=""" +Memory_Timeline_ExtractedEntity = """ MATCH (n)-[r1]-(e)-[r2]-(ms) WHERE elementId(n) = $id AND (ms:ExtractedEntity OR ms:MemorySummary) @@ -869,7 +914,7 @@ RETURN """ -Memory_Timeline_MemorySummary=""" +Memory_Timeline_MemorySummary = """ MATCH (n)-[r1]-(e)-[r2]-(ms) WHERE elementId(n) =$id AND (ms:MemorySummary OR ms:ExtractedEntity) @@ -904,7 +949,7 @@ RETURN } ) AS statement; """ -Memory_Timeline_Statement=""" +Memory_Timeline_Statement = """ MATCH (n) WHERE elementId(n) = $id @@ -947,7 +992,7 @@ RETURN """ '''针对当前节点,主要获取更加完整的句子节点''' -Memory_Space_Emotion_Statement=""" +Memory_Space_Emotion_Statement = """ MATCH (n) WHERE elementId(n) = $id RETURN @@ -957,7 +1002,7 @@ RETURN n.statement AS statement; """ -Memory_Space_Emotion_MemorySummary=""" +Memory_Space_Emotion_MemorySummary = """ MATCH (n)-[]-(e) WHERE elementId(n) = $id AND EXISTS { @@ -970,7 +1015,7 @@ RETURN DISTINCT e.emotion_type AS emotion_type, e.statement AS statement; """ -Memory_Space_Emotion_ExtractedEntity=""" +Memory_Space_Emotion_ExtractedEntity = """ MATCH (n)-[]-(e) WHERE elementId(n) = $id AND EXISTS { @@ -985,18 +1030,18 @@ RETURN DISTINCT '''获取实体''' -Memory_Space_User=""" +Memory_Space_User = """ MATCH (n)-[r]->(m) WHERE n.end_user_id = $end_user_id AND m.name="用户" return DISTINCT elementId(m) as id """ -Memory_Space_Entity=""" +Memory_Space_Entity = """ MATCH (n)-[]-(m) WHERE elementId(m) = $id AND m.entity_type = "Person" RETURN DISTINCT m.name as name,m.end_user_id as end_user_id """ -Memory_Space_Associative=""" +Memory_Space_Associative = """ MATCH (u)-[]-(x)-[]-(h) WHERE elementId(u) = $user_id AND elementId(h) = $id @@ -1005,61 +1050,69 @@ RETURN DISTINCT """ Graph_Node_query = """ - MATCH (n:MemorySummary) - WHERE n.end_user_id = $end_user_id - RETURN - elementId(n) AS id, - labels(n) AS labels, - properties(n) AS properties, - 0 AS priority - LIMIT $limit +MATCH (n:MemorySummary) +WHERE n.end_user_id = $end_user_id +RETURN + elementId(n) AS id, + labels(n) AS labels, + properties(n) AS properties, + 0 AS priority +LIMIT $limit - UNION ALL +UNION ALL - MATCH (n:Dialogue) - WHERE n.end_user_id = $end_user_id - RETURN - elementId(n) AS id, - labels(n) AS labels, - properties(n) AS properties, - 1 AS priority - LIMIT 1 +MATCH (n:Dialogue) +WHERE n.end_user_id = $end_user_id +RETURN + elementId(n) AS id, + labels(n) AS labels, + properties(n) AS properties, + 1 AS priority +LIMIT 1 - UNION ALL +UNION ALL - MATCH (n:Statement) - WHERE n.end_user_id = $end_user_id - RETURN - elementId(n) AS id, - labels(n) AS labels, - properties(n) AS properties, - 1 AS priority - LIMIT $limit +MATCH (n:Statement) +WHERE n.end_user_id = $end_user_id +RETURN + elementId(n) AS id, + labels(n) AS labels, + properties(n) AS properties, + 1 AS priority +LIMIT $limit - UNION ALL +UNION ALL - MATCH (n:ExtractedEntity) - WHERE n.end_user_id = $end_user_id - RETURN - elementId(n) AS id, - labels(n) AS labels, - properties(n) AS properties, - 2 AS priority - LIMIT $limit +MATCH (n:ExtractedEntity) +WHERE n.end_user_id = $end_user_id +RETURN + elementId(n) AS id, + labels(n) AS labels, + properties(n) AS properties, + 2 AS priority +LIMIT $limit - UNION ALL +UNION ALL - MATCH (n:Chunk) - WHERE n.end_user_id = $end_user_id - RETURN - elementId(n) AS id, - labels(n) AS labels, - properties(n) AS properties, - 3 AS priority - LIMIT $limit +MATCH (n:Chunk) +WHERE n.end_user_id = $end_user_id +RETURN + elementId(n) AS id, + labels(n) AS labels, + properties(n) AS properties, + 3 AS priority +LIMIT $limit - """ +UNION ALL +MATCH (n:Perceptual) +WHERE n.end_user_id = $end_user_id +RETURN + elementId(n) AS id, + labels(n) AS labels, + properties(n) AS properties, + 4 AS priority +""" # ============================================================ # Community 节点 & BELONGS_TO_COMMUNITY 边 @@ -1069,6 +1122,7 @@ Graph_Node_query = """ COMMUNITY_NODE_UPSERT = """ MERGE (c:Community {community_id: $community_id}) +ON CREATE SET c.id = $community_id SET c.end_user_id = $end_user_id, c.member_count = $member_count, c.updated_at = datetime() @@ -1122,21 +1176,43 @@ RETURN e.id AS id, CASE WHEN c IS NOT NULL THEN c.community_id ELSE null END AS community_id """ +GET_ENTITY_COUNT_FOR_USER = """ +MATCH (e:ExtractedEntity {end_user_id: $end_user_id}) +RETURN count(e) AS entity_count +""" + +GET_ALL_ENTITY_IDS_FOR_USER = """ +MATCH (e:ExtractedEntity {end_user_id: $end_user_id}) +RETURN e.id AS id +""" + GET_COMMUNITY_MEMBERS = """ MATCH (e:ExtractedEntity {end_user_id: $end_user_id})-[:BELONGS_TO_COMMUNITY]->(c:Community {community_id: $community_id}) RETURN e.id AS id, e.name AS name, e.entity_type AS entity_type, e.importance_score AS importance_score, e.activation_value AS activation_value, - e.name_embedding AS name_embedding + e.name_embedding AS name_embedding, + e.aliases AS aliases, e.description AS description, + e.example AS example ORDER BY coalesce(e.activation_value, 0) DESC """ +GET_COMMUNITY_RELATIONSHIPS = """ +MATCH (e1:ExtractedEntity {end_user_id: $end_user_id})-[:BELONGS_TO_COMMUNITY]->(c:Community {community_id: $community_id}) +MATCH (e2:ExtractedEntity {end_user_id: $end_user_id})-[:BELONGS_TO_COMMUNITY]->(c) +MATCH (e1)-[r:EXTRACTED_RELATIONSHIP]->(e2) +RETURN e1.name AS subject, r.predicate AS predicate, e2.name AS object +ORDER BY e1.name, r.predicate, e2.name +LIMIT 20 +""" + GET_ALL_COMMUNITY_MEMBERS_BATCH = """ MATCH (e:ExtractedEntity {end_user_id: $end_user_id})-[:BELONGS_TO_COMMUNITY]->(c:Community) -WHERE c.community_id IN $community_ids RETURN c.community_id AS community_id, - e.id AS id, + e.id AS id, e.name AS name, e.entity_type AS entity_type, + e.importance_score AS importance_score, e.activation_value AS activation_value, e.name_embedding AS name_embedding, - e.activation_value AS activation_value + e.aliases AS aliases, e.description AS description +ORDER BY c.community_id, coalesce(e.activation_value, 0) DESC """ CHECK_USER_HAS_COMMUNITIES = """ @@ -1153,7 +1229,8 @@ RETURN c.community_id AS community_id, cnt AS member_count UPDATE_COMMUNITY_METADATA = """ MATCH (c:Community {community_id: $community_id, end_user_id: $end_user_id}) -SET c.name = $name, +SET c.id = coalesce(c.id, $community_id), + c.name = $name, c.summary = $summary, c.core_entities = $core_entities, c.summary_embedding = $summary_embedding, @@ -1161,6 +1238,51 @@ SET c.name = $name, RETURN c.community_id AS community_id """ +BATCH_UPDATE_COMMUNITY_METADATA = """ +UNWIND $communities AS row +MATCH (c:Community {community_id: row.community_id, end_user_id: row.end_user_id}) +SET c.id = coalesce(c.id, row.community_id), + c.name = row.name, + c.summary = row.summary, + c.core_entities = row.core_entities, + c.summary_embedding = row.summary_embedding, + c.updated_at = datetime() +RETURN c.community_id AS community_id +""" + +GET_ENTITIES_PAGE = """ +MATCH (e:ExtractedEntity {end_user_id: $end_user_id}) +OPTIONAL MATCH (e)-[:BELONGS_TO_COMMUNITY]->(c:Community) +RETURN e.id AS id, + e.name AS name, + e.name_embedding AS name_embedding, + e.activation_value AS activation_value, + CASE WHEN c IS NOT NULL THEN c.community_id ELSE null END AS community_id +ORDER BY e.id +SKIP $skip LIMIT $limit +""" + +GET_ENTITY_NEIGHBORS_BATCH_FOR_IDS = """ +// 批量拉取指定实体列表的邻居(用于分批全量聚类) +MATCH (e:ExtractedEntity {end_user_id: $end_user_id}) +WHERE e.id IN $entity_ids +OPTIONAL MATCH (e)-[:EXTRACTED_RELATIONSHIP]-(nb1:ExtractedEntity {end_user_id: $end_user_id}) +OPTIONAL MATCH (s:Statement)-[:REFERENCES_ENTITY]->(e) +OPTIONAL MATCH (s)-[:REFERENCES_ENTITY]->(nb2:ExtractedEntity {end_user_id: $end_user_id}) +WHERE nb2.id <> e.id +WITH e, collect(DISTINCT nb1) + collect(DISTINCT nb2) AS all_neighbors +UNWIND all_neighbors AS nb +WITH e, nb WHERE nb IS NOT NULL +OPTIONAL MATCH (nb)-[:BELONGS_TO_COMMUNITY]->(c:Community) +RETURN DISTINCT + e.id AS entity_id, + nb.id AS id, + nb.name AS name, + nb.name_embedding AS name_embedding, + nb.activation_value AS activation_value, + CASE WHEN c IS NOT NULL THEN c.community_id ELSE null END AS community_id +""" + GET_ALL_ENTITY_NEIGHBORS_BATCH = """ // 批量拉取某用户下所有实体的邻居(用于全量聚类预加载) MATCH (e:ExtractedEntity {end_user_id: $end_user_id}) @@ -1238,3 +1360,92 @@ WHERE c.name IS NULL OR c.name = '' OR (c.summary_embedding IS NULL AND c.summary IS NOT NULL AND c.summary <> '(empty)') RETURN c.community_id AS community_id """ + +# Community keyword search: matches name or summary via fulltext index +SEARCH_COMMUNITIES_BY_KEYWORD = """ +CALL db.index.fulltext.queryNodes("communitiesFulltext", $q) YIELD node AS c, score +WHERE ($end_user_id IS NULL OR c.end_user_id = $end_user_id) +RETURN c.community_id AS id, + c.name AS name, + c.summary AS content, + c.core_entities AS core_entities, + c.member_count AS member_count, + c.end_user_id AS end_user_id, + c.updated_at AS updated_at, + score +ORDER BY score DESC +LIMIT $limit +""" + +# Community 向量检索 ────────────────────────────────────────────────── +# Community embedding-based search: cosine similarity on Community.summary_embedding +COMMUNITY_EMBEDDING_SEARCH = """ +CALL db.index.vector.queryNodes('community_summary_embedding_index', $limit * 100, $embedding) +YIELD node AS c, score +WHERE c.summary_embedding IS NOT NULL + AND ($end_user_id IS NULL OR c.end_user_id = $end_user_id) +RETURN c.community_id AS id, + c.name AS name, + c.summary AS content, + c.core_entities AS core_entities, + c.member_count AS member_count, + c.end_user_id AS end_user_id, + c.updated_at AS updated_at, + score +ORDER BY score DESC +LIMIT $limit +""" + +# Community 展开检索 ────────────────────────────────────────────────── +# 命中社区后,拉取该社区所有成员实体关联的 Statement 节点(主题→细节两级检索) +EXPAND_COMMUNITY_STATEMENTS = """ +MATCH (c:Community {community_id: $community_id}) +MATCH (e:ExtractedEntity)-[:BELONGS_TO_COMMUNITY]->(c) +MATCH (s:Statement)-[:REFERENCES_ENTITY]->(e) +WHERE s.end_user_id = $end_user_id +RETURN s.statement AS statement, + s.id AS id, + s.end_user_id AS end_user_id, + s.created_at AS created_at, + s.valid_at AS valid_at, + s.invalid_at AS invalid_at, + COALESCE(s.activation_value, s.importance_score, 0.5) AS activation_value, + COALESCE(s.importance_score, 0.5) AS importance_score, + e.name AS source_entity, + c.name AS community_name +ORDER BY COALESCE(s.activation_value, 0) DESC +LIMIT $limit +""" + +# 感知记忆节点保存 +PERCEPTUAL_NODE_SAVE = """ +UNWIND $perceptuals AS p +MERGE (n:Perceptual {id: p.id}) +SET n += { + id: p.id, + end_user_id: p.end_user_id, + perceptual_type: p.perceptual_type, + file_path: p.file_path, + file_name: p.file_name, + file_ext: p.file_ext, + summary: p.summary, + keywords: p.keywords, + topic: p.topic, + domain: p.domain, + created_at: p.created_at, + file_type: p.file_type, + summary_embedding: p.summary_embedding +} +RETURN n.id AS uuid +""" + +# 感知记忆与对话的关联边 +PERCEPTUAL_CHUNK_EDGE_SAVE = """ +UNWIND $edges AS edge +MATCH (p:Perceptual {id: edge.perceptual_id, end_user_id: edge.end_user_id}) +MATCH (c:Chunk {id: edge.chunk_id, end_user_id: edge.end_user_id}) +MERGE (c)-[r:HAS_PERCEPTUAL]->(p) +ON CREATE SET r.end_user_id = edge.end_user_id, + r.created_at = edge.created_at +RETURN elementId(r) AS uuid +""" diff --git a/api/app/repositories/neo4j/graph_saver.py b/api/app/repositories/neo4j/graph_saver.py index cbd2b532..adc266fe 100644 --- a/api/app/repositories/neo4j/graph_saver.py +++ b/api/app/repositories/neo4j/graph_saver.py @@ -22,13 +22,18 @@ from app.core.memory.models.graph_models import ( StatementNode, ExtractedEntityNode, EntityEntityEdge, + PerceptualNode, + PerceptualEdge, ) import logging + logger = logging.getLogger(__name__) + + async def save_entities_and_relationships( - entity_nodes: List[ExtractedEntityNode], - entity_entity_edges: List[EntityEntityEdge], - connector: Neo4jConnector + entity_nodes: List[ExtractedEntityNode], + entity_entity_edges: List[EntityEntityEdge], + connector: Neo4jConnector ): """Save entities and their relationships using graph models""" all_entities = [entity.model_dump() for entity in entity_nodes] @@ -73,8 +78,8 @@ async def save_entities_and_relationships( async def save_chunk_nodes( - chunk_nodes: List[ChunkNode], - connector: Neo4jConnector + chunk_nodes: List[ChunkNode], + connector: Neo4jConnector ): """Save chunk nodes using graph models""" if not chunk_nodes: @@ -89,8 +94,8 @@ async def save_chunk_nodes( async def save_statement_chunk_edges( - statement_chunk_edges: List[StatementChunkEdge], - connector: Neo4jConnector + statement_chunk_edges: List[StatementChunkEdge], + connector: Neo4jConnector ): """Save statement-chunk edges using graph models""" if not statement_chunk_edges: @@ -118,8 +123,8 @@ async def save_statement_chunk_edges( async def save_statement_entity_edges( - statement_entity_edges: List[StatementEntityEdge], - connector: Neo4jConnector + statement_entity_edges: List[StatementEntityEdge], + connector: Neo4jConnector ): """Save statement-entity edges using graph models""" if not statement_entity_edges: @@ -142,7 +147,7 @@ async def save_statement_entity_edges( if all_se_edges: try: await connector.execute_query( - STATEMENT_ENTITY_EDGE_SAVE, + STATEMENT_ENTITY_EDGE_SAVE, relationships=all_se_edges ) except Exception: @@ -154,23 +159,28 @@ async def save_dialog_and_statements_to_neo4j( chunk_nodes: List[ChunkNode], statement_nodes: List[StatementNode], entity_nodes: List[ExtractedEntityNode], + perceptual_nodes: List[PerceptualNode], entity_edges: List[EntityEntityEdge], statement_chunk_edges: List[StatementChunkEdge], statement_entity_edges: List[StatementEntityEdge], + perceptual_edges: List[PerceptualEdge], connector: Neo4jConnector, - config_id: Optional[str] = None, - llm_model_id: Optional[str] = None, ) -> bool: """Save dialogue nodes, chunk nodes, statement nodes, entities, and all relationships to Neo4j using graph models. + 只负责数据写入,不触发聚类。聚类由调用方在写入成功后通过 + _trigger_clustering_sync() 显式触发。 + Args: dialogue_nodes: List of DialogueNode objects to save chunk_nodes: List of ChunkNode objects to save statement_nodes: List of StatementNode objects to save entity_nodes: List of ExtractedEntityNode objects to save + perceptual_nodes: List of PerceptualNode objects to save entity_edges: List of EntityEntityEdge objects to save statement_chunk_edges: List of StatementChunkEdge objects to save statement_entity_edges: List of StatementEntityEdge objects to save + perceptual_edges: List of PerceptualEdge objects to save connector: Neo4j connector instance Returns: @@ -189,7 +199,7 @@ async def save_dialog_and_statements_to_neo4j( result = await tx.run(DIALOGUE_NODE_SAVE, dialogues=dialogue_data) dialogue_uuids = [record["uuid"] async for record in result] results['dialogues'] = dialogue_uuids - print(f"Dialogues saved to Neo4j with UUIDs: {dialogue_uuids}") + logger.info(f"Dialogues saved to Neo4j with UUIDs: {dialogue_uuids}") # 2. Save all chunk nodes in batch if chunk_nodes: @@ -200,6 +210,14 @@ async def save_dialog_and_statements_to_neo4j( results['chunks'] = chunk_uuids logger.info(f"Successfully saved {len(chunk_uuids)} chunk nodes to Neo4j") + if perceptual_nodes: + from app.repositories.neo4j.cypher_queries import PERCEPTUAL_NODE_SAVE + perceptual_data = [node.model_dump() for node in perceptual_nodes] + result = await tx.run(PERCEPTUAL_NODE_SAVE, perceptuals=perceptual_data) + perceptual_uuids = [record["uuid"] async for record in result] + results["perceptuals"] = perceptual_uuids + logger.info(f"Successfully saved {len(perceptual_uuids)} perceptual nodes to Neo4j") + # 3. Save all statement nodes in batch if statement_nodes: from app.repositories.neo4j.cypher_queries import STATEMENT_NODE_SAVE @@ -280,6 +298,22 @@ async def save_dialog_and_statements_to_neo4j( results['statement_entity_edges'] = se_uuids logger.info(f"Successfully saved {len(se_uuids)} statement-entity edges to Neo4j") + if perceptual_edges: + from app.repositories.neo4j.cypher_queries import PERCEPTUAL_CHUNK_EDGE_SAVE + perceptual_edge_data = [] + for edge in perceptual_edges: + print(edge.source, edge.target) + perceptual_edge_data.append({ + "perceptual_id": edge.source, + "chunk_id": edge.target, + "end_user_id": edge.end_user_id, + "created_at": edge.created_at.isoformat() if edge.created_at else None, + }) + result = await tx.run(PERCEPTUAL_CHUNK_EDGE_SAVE, edges=perceptual_edge_data) + perceptual_edges_uuids = [record["uuid"] async for record in result] + results['perceptual_chunk_edges'] = perceptual_edges_uuids + logger.info(f"Successfully saved {len(perceptual_edges_uuids)} perceptual-chunk edges to Neo4j") + return results try: @@ -293,9 +327,6 @@ async def save_dialog_and_statements_to_neo4j( logger.info("Transaction completed. Summary: %s", summary) logger.debug("Full transaction results: %r", results) - # 写入成功后,异步触发聚类(不阻塞写入响应) - schedule_clustering_after_write(entity_nodes, config_id=config_id, llm_model_id=llm_model_id) - return True except Exception as e: @@ -305,16 +336,13 @@ async def save_dialog_and_statements_to_neo4j( return False -def schedule_clustering_after_write( - entity_nodes: List, - config_id: Optional[str] = None, - llm_model_id: Optional[str] = None, +async def _trigger_clustering_sync( + entity_nodes: List, + llm_model_id: Optional[str] = None, + embedding_model_id: Optional[str] = None, ) -> None: """ - 写入 Neo4j 成功后,调度后台聚类任务。 - - 可通过环境变量 CLUSTERING_ENABLED=false 禁用(用于基准测试对比)。 - 使用 asyncio.create_task 异步触发,不阻塞写入响应。 + 同步等待聚类完成,避免与其他 LLM 任务并发冲突。 """ if not entity_nodes: return @@ -326,15 +354,16 @@ def schedule_clustering_after_write( end_user_id = entity_nodes[0].end_user_id new_entity_ids = [e.id for e in entity_nodes] - logger.info(f"[Clustering] 准备触发聚类,实体数: {len(new_entity_ids)}, end_user_id: {end_user_id}") - asyncio.create_task(_trigger_clustering(new_entity_ids, end_user_id, config_id=config_id, llm_model_id=llm_model_id)) + logger.info(f"[Clustering] 准备触发聚类(同步),实体数: {len(new_entity_ids)}, end_user_id: {end_user_id}") + await _trigger_clustering(new_entity_ids, end_user_id, llm_model_id=llm_model_id, + embedding_model_id=embedding_model_id) async def _trigger_clustering( - new_entity_ids: List[str], - end_user_id: str, - config_id: Optional[str] = None, - llm_model_id: Optional[str] = None, + new_entity_ids: List[str], + end_user_id: str, + llm_model_id: Optional[str] = None, + embedding_model_id: Optional[str] = None, ) -> None: """ 聚类触发函数,自动判断全量初始化还是增量更新。 @@ -344,7 +373,7 @@ async def _trigger_clustering( from app.core.memory.storage_services.clustering_engine import LabelPropagationEngine logger.info(f"[Clustering] 开始聚类,end_user_id={end_user_id}, 实体数={len(new_entity_ids)}") connector = Neo4jConnector() - engine = LabelPropagationEngine(connector, config_id=config_id, llm_model_id=llm_model_id) + engine = LabelPropagationEngine(connector, llm_model_id=llm_model_id, embedding_model_id=embedding_model_id) await engine.run(end_user_id=end_user_id, new_entity_ids=new_entity_ids) logger.info(f"[Clustering] 聚类完成,end_user_id={end_user_id}") except Exception as e: diff --git a/api/app/repositories/neo4j/graph_search.py b/api/app/repositories/neo4j/graph_search.py index e8f52535..c5d3bcca 100644 --- a/api/app/repositories/neo4j/graph_search.py +++ b/api/app/repositories/neo4j/graph_search.py @@ -4,12 +4,16 @@ from typing import Any, Dict, List, Optional from app.repositories.neo4j.cypher_queries import ( CHUNK_EMBEDDING_SEARCH, + COMMUNITY_EMBEDDING_SEARCH, ENTITY_EMBEDDING_SEARCH, + EXPAND_COMMUNITY_STATEMENTS, MEMORY_SUMMARY_EMBEDDING_SEARCH, SEARCH_CHUNK_BY_CHUNK_ID, SEARCH_CHUNKS_BY_CONTENT, + SEARCH_COMMUNITIES_BY_KEYWORD, SEARCH_DIALOGUE_BY_DIALOG_ID, SEARCH_ENTITIES_BY_NAME, + SEARCH_ENTITIES_BY_NAME_OR_ALIAS, SEARCH_MEMORY_SUMMARIES_BY_KEYWORD, SEARCH_STATEMENTS_BY_CREATED_AT, SEARCH_STATEMENTS_BY_KEYWORD, @@ -261,7 +265,7 @@ async def search_graph( if "entities" in include: tasks.append(connector.execute_query( - SEARCH_ENTITIES_BY_NAME, + SEARCH_ENTITIES_BY_NAME_OR_ALIAS, q=q, end_user_id=end_user_id, limit=limit, @@ -285,6 +289,15 @@ async def search_graph( limit=limit, )) task_keys.append("summaries") + + if "communities" in include: + tasks.append(connector.execute_query( + SEARCH_COMMUNITIES_BY_KEYWORD, + q=q, + end_user_id=end_user_id, + limit=limit, + )) + task_keys.append("communities") # Execute all queries in parallel task_results = await asyncio.gather(*tasks, return_exceptions=True) @@ -293,6 +306,7 @@ async def search_graph( results = {} for key, result in zip(task_keys, task_results): if isinstance(result, Exception): + logger.warning(f"search_graph: {key} 关键词查询异常: {result}") results[key] = [] else: results[key] = result @@ -349,7 +363,11 @@ async def search_graph_by_embedding( print(f"[PERF] Embedding generation took: {embed_time:.4f}s") if not embeddings or not embeddings[0]: - return {"statements": [], "chunks": [], "entities": [], "summaries": []} + logger.warning( + f"search_graph_by_embedding: embedding 生成失败或为空," + f"query='{query_text[:50]}', end_user_id={end_user_id},向量检索跳过" + ) + return {"statements": [], "chunks": [], "entities": [], "summaries": [], "communities": []} embedding = embeddings[0] # Prepare tasks for parallel execution @@ -396,6 +414,16 @@ async def search_graph_by_embedding( )) task_keys.append("summaries") + # Communities (向量语义匹配) + if "communities" in include: + tasks.append(connector.execute_query( + COMMUNITY_EMBEDDING_SEARCH, + embedding=embedding, + end_user_id=end_user_id, + limit=limit, + )) + task_keys.append("communities") + # Execute all queries in parallel query_start = time.time() task_results = await asyncio.gather(*tasks, return_exceptions=True) @@ -408,10 +436,12 @@ async def search_graph_by_embedding( "chunks": [], "entities": [], "summaries": [], + "communities": [], } for key, result in zip(task_keys, task_results): if isinstance(result, Exception): + logger.warning(f"search_graph_by_embedding: {key} 向量查询异常: {result}") results[key] = [] else: results[key] = result @@ -661,6 +691,62 @@ async def search_graph_by_chunk_id( return {"chunks": chunks} +async def search_graph_community_expand( + connector: Neo4jConnector, + community_ids: List[str], + end_user_id: str, + limit: int = 10, +) -> Dict[str, List[Dict[str, Any]]]: + """ + 三期:社区展开检索 —— 主题 → 细节两级检索。 + + 命中 Community 节点后,沿 BELONGS_TO_COMMUNITY 关系拉取成员实体, + 再沿 REFERENCES_ENTITY 关系拉取关联的 Statement 节点, + 按 activation_value 降序返回,实现"主题摘要 → 具体记忆"的深度召回。 + + Args: + connector: Neo4j 连接器 + community_ids: 已命中的社区 ID 列表 + end_user_id: 用户 ID,用于数据隔离 + limit: 每个社区最多返回的 Statement 数量 + + Returns: + {"expanded_statements": [Statement 列表,含 community_name / source_entity 字段]} + """ + if not community_ids or not end_user_id: + return {"expanded_statements": []} + + tasks = [ + connector.execute_query( + EXPAND_COMMUNITY_STATEMENTS, + community_id=cid, + end_user_id=end_user_id, + limit=limit, + ) + for cid in community_ids + ] + + task_results = await asyncio.gather(*tasks, return_exceptions=True) + + expanded: List[Dict[str, Any]] = [] + for cid, result in zip(community_ids, task_results): + if isinstance(result, Exception): + logger.warning(f"社区展开检索失败 community_id={cid}: {result}") + else: + expanded.extend(result) + + # 按 activation_value 全局排序后去重 + from app.core.memory.src.search import _deduplicate_results + expanded.sort( + key=lambda x: float(x.get("activation_value") or 0), + reverse=True, + ) + expanded = _deduplicate_results(expanded) + + logger.info(f"社区展开检索完成: community_ids={community_ids}, 展开 statements={len(expanded)}") + return {"expanded_statements": expanded} + + async def search_graph_by_created_at( connector: Neo4jConnector, end_user_id: Optional[str] = None, diff --git a/api/app/repositories/user_repository.py b/api/app/repositories/user_repository.py index b4c11aa4..af4449e5 100644 --- a/api/app/repositories/user_repository.py +++ b/api/app/repositories/user_repository.py @@ -19,18 +19,22 @@ class UserRepository: self.db = db def get_user_by_id(self, user_id: uuid.UUID) -> Optional[User]: - """根据ID获取用户""" - db_logger.debug(f"根据ID查询用户: user_id={user_id}") + """根据 ID 获取用户(租户禁用时返回 None)""" + db_logger.debug(f"根据 ID 查询用户:user_id={user_id}") try: user = self.db.query(User).options(joinedload(User.tenant)).filter(User.id == user_id).first() if user: - db_logger.debug(f"用户查询成功: {user.username} (ID: {user_id})") + # 检查租户状态,租户禁用时返回 None + if user.tenant and not user.tenant.is_active: + db_logger.warning(f"用户 {user.username} (ID: {user_id}) 所属租户 {user.tenant_id} 已被禁用") + return None + db_logger.debug(f"用户查询成功:{user.username} (ID: {user_id})") else: - db_logger.debug(f"用户不存在: user_id={user_id}") + db_logger.debug(f"用户不存在:user_id={user_id}") return user except Exception as e: - db_logger.error(f"根据ID查询用户失败: user_id={user_id} - {str(e)}") + db_logger.error(f"根据 ID 查询用户失败:user_id={user_id} - {str(e)}") raise def get_user_by_email(self, email: str) -> Optional[User]: @@ -154,22 +158,26 @@ class UserRepository: raise def get_users_by_tenant( - self, - tenant_id: uuid.UUID, - skip: int = 0, + self, + tenant_id: uuid.UUID, + skip: int = 0, limit: int = 100, is_active: Optional[bool] = None, + is_superuser: Optional[bool] = None, search: Optional[str] = None ) -> List[User]: """获取租户下的用户列表""" db_logger.debug(f"查询租户用户: tenant_id={tenant_id}") - + try: query = self.db.query(User).options(joinedload(User.tenant)).filter(User.tenant_id == tenant_id) - + if is_active is not None: query = query.filter(User.is_active == is_active) - + + if is_superuser is not None: + query = query.filter(User.is_superuser == is_superuser) + if search: query = query.filter( or_( @@ -177,7 +185,7 @@ class UserRepository: User.email.ilike(f"%{search}%") ) ) - + users = query.offset(skip).limit(limit).all() db_logger.debug(f"租户用户查询成功: tenant_id={tenant_id}, count={len(users)}") return users @@ -186,18 +194,22 @@ class UserRepository: raise def count_users_by_tenant( - self, + self, tenant_id: uuid.UUID, is_active: Optional[bool] = None, + is_superuser: Optional[bool] = None, search: Optional[str] = None ) -> int: """统计租户下的用户数量""" try: query = self.db.query(func.count(User.id)).filter(User.tenant_id == tenant_id) - + if is_active is not None: query = query.filter(User.is_active == is_active) - + + if is_superuser is not None: + query = query.filter(User.is_superuser == is_superuser) + if search: query = query.filter( or_( @@ -205,7 +217,7 @@ class UserRepository: User.email.ilike(f"%{search}%") ) ) - + return query.scalar() except Exception as e: db_logger.error(f"统计租户用户失败: tenant_id={tenant_id} - {str(e)}") diff --git a/api/app/schemas/app_log_schema.py b/api/app/schemas/app_log_schema.py new file mode 100644 index 00000000..bda78138 --- /dev/null +++ b/api/app/schemas/app_log_schema.py @@ -0,0 +1,53 @@ +"""应用日志(消息记录)Schema""" +import uuid +import datetime +from typing import Optional, Dict, Any, List + +from pydantic import BaseModel, Field, ConfigDict, field_serializer + + +class AppLogMessage(BaseModel): + """单条消息记录""" + model_config = ConfigDict(from_attributes=True) + + id: uuid.UUID + conversation_id: uuid.UUID + role: str = Field(description="角色: user / assistant / system") + content: str + meta_data: Optional[Dict[str, Any]] = None + created_at: datetime.datetime + + @field_serializer("created_at", when_used="json") + def _serialize_created_at(self, dt: datetime.datetime): + return int(dt.timestamp() * 1000) if dt else None + + @field_serializer("meta_data", when_used="json") + def _serialize_meta_data(self, data: Optional[Dict[str, Any]]): + return data or {} + + +class AppLogConversation(BaseModel): + """会话摘要(用于列表)""" + model_config = ConfigDict(from_attributes=True) + + id: uuid.UUID + app_id: uuid.UUID + user_id: Optional[str] = None + title: Optional[str] = None + message_count: int = 0 + is_draft: bool + created_at: datetime.datetime + updated_at: datetime.datetime + + @field_serializer("created_at", when_used="json") + def _serialize_created_at(self, dt: datetime.datetime): + return int(dt.timestamp() * 1000) if dt else None + + @field_serializer("updated_at", when_used="json") + def _serialize_updated_at(self, dt: datetime.datetime): + return int(dt.timestamp() * 1000) if dt else None + + +class AppLogConversationDetail(AppLogConversation): + """会话详情(包含消息列表)""" + messages: List[AppLogMessage] = Field(default_factory=list) diff --git a/api/app/schemas/app_schema.py b/api/app/schemas/app_schema.py index 1582d862..f1e9132f 100644 --- a/api/app/schemas/app_schema.py +++ b/api/app/schemas/app_schema.py @@ -196,6 +196,13 @@ class CitationConfig(BaseModel): enabled: bool = Field(default=False) +class Citation(BaseModel): + document_id: str + file_name: str + knowledge_id: str + score: float + + class WebSearchConfig(BaseModel): """联网搜索配置""" enabled: bool = Field(default=False) @@ -269,7 +276,7 @@ class AgentConfigCreate(BaseModel): # 记忆配置 memory: MemoryConfig = Field( - default_factory=lambda: MemoryConfig(enabled=True), + default_factory=lambda: MemoryConfig(enabled=False), description="对话历史记忆配置" ) diff --git a/api/app/schemas/end_user_info_schema.py b/api/app/schemas/end_user_info_schema.py new file mode 100644 index 00000000..60bdf9d6 --- /dev/null +++ b/api/app/schemas/end_user_info_schema.py @@ -0,0 +1,35 @@ +import uuid +import datetime +from typing import Optional, Dict, Any, List +from pydantic import BaseModel, Field +from pydantic import ConfigDict + + +class EndUserInfoBase(BaseModel): + """终端用户信息基础模型""" + other_name: str = Field(description="关联的用户名称") + aliases: Optional[List[str]] = Field(description="用户别名列表", default=None) + meta_data: Optional[Dict[str, Any]] = Field(description="用户相关的扩展信息", default=None) + + +class EndUserInfoCreate(EndUserInfoBase): + """创建终端用户信息请求模型""" + end_user_id: str = Field(description="关联的终端用户ID") + + +class EndUserInfoUpdate(BaseModel): + """更新终端用户信息请求模型""" + end_user_id: str = Field(description="终端用户ID") + other_name: Optional[str] = Field(description="用户名称", default=None) + aliases: Optional[List[str]] = Field(description="用户别名列表", default=None) + meta_data: Optional[Dict[str, Any]] = Field(description="用户相关的扩展信息", default=None) + + +class EndUserInfoResponse(EndUserInfoBase): + """终端用户信息响应模型""" + model_config = ConfigDict(from_attributes=True) + + end_user_info_id: uuid.UUID = Field(description="终端用户信息记录ID") + end_user_id: uuid.UUID = Field(description="关联的终端用户ID") + created_at: datetime.datetime = Field(description="创建时间") + updated_at: datetime.datetime = Field(description="更新时间") diff --git a/api/app/schemas/end_user_schema.py b/api/app/schemas/end_user_schema.py index bbb6fd5c..c2498203 100644 --- a/api/app/schemas/end_user_schema.py +++ b/api/app/schemas/end_user_schema.py @@ -1,6 +1,6 @@ import uuid import datetime -from typing import Optional +from typing import Optional, List from pydantic import BaseModel, Field from pydantic import ConfigDict @@ -17,40 +17,6 @@ class EndUser(BaseModel): created_at: datetime.datetime = Field(description="创建时间", default_factory=datetime.datetime.now) updated_at: datetime.datetime = Field(description="更新时间", default_factory=datetime.datetime.now) - # 用户基本信息字段 - position: Optional[str] = Field(description="职位", default=None) - department: Optional[str] = Field(description="部门", default=None) - contact: Optional[str] = Field(description="联系方式", default=None) - phone: Optional[str] = Field(description="电话", default=None) - hire_date: Optional[datetime.datetime] = Field(description="入职日期", default=None) - updatetime_profile: Optional[datetime.datetime] = Field(description="核心档案信息最后更新时间", default=None) - # 用户摘要和洞察更新时间 user_summary_updated_at: Optional[datetime.datetime] = Field(description="用户摘要最后更新时间", default=None) - memory_insight_updated_at: Optional[datetime.datetime] = Field(description="洞察报告最后更新时间", default=None) - - -class EndUserProfileResponse(BaseModel): - """终端用户基本信息响应模型""" - model_config = ConfigDict(from_attributes=True) - - id: uuid.UUID = Field(description="终端用户ID") - other_name: Optional[str] = Field(description="其他名称", default="") - position: Optional[str] = Field(description="职位", default=None) - department: Optional[str] = Field(description="部门", default=None) - contact: Optional[str] = Field(description="联系方式", default=None) - phone: Optional[str] = Field(description="电话", default=None) - hire_date: Optional[datetime.datetime] = Field(description="入职日期", default=None) - updatetime_profile: Optional[datetime.datetime] = Field(description="核心档案信息最后更新时间", default=None) - - - -class EndUserProfileUpdate(BaseModel): - """终端用户基本信息更新请求模型""" - end_user_id: str = Field(description="终端用户ID") - other_name: Optional[str] = Field(description="其他名称", default="") - position: Optional[str] = Field(description="职位", default=None) - department: Optional[str] = Field(description="部门", default=None) - contact: Optional[str] = Field(description="联系方式", default=None) - phone: Optional[str] = Field(description="电话", default=None) - hire_date: Optional[int] = Field(description="入职日期(时间戳,毫秒)", default=None) \ No newline at end of file + memory_insight_updated_at: Optional[datetime.datetime] = Field(description="洞察报告最后更新时间", default=None) \ No newline at end of file diff --git a/api/app/schemas/memory_api_schema.py b/api/app/schemas/memory_api_schema.py index 98d257c1..84a34e8a 100644 --- a/api/app/schemas/memory_api_schema.py +++ b/api/app/schemas/memory_api_schema.py @@ -21,7 +21,7 @@ class MemoryWriteRequest(BaseModel): """ end_user_id: str = Field(..., description="End user ID (required)") message: str = Field(..., description="Message content to store") - config_id: Optional[str] = Field(None, description="Memory configuration ID") + config_id: str = Field(..., description="Memory configuration ID (required)") storage_type: str = Field("neo4j", description="Storage type: neo4j or rag") user_rag_memory_id: Optional[str] = Field(None, description="RAG memory ID") @@ -68,7 +68,7 @@ class MemoryReadRequest(BaseModel): "0", description="Search mode: 0=verify, 1=direct, 2=context" ) - config_id: Optional[str] = Field(None, description="Memory configuration ID") + config_id: str = Field(..., description="Memory configuration ID (required)") storage_type: str = Field("neo4j", description="Storage type: neo4j or rag") user_rag_memory_id: Optional[str] = Field(None, description="RAG memory ID") @@ -132,3 +132,79 @@ class MemoryReadResponse(BaseModel): description="Intermediate retrieval outputs" ) end_user_id: str = Field(..., description="End user ID") + + +class CreateEndUserRequest(BaseModel): + """Request schema for creating an end user. + + Attributes: + workspace_id: Workspace ID (required) + other_id: External user identifier (required) + other_name: Display name for the end user + """ + workspace_id: str = Field(..., description="Workspace ID (required)") + other_id: str = Field(..., description="External user identifier (required)") + other_name: Optional[str] = Field("", description="Display name") + + @field_validator("workspace_id") + @classmethod + def validate_workspace_id(cls, v: str) -> str: + """Validate that workspace_id is not empty.""" + if not v or not v.strip(): + raise ValueError("workspace_id is required and cannot be empty") + return v.strip() + + @field_validator("other_id") + @classmethod + def validate_other_id(cls, v: str) -> str: + """Validate that other_id is not empty.""" + if not v or not v.strip(): + raise ValueError("other_id is required and cannot be empty") + return v.strip() + + +class CreateEndUserResponse(BaseModel): + """Response schema for end user creation. + + Attributes: + id: Created end user UUID + other_id: External user identifier + other_name: Display name + workspace_id: Workspace the user belongs to + """ + id: str = Field(..., description="End user UUID") + other_id: str = Field(..., description="External user identifier") + other_name: str = Field("", description="Display name") + workspace_id: str = Field(..., description="Workspace ID") + + +class MemoryConfigItem(BaseModel): + """Schema for a single memory config in the list response. + + Attributes: + config_id: Configuration UUID + config_name: Configuration name + config_desc: Configuration description + is_default: Whether this is the workspace default config + scene_name: Associated ontology scene name + created_at: Creation timestamp + updated_at: Last update timestamp + """ + config_id: str = Field(..., description="Configuration ID") + config_name: str = Field(..., description="Configuration name") + config_desc: Optional[str] = Field(None, description="Configuration description") + is_default: bool = Field(False, description="Whether this is the workspace default") + scene_name: Optional[str] = Field(None, description="Associated ontology scene name") + created_at: Optional[str] = Field(None, description="Creation timestamp") + updated_at: Optional[str] = Field(None, description="Last update timestamp") + + +class ListConfigsResponse(BaseModel): + """Response schema for listing memory configs. + + Attributes: + configs: List of memory config items + total: Total number of configs + """ + configs: List[MemoryConfigItem] = Field(default_factory=list, description="List of configs") + total: int = Field(0, description="Total number of configs") diff --git a/api/app/schemas/memory_config_schema.py b/api/app/schemas/memory_config_schema.py index 0c359d70..e186e54b 100644 --- a/api/app/schemas/memory_config_schema.py +++ b/api/app/schemas/memory_config_schema.py @@ -387,6 +387,12 @@ class MemoryConfig: rerank_model_id: Optional[UUID] = None rerank_model_name: Optional[str] = None + video_model_id: Optional[UUID] = None + video_model_name: Optional[str] = None + vision_model_id: Optional[UUID] = None + vision_model_name: Optional[str] = None + audio_model_id: Optional[UUID] = None + audio_model_name: Optional[str] = None llm_params: Dict[str, Any] = field(default_factory=dict) embedding_params: Dict[str, Any] = field(default_factory=dict) @@ -417,7 +423,7 @@ class MemoryConfig: # Ontology scene association scene_id: Optional[UUID] = None - ontology_classes: Optional[list] = field(default=None) + ontology_class_infos: list[dict] = field(default_factory=list) def __post_init__(self): """Validate configuration after initialization.""" diff --git a/api/app/schemas/memory_storage_schema.py b/api/app/schemas/memory_storage_schema.py index 046b79e7..bfcf6337 100644 --- a/api/app/schemas/memory_storage_schema.py +++ b/api/app/schemas/memory_storage_schema.py @@ -8,9 +8,6 @@ import uuid from pydantic import BaseModel, Field, ConfigDict, field_validator, model_validator - - - # ============================================================================ # 从 json_schema.py 迁移的 Schema # ============================================================================ @@ -58,10 +55,13 @@ class MemoryVerifySchema(BaseModel): class ConflictResultSchema(BaseModel): """Schema for the conflict result data in the reflexion_data.json file.""" - data: List[BaseDataSchema] = Field(..., description="The conflict memory data. Only contains conflicting records when conflict is True.") + data: List[BaseDataSchema] = Field(..., + description="The conflict memory data. Only contains conflicting records when conflict is True.") conflict: bool = Field(..., description="Whether the memory is in conflict.") - quality_assessment: Optional[QualityAssessmentSchema] = Field(None, description="The quality assessment object. Contains score and summary when quality_assessment is enabled, null otherwise.") - memory_verify: Optional[MemoryVerifySchema] = Field(None, description="The memory privacy verification object. Contains privacy detection results when memory_verify is enabled, null otherwise.") + quality_assessment: Optional[QualityAssessmentSchema] = Field(None, + description="The quality assessment object. Contains score and summary when quality_assessment is enabled, null otherwise.") + memory_verify: Optional[MemoryVerifySchema] = Field(None, + description="The memory privacy verification object. Contains privacy detection results when memory_verify is enabled, null otherwise.") @model_validator(mode="before") def _normalize_data(cls, v): @@ -101,16 +101,19 @@ class ChangeRecordSchema(BaseModel): - entity2等嵌套对象的字段也遵循 [old_value, new_value] 格式 """ field: List[Dict[str, Any]] = Field( - ..., + ..., description="List of field changes. First item: {id: value}, followed by changed fields as {field_name: [old_value, new_value]} or {field_name: new_value} or nested structures like {entity2: {field_name: [old, new]}}" ) + class ResolvedSchema(BaseModel): """Schema for the resolved memory data in the reflexion_data""" original_memory_id: Optional[str] = Field(None, description="The original memory identifier.") # resolved_memory: Optional[BaseDataSchema] = Field(None, description="The resolved memory data (only contains records that need modification).") - resolved_memory: Optional[Union[BaseDataSchema, List[BaseDataSchema]]] = Field(None, description="The resolved memory data (only contains records that need modification). Can be a single record or list of records.") - change: Optional[List[ChangeRecordSchema]] = Field(None, description="List of detailed change records with IDs and field information.") + resolved_memory: Optional[Union[BaseDataSchema, List[BaseDataSchema]]] = Field(None, + description="The resolved memory data (only contains records that need modification). Can be a single record or list of records.") + change: Optional[List[ChangeRecordSchema]] = Field(None, + description="List of detailed change records with IDs and field information.") class SingleReflexionResultSchema(BaseModel): @@ -120,9 +123,11 @@ class SingleReflexionResultSchema(BaseModel): resolved: Optional[ResolvedSchema] = Field(None, description="The resolved memory data for this conflict.") type: str = Field("reflexion_result", description="The type identifier.") + class ReflexionResultSchema(BaseModel): """Schema for the complete reflexion result data - a list of individual conflict resolutions.""" - results: List[SingleReflexionResultSchema] = Field(..., description="List of individual conflict resolution results, grouped by conflict type.") + results: List[SingleReflexionResultSchema] = Field(..., + description="List of individual conflict resolution results, grouped by conflict type.") @model_validator(mode="before") def _normalize_resolved(cls, v): @@ -147,9 +152,9 @@ class ReflexionResultSchema(BaseModel): # Composite key identifying a config row class ConfigKey(BaseModel): # 配置参数键模型 model_config = ConfigDict(populate_by_name=True, extra="forbid") - config_id:Union[uuid.UUID, int, str] = Field(..., description="配置唯一标识(UUID或int)") - user_id: str = Field("user_id", description="用户标识(字符串)") - apply_id: str = Field("apply_id", description="应用或场景标识(字符串)") + config_id: Union[uuid.UUID, int, str] = Field(..., description="配置唯一标识(UUID或int)") + user_id: str | None = Field(default=None, description="用户标识(字符串)") + apply_id: str | None = Field(default=None, description="应用或场景标识(字符串)") # Allowed chunking strategies (extendable later) @@ -228,23 +233,25 @@ class ConfigParamsCreate(BaseModel): # 创建配置参数模型(仅 body, config_name: str = Field("配置名称", description="配置名称(字符串)") config_desc: str = Field("配置描述", description="配置描述(字符串)") workspace_id: Optional[uuid.UUID] = Field(None, description="工作空间ID(UUID)") - + # 本体场景关联(可选) scene_id: Optional[uuid.UUID] = Field(None, description="本体场景ID(UUID),关联ontology_scene表") - + # 语义剪枝场景(由 service 层根据 scene_id 自动推导,值为关联场景的 scene_name,前端无需传入) pruning_scene: Optional[str] = Field(None, description="语义剪枝场景,由 scene_id 对应的 scene_name 自动填充") - + # 模型配置字段(可选,用于手动指定或自动填充) llm_id: Optional[str] = Field(None, description="LLM模型配置ID") embedding_id: Optional[str] = Field(None, description="嵌入模型配置ID") rerank_id: Optional[str] = Field(None, description="重排序模型配置ID") reflection_model_id: Optional[str] = Field(None, description="反思模型ID,默认与llm_id一致") emotion_model_id: Optional[str] = Field(None, description="情绪分析模型ID,默认与llm_id一致") + + class ConfigParamsDelete(BaseModel): # 删除配置参数模型(请求体) model_config = ConfigDict(populate_by_name=True, extra="forbid") # config_name: str = Field("配置名称", description="配置名称(字符串)") - config_id:Union[uuid.UUID, int, str] = Field(..., description="配置ID(支持UUID、整数或字符串)") + config_id: Union[uuid.UUID, int, str] = Field(..., description="配置ID(支持UUID、整数或字符串)") class ConfigUpdate(BaseModel): # 更新记忆萃取引擎配置参数时使用的模型 @@ -255,8 +262,11 @@ class ConfigUpdate(BaseModel): # 更新记忆萃取引擎配置参数时使用 class ConfigUpdateExtracted(BaseModel): # 更新记忆萃取引擎配置参数时使用的模型 - config_id:Union[uuid.UUID, int, str] = None + config_id: Union[uuid.UUID, int, str] = None llm_id: Optional[str] = Field(None, description="LLM模型配置ID") + audio_id: Optional[str] = Field(None, description="语音模型ID") + vision_id: Optional[str] = Field(None, description="视觉模型ID") + video_id: Optional[str] = Field(None, description="视频模型ID") embedding_id: Optional[str] = Field(None, description="嵌入模型配置ID") rerank_id: Optional[str] = Field(None, description="重排序模型配置ID") enable_llm_dedup_blockwise: Optional[bool] = None @@ -322,14 +332,14 @@ class ConfigUpdateExtracted(BaseModel): # 更新记忆萃取引擎配置参数 class ConfigUpdateForget(BaseModel): # 更新遗忘引擎配置参数时使用的模型 # 遗忘引擎配置参数更新模型 - config_id:Union[uuid.UUID, int, str] = None + config_id: Union[uuid.UUID, int, str] = None lambda_time: Optional[float] = Field(0.5, ge=0.0, le=1.0, description="最低保持度,0-1 小数;默认 0.5") lambda_mem: Optional[float] = Field(0.5, ge=0.0, le=1.0, description="遗忘率,0-1 小数;默认 0.5") offset: Optional[float] = Field(0.0, ge=0.0, le=1.0, description="偏移度,0-1 小数;默认 0.0") class ConfigPilotRun(BaseModel): # 试运行触发请求模型 - config_id:Union[uuid.UUID, int, str] = Field(..., description="配置ID(唯一,支持UUID、整数或字符串)") + config_id: Union[uuid.UUID, int, str] = Field(..., description="配置ID(唯一,支持UUID、整数或字符串)") dialogue_text: str = Field(..., description="前端传入的对话文本,格式如 '用户: ...\nAI: ...' 可多行,试运行必填") custom_text: Optional[str] = Field(None, description="自定义输入文本,当配置关联本体场景时使用此字段进行试运行") model_config = ConfigDict(populate_by_name=True, extra="forbid") @@ -364,11 +374,11 @@ def ok(msg: str = "OK", data: Optional[Any] = None, time: Optional[int] = None) def fail( - msg: str, - error_code: str = "ERROR", - data: Optional[Any] = None, - time: Optional[int] = None, - query_preview: Optional[str] = None, + msg: str, + error_code: str = "ERROR", + data: Optional[Any] = None, + time: Optional[int] = None, + query_preview: Optional[str] = None, ) -> ApiResponse: payload = data if query_preview is not None: @@ -387,12 +397,13 @@ def fail( time=time or _now_ms(), ) + class GenerateCacheRequest(BaseModel): """缓存生成请求模型""" model_config = ConfigDict(populate_by_name=True, extra="forbid") - + end_user_id: Optional[str] = Field( - None, + None, description="终端用户ID(UUID格式)。如果提供,只为该用户生成;如果不提供,为当前工作空间的所有用户生成" ) @@ -404,7 +415,7 @@ class GenerateCacheRequest(BaseModel): class ForgettingTriggerRequest(BaseModel): """手动触发遗忘周期请求模型""" model_config = ConfigDict(populate_by_name=True, extra="forbid") - + end_user_id: str = Field(..., description="组ID(即终端用户ID,必填)") max_merge_batch_size: int = Field(100, ge=1, le=1000, description="单次最大融合节点对数(默认100)") min_days_since_access: int = Field(30, ge=1, le=365, description="最小未访问天数(默认30天)") @@ -413,7 +424,7 @@ class ForgettingTriggerRequest(BaseModel): class ForgettingConfigResponse(BaseModel): """遗忘引擎配置响应模型""" model_config = ConfigDict(populate_by_name=True, extra="forbid") - + config_id: Union[uuid.UUID, int, str] = Field(..., description="配置ID(支持UUID、整数或字符串)") decay_constant: float = Field(..., description="衰减常数 d") lambda_time: float = Field(..., description="时间衰减参数") @@ -432,7 +443,7 @@ class ForgettingConfigUpdateRequest(BaseModel): """遗忘引擎配置更新请求模型""" model_config = ConfigDict(populate_by_name=True, extra="forbid") - config_id: Union[uuid.UUID, int,str] = Field(..., description="配置唯一标识(UUID或int)") + config_id: Union[uuid.UUID, int, str] = Field(..., description="配置唯一标识(UUID或int)") decay_constant: Optional[float] = Field(None, ge=0.0, le=1.0, description="衰减常数 d") lambda_time: Optional[float] = Field(None, ge=0.0, le=1.0, description="时间衰减参数") lambda_mem: Optional[float] = Field(None, ge=0.0, le=1.0, description="记忆衰减参数") @@ -448,7 +459,7 @@ class ForgettingConfigUpdateRequest(BaseModel): class ForgettingCycleHistoryPoint(BaseModel): """遗忘周期历史数据点模型(用于趋势图)""" model_config = ConfigDict(populate_by_name=True, extra="forbid") - + date: str = Field(..., description="日期(格式: '1/1', '1/2')") merged_count: int = Field(..., description="每日融合节点数") average_activation: Optional[float] = Field(None, description="平均激活值") @@ -459,7 +470,7 @@ class ForgettingCycleHistoryPoint(BaseModel): class PendingForgettingNode(BaseModel): """待遗忘节点模型""" model_config = ConfigDict(populate_by_name=True, extra="forbid") - + node_id: str = Field(..., description="节点ID") node_type: str = Field(..., description="节点类型:statement/entity/summary") content_summary: str = Field(..., description="内容摘要") @@ -467,20 +478,36 @@ class PendingForgettingNode(BaseModel): last_access_time: int = Field(..., description="最后访问时间(Unix时间戳,秒)") +class PageInfo(BaseModel): + """分页信息模型""" + model_config = ConfigDict(populate_by_name=True, extra="forbid") + page: int = Field(..., description="当前页码(从1开始)") + pagesize: int = Field(..., description="每页数量") + total: int = Field(..., description="总记录数") + hasnext: bool = Field(..., description="是否有下一页") + + +class PendingNodesResponse(BaseModel): + """待遗忘节点列表响应模型(独立分页接口)""" + model_config = ConfigDict(populate_by_name=True, extra="forbid") + items: List[PendingForgettingNode] = Field(..., description="待遗忘节点列表") + page: PageInfo = Field(..., description="分页信息") + + class ForgettingStatsResponse(BaseModel): """遗忘引擎统计信息响应模型""" model_config = ConfigDict(populate_by_name=True, extra="forbid") activation_metrics: Dict[str, Any] = Field(..., description="激活值相关指标") node_distribution: Dict[str, int] = Field(..., description="节点类型分布") - recent_trends: List[ForgettingCycleHistoryPoint] = Field(..., description="最近7个日期的遗忘趋势数据(每天取最后一次执行)") - pending_nodes: List[PendingForgettingNode] = Field(..., description="待遗忘节点列表(前20个满足遗忘条件的节点)") + recent_trends: List[ForgettingCycleHistoryPoint] = Field(..., + description="最近7个日期的遗忘趋势数据(每天取最后一次执行)") timestamp: int = Field(..., description="统计时间(时间戳)") class ForgettingReportResponse(BaseModel): """遗忘周期报告响应模型""" model_config = ConfigDict(populate_by_name=True, extra="forbid") - + merged_count: int = Field(..., description="融合的节点对数量") nodes_before: int = Field(..., description="遗忘前的节点总数") nodes_after: int = Field(..., description="遗忘后的节点总数") @@ -495,7 +522,7 @@ class ForgettingReportResponse(BaseModel): class ForgettingCurvePoint(BaseModel): """遗忘曲线数据点模型""" model_config = ConfigDict(populate_by_name=True, extra="forbid") - + day: int = Field(..., description="天数") activation: float = Field(..., description="激活值") retention_rate: float = Field(..., description="保持率(与激活值相同)") @@ -504,7 +531,7 @@ class ForgettingCurvePoint(BaseModel): class ForgettingCurveRequest(BaseModel): """遗忘曲线请求模型""" model_config = ConfigDict(populate_by_name=True, extra="forbid") - + importance_score: float = Field(0.5, ge=0.0, le=1.0, description="重要性分数(0-1)") days: int = Field(60, ge=1, le=365, description="模拟天数(默认60天)") config_id: Union[uuid.UUID, int, str] = Field(..., description="配置唯一标识(UUID或int)") @@ -513,6 +540,6 @@ class ForgettingCurveRequest(BaseModel): class ForgettingCurveResponse(BaseModel): """遗忘曲线响应模型""" model_config = ConfigDict(populate_by_name=True, extra="forbid") - + curve_data: List[ForgettingCurvePoint] = Field(..., description="遗忘曲线数据点列表") config: Dict[str, Any] = Field(..., description="使用的配置参数") diff --git a/api/app/schemas/model_schema.py b/api/app/schemas/model_schema.py index 058f082d..668a84a8 100644 --- a/api/app/schemas/model_schema.py +++ b/api/app/schemas/model_schema.py @@ -81,6 +81,12 @@ class ModelConfig(ModelConfigBase): updated_at: datetime.datetime api_keys: List["ModelApiKey"] = [] + @staticmethod + def mask_api_key(key: str, prefix: int = 4, suffix: int = 4) -> str: + if not key or len(key) <= prefix + suffix: + return "*" * len(key) + return key[:prefix] + "*" * (len(key) - prefix - suffix) + key[-suffix:] + @field_validator("api_keys", mode="after") @classmethod def filter_active_api_keys(cls, api_keys: List["ModelApiKey"]) -> List["ModelApiKey"]: @@ -90,6 +96,15 @@ class ModelConfig(ModelConfigBase): def _serialize_created_at(self, dt: datetime.datetime | None): return int(dt.timestamp() * 1000) if dt else None + @field_serializer("api_keys", when_used="json") + def _serialize_api_keys(self, api_keys: List["ModelApiKey"]): + result = [] + for api_key in api_keys: + data = api_key.model_dump() + data["api_key"] = self.mask_api_key(api_key.api_key) + result.append(data) + return result + @field_serializer("updated_at", when_used="json") def _serialize_updated_at(self, dt: datetime.datetime): return int(dt.timestamp() * 1000) if dt else None @@ -165,20 +180,20 @@ class ModelApiKey(ModelApiKeyBase): if hasattr(self.model_configs, '__iter__') and not isinstance(self.model_configs, dict): self.model_config_ids = [ mc.id for mc in self.model_configs - if hasattr(mc, 'id') - and not getattr(mc, 'is_composite', False) - and getattr(mc, 'name', None) == self.model_name + if hasattr(mc, 'id') + and not getattr(mc, 'is_composite', False) + and getattr(mc, 'name', None) == self.model_name ] # 情况2:字典列表 elif isinstance(self.model_configs, list): self.model_config_ids = [ mc['id'] if isinstance(mc, dict) else mc.id for mc in self.model_configs - if ((isinstance(mc, dict) - and 'id' in mc + if ((isinstance(mc, dict) + and 'id' in mc and not mc.get('is_composite', False) - and mc.get('name') == self.model_name) or - (hasattr(mc, 'id') + and mc.get('name') == self.model_name) or + (hasattr(mc, 'id') and not getattr(mc, 'is_composite', False) and getattr(mc, 'name', None) == self.model_name)) ] @@ -193,11 +208,10 @@ class ModelApiKey(ModelApiKeyBase): validate_assignment=True # 确保赋值触发校验 ) - @field_serializer("created_at", when_used="json") def _serialize_created_at(self, dt: datetime.datetime): return int(dt.timestamp() * 1000) if dt else None - + @field_serializer("updated_at", when_used="json") def _serialize_updated_at(self, dt: datetime.datetime): return int(dt.timestamp() * 1000) if dt else None @@ -211,6 +225,7 @@ class ModelConfigQuery(BaseModel): """模型配置查询Schema""" type: Optional[List[ModelType]] = Field(None, description="模型类型筛选(支持多个)") provider: Optional[ModelProvider] = Field(None, description="提供商筛选(通过API Key)") + capability: Optional[List[str]] = Field(None, description="能力筛选(支持多个)") is_active: Optional[bool] = Field(None, description="激活状态筛选") is_public: Optional[bool] = Field(None, description="公开状态筛选") search: Optional[str] = Field(None, description="搜索关键词", max_length=255) @@ -228,6 +243,7 @@ class ModelConfigQueryNew(BaseModel): is_composite: Optional[bool] = Field(None, description="组合模型筛选") search: Optional[str] = Field(None, description="搜索关键词", max_length=255) + class ModelMarketplace(BaseModel): """模型广场响应Schema""" llm_models: List[ModelConfig] = [] @@ -304,7 +320,7 @@ class ModelBaseUpdate(BaseModel): class ModelBase(BaseModel): """基础模型Schema""" model_config = ConfigDict(from_attributes=True) - + id: uuid.UUID name: str type: str @@ -327,6 +343,7 @@ class ModelBaseQuery(BaseModel): is_deprecated: Optional[bool] = Field(None, description="是否弃用") search: Optional[str] = Field(None, description="搜索关键词", max_length=255) + class ModelInfo(BaseModel): """模型信息Schema""" model_name: str = Field(..., description="模型名称") @@ -336,4 +353,3 @@ class ModelInfo(BaseModel): is_omni: bool = Field(default=False, description="是否为omni模型") model_type: ModelType = Field(..., description="模型类型") capability: List[str] = Field(default_factory=list, description="模型能力列表") - diff --git a/api/app/schemas/user_schema.py b/api/app/schemas/user_schema.py index 6b880696..aa9ac256 100644 --- a/api/app/schemas/user_schema.py +++ b/api/app/schemas/user_schema.py @@ -1,6 +1,6 @@ from dataclasses import field from pydantic import BaseModel, EmailStr, Field, field_validator, validator, ConfigDict -from typing import Optional +from typing import Optional, List import datetime import uuid @@ -20,6 +20,7 @@ class UserCreate(UserBase): class UserUpdate(BaseModel): username: Optional[str] = None email: Optional[EmailStr] = None + phone: Optional[str] = None is_active: Optional[bool] = None is_superuser: Optional[bool] = None @@ -85,6 +86,8 @@ class User(UserBase): current_workspace_name: Optional[str] = None role: Optional[WorkspaceRole] = None preferred_language: Optional[str] = "zh" # 用户语言偏好 + phone: Optional[str] = None # 用户电话 + permissions: Optional[List[str]] = None # 用户权限列表,由 external_source 的 permissions 控制 # 将 datetime 转换为毫秒时间戳 @validator("created_at", pre=True) diff --git a/api/app/services/app_chat_service.py b/api/app/services/app_chat_service.py index 6fcf680b..bdccd787 100644 --- a/api/app/services/app_chat_service.py +++ b/api/app/services/app_chat_service.py @@ -93,7 +93,8 @@ class AppChatService: tools.extend(skill_tools) if skill_prompts: system_prompt = f"{system_prompt}\n\n{skill_prompts}" - tools.extend(self.agent_service.load_knowledge_retrieval_config(config.knowledge_retrieval, user_id)) + kb_tools, citations_collector = self.agent_service.load_knowledge_retrieval_config(config.knowledge_retrieval, user_id) + tools.extend(kb_tools) memory_flag = False if memory: memory_tools, memory_flag = self.agent_service.load_memory_config( @@ -128,46 +129,38 @@ class AppChatService: model_type=ModelType.LLM ) - # 加载历史消息 - messages = self.conversation_service.get_messages( + # 加载历史消息(包含开场白) + history = await self.conversation_service.get_conversation_history( conversation_id=conversation_id, - limit=10 + max_history=10, + current_provider=api_key_obj.provider, + current_is_omni=api_key_obj.is_omni ) - history = [] - for msg in messages: - content = [{"type": "text", "text": msg.content}] - # 处理 meta_data 中的 files - if msg.meta_data and msg.meta_data.get("files"): - files = msg.meta_data.get("files", []) - # 使用 MultimodalService 处理文件 - multimodal_service = MultimodalService(self.db, api_config=model_info) - - # 将 files 转换为 FileInput 格式 - file_inputs = [] - for file in files: - from app.schemas.app_schema import FileInput, TransferMethod - file_input = FileInput( - type=file.get("type"), - transfer_method=TransferMethod.REMOTE_URL, - url=file.get("url") - ) - file_inputs.append(file_input) - - history_processed_files = await multimodal_service.history_process_files(files=file_inputs) - - content.extend(history_processed_files) - - history.append({ - "role": msg.role, - "content": content - }) + # 如果是新会话且有开场白,作为第一条 assistant 消息写入数据库 + is_new_conversation = len(history) == 0 + if is_new_conversation: + opening, suggested_questions = self.agent_service._get_opening_statement(features_config, True, variables) + if opening: + self.conversation_service.add_message( + conversation_id=conversation_id, + role="assistant", + content=opening, + meta_data={"suggested_questions": suggested_questions} + ) + # 重新加载历史(包含刚写入的开场白) + history = await self.conversation_service.get_conversation_history( + conversation_id=conversation_id, + max_history=10, + current_provider=api_key_obj.provider, + current_is_omni=api_key_obj.is_omni + ) # 处理多模态文件 processed_files = None if files: multimodal_service = MultimodalService(self.db, model_info) - processed_files = await multimodal_service.process_files(user_id, files) + processed_files = await multimodal_service.process_files(files) logger.info(f"处理了 {len(processed_files)} 个文件") # 调用 Agent(支持多模态) @@ -204,14 +197,19 @@ class AppChatService: tenant_id=tenant_id, workspace_id=workspace_id ) + # 过滤 citations(只调用一次) + filtered_citations = self.agent_service._filter_citations(features_config, citations_collector) + # 构建用户消息内容(含多模态文件) human_meta = { - "files": [] + "files": [], + "history_files": {} } assistant_meta = { "model": api_key_obj.model_name, "usage": result.get("usage", {"prompt_tokens": 0, "completion_tokens": 0, "total_tokens": 0}), - "audio_url": None + "audio_url": None, + "citations": filtered_citations } if files: for f in files: @@ -221,6 +219,13 @@ class AppChatService: "url": f.url }) + if processed_files: + human_meta["history_files"] = { + "content": processed_files, + "provider": api_key_obj.provider, + "is_omni": api_key_obj.is_omni + } + # 保存消息 if audio_url: assistant_meta["audio_url"] = audio_url @@ -249,8 +254,9 @@ class AppChatService: }), "elapsed_time": elapsed_time, "suggested_questions": suggested_questions, - "citations": self.agent_service._filter_citations(features_config, result.get("citations", [])), + "citations": filtered_citations, "audio_url": audio_url, + "audio_status": "pending" } async def agnet_chat_stream( @@ -313,7 +319,8 @@ class AppChatService: tools.extend(skill_tools) if skill_prompts: system_prompt = f"{system_prompt}\n\n{skill_prompts}" - tools.extend(self.agent_service.load_knowledge_retrieval_config(config.knowledge_retrieval, user_id)) + kb_tools, citations_collector = self.agent_service.load_knowledge_retrieval_config(config.knowledge_retrieval, user_id) + tools.extend(kb_tools) # 添加长期记忆工具 memory_flag = False if memory: @@ -349,46 +356,38 @@ class AppChatService: model_type=ModelType.LLM ) - # 加载历史消息 - messages = self.conversation_service.get_messages( + # 加载历史消息(包含开场白) + history = await self.conversation_service.get_conversation_history( conversation_id=conversation_id, - limit=10 + max_history=10, + current_provider=api_key_obj.provider, + current_is_omni=api_key_obj.is_omni ) - history = [] - for msg in messages: - content = [{"type": "text", "text": msg.content}] - # 处理 meta_data 中的 files - if msg.meta_data and msg.meta_data.get("files"): - history_files = msg.meta_data.get("files", []) - # 使用 MultimodalService 处理文件 - multimodal_service = MultimodalService(self.db, api_config=model_info) - - # 将 files 转换为 FileInput 格式 - file_inputs = [] - for file in history_files: - from app.schemas.app_schema import FileInput, TransferMethod - file_input = FileInput( - type=file.get("type"), - transfer_method=TransferMethod.REMOTE_URL, - url=file.get("url") - ) - file_inputs.append(file_input) - - history_processed_files = await multimodal_service.history_process_files(files=file_inputs) - - content.extend(history_processed_files) - - history.append({ - "role": msg.role, - "content": content - }) + # 如果是新会话且有开场白,作为第一条 assistant 消息写入数据库 + is_new_conversation = len(history) == 0 + if is_new_conversation: + opening, suggested_questions = self.agent_service._get_opening_statement(features_config, True, variables) + if opening: + self.conversation_service.add_message( + conversation_id=conversation_id, + role="assistant", + content=opening, + meta_data={"suggested_questions": suggested_questions} + ) + # 重新加载历史(包含刚写入的开场白) + history = await self.conversation_service.get_conversation_history( + conversation_id=conversation_id, + max_history=10, + current_provider=api_key_obj.provider, + current_is_omni=api_key_obj.is_omni + ) # 处理多模态文件 processed_files = None if files: multimodal_service = MultimodalService(self.db, model_info) - processed_files = await multimodal_service.process_files(user_id, files) + processed_files = await multimodal_service.process_files(files) logger.info(f"处理了 {len(processed_files)} 个文件") # 流式调用 Agent(支持多模态),同时并行启动 TTS @@ -433,7 +432,7 @@ class AppChatService: elapsed_time = time.time() - start_time ModelApiKeyService.record_api_key_usage(self.db, api_key_obj.id) - # 发送结束事件(包含 suggested_questions、tts、citations) + # 发送结束事件(包含 suggested_questions、tts、audio_status、citations) end_data: dict = {"elapsed_time": elapsed_time, "message_length": len(full_content), "error": None} sq_config = features_config.get("suggested_questions_after_answer", {}) if isinstance(sq_config, dict) and sq_config.get("enabled"): @@ -443,25 +442,45 @@ class AppChatService: "api_base": api_key_obj.api_base}, {} ) end_data["audio_url"] = stream_audio_url - end_data["citations"] = self.agent_service._filter_citations(features_config, []) + # 检查TTS是否已完成(非阻塞,不取消任务) + audio_status = "pending" + if tts_task is not None and tts_task.done(): + # 任务已完成,检查是否有异常 + try: + tts_task.result() + audio_status = "completed" + except Exception as e: + logger.warning(f"TTS任务异常: {e}") + audio_status = "failed" + end_data["audio_status"] = audio_status if stream_audio_url else None + # 过滤 citations(只调用一次) + filtered_citations = self.agent_service._filter_citations(features_config, citations_collector) + end_data["citations"] = filtered_citations # 保存消息 human_meta = { - "files":[] + "files":[], + "history_files": {} } assistant_meta = { "model": api_key_obj.model_name, "usage": {"prompt_tokens": 0, "completion_tokens": 0, "total_tokens": total_tokens}, - "audio_url": None + "audio_url": None, + "citations": filtered_citations } if files: for f in files: - # url = await MultimodalService(self.db).get_file_url(f) human_meta["files"].append({ "type": f.type, "url": f.url }) + if processed_files: + human_meta["history_files"] = { + "content": processed_files, + "provider": api_key_obj.provider, + "is_omni": api_key_obj.is_omni + } if stream_audio_url: assistant_meta["audio_url"] = stream_audio_url diff --git a/api/app/services/app_log_service.py b/api/app/services/app_log_service.py new file mode 100644 index 00000000..856045d1 --- /dev/null +++ b/api/app/services/app_log_service.py @@ -0,0 +1,128 @@ +"""应用日志服务层""" +import uuid +from typing import Optional, Tuple +from datetime import datetime + +from sqlalchemy.orm import Session + +from app.core.logging_config import get_business_logger +from app.models.conversation_model import Conversation, Message +from app.repositories.conversation_repository import ConversationRepository, MessageRepository + +logger = get_business_logger() + + +class AppLogService: + """应用日志服务""" + + def __init__(self, db: Session): + self.db = db + self.conversation_repository = ConversationRepository(db) + self.message_repository = MessageRepository(db) + + def list_conversations( + self, + app_id: uuid.UUID, + workspace_id: uuid.UUID, + page: int = 1, + pagesize: int = 20, + is_draft: Optional[bool] = None, + ) -> Tuple[list[Conversation], int]: + """ + 查询应用日志会话列表 + + Args: + app_id: 应用 ID + workspace_id: 工作空间 ID + page: 页码(从 1 开始) + pagesize: 每页数量 + is_draft: 是否草稿会话(None 表示不过滤) + + Returns: + Tuple[list[Conversation], int]: (会话列表,总数) + """ + logger.info( + "查询应用日志会话列表", + extra={ + "app_id": str(app_id), + "workspace_id": str(workspace_id), + "page": page, + "pagesize": pagesize, + "is_draft": is_draft + } + ) + + # 使用 Repository 查询 + conversations, total = self.conversation_repository.list_app_conversations( + app_id=app_id, + workspace_id=workspace_id, + is_draft=is_draft, + page=page, + pagesize=pagesize + ) + + logger.info( + "查询应用日志会话列表成功", + extra={ + "app_id": str(app_id), + "total": total, + "returned": len(conversations) + } + ) + + return conversations, total + + def get_conversation_detail( + self, + app_id: uuid.UUID, + conversation_id: uuid.UUID, + workspace_id: uuid.UUID + ) -> Conversation: + """ + 查询会话详情(包含消息) + + Args: + app_id: 应用 ID + conversation_id: 会话 ID + workspace_id: 工作空间 ID + + Returns: + Conversation: 包含消息的会话对象 + + Raises: + ResourceNotFoundException: 当会话不存在时 + """ + logger.info( + "查询应用日志会话详情", + extra={ + "app_id": str(app_id), + "conversation_id": str(conversation_id), + "workspace_id": str(workspace_id) + } + ) + + # 查询会话 + conversation = self.conversation_repository.get_conversation_for_app_log( + conversation_id=conversation_id, + app_id=app_id, + workspace_id=workspace_id + ) + + # 查询消息(按时间正序) + messages = self.message_repository.get_messages_by_conversation( + conversation_id=conversation_id + ) + + # 将消息附加到会话对象 + conversation.messages = messages + + logger.info( + "查询应用日志会话详情成功", + extra={ + "app_id": str(app_id), + "conversation_id": str(conversation_id), + "message_count": len(messages) + } + ) + + return conversation diff --git a/api/app/services/app_service.py b/api/app/services/app_service.py index 19aaac42..377f9479 100644 --- a/api/app/services/app_service.py +++ b/api/app/services/app_service.py @@ -1084,7 +1084,6 @@ class AppService: if not exists: cleaned["memory_config_id"] = None cleaned.pop("memory_content", None) - cleaned["enabled"] = False return cleaned exists = self.db.query( @@ -1096,7 +1095,6 @@ class AppService: if not exists: cleaned["memory_config_id"] = None cleaned.pop("memory_content", None) - cleaned["enabled"] = False return cleaned @@ -1638,7 +1636,7 @@ class AppService: # ==================== 记忆配置提取方法 ==================== - def _extract_memory_config_id( + def _get_memory_config_id_from_release( self, app_type: str, config: Dict[str, Any] @@ -1684,15 +1682,15 @@ class AppService: return config.config_id - def _update_endusers_memory_config_by_workspace( + def _update_endusers_memory_config_by_app( self, - workspace_id: uuid.UUID, + app_id: uuid.UUID, memory_config_id: uuid.UUID ) -> int: """批量更新应用下所有终端用户的 memory_config_id Args: - workspace_id: 工作空间ID + app_id: 应用ID memory_config_id: 新的记忆配置ID Returns: @@ -1701,8 +1699,8 @@ class AppService: from app.repositories.end_user_repository import EndUserRepository repo = EndUserRepository(self.db) - updated_count = repo.batch_update_memory_config_id_by_workspace( - workspace_id=workspace_id, + updated_count = repo.batch_update_memory_config_id_by_app( + app_id=app_id, memory_config_id=memory_config_id ) @@ -1753,12 +1751,16 @@ class AppService: miss_params = [] if agent_cfg.default_model_config_id is None: - miss_params.append("model config") + miss_params.append("模型配置") if agent_cfg.memory.get("enabled") and not agent_cfg.memory.get("memory_config_id"): - miss_params.append("memory config") + miss_params.append("记忆配置") if miss_params: - raise BusinessException(f"{', '.join(miss_params)} is required") + raise BusinessException( + f"应用发布失败:检测到以下必要配置尚未完成:{', '.join(miss_params)}。请返回应用编辑页面完成相关配置后再尝试发布。", + BizCode.CONFIG_MISSING, + context={"missing_params": miss_params}, + ) config = { "system_prompt": agent_cfg.system_prompt, @@ -1863,7 +1865,7 @@ class AppService: self.db.flush() # 先 flush,确保 release 已插入数据库 # 提取记忆配置ID并更新终端用户 - memory_config_id, is_legacy_int = self._extract_memory_config_id(app.type, config) + memory_config_id, is_legacy_int = self._get_memory_config_id_from_release(app.type, config) # 如果检测到旧格式 int 数据,回退到工作空间默认配置 if is_legacy_int and not memory_config_id: @@ -1877,8 +1879,8 @@ class AppService: if memory_config_id: app = self.db.query(App).filter(App.id == app_id).first() if app: - updated_count = self._update_endusers_memory_config_by_workspace( - app.workspace_id, memory_config_id + updated_count = self._update_endusers_memory_config_by_app( + app_id, memory_config_id ) logger.info( f"发布时更新终端用户记忆配置: app_id={app_id}, workspace_id={app.workspace_id}, " @@ -2001,7 +2003,7 @@ class AppService: raise ResourceNotFoundException("发布版本", f"app_id={app_id}, version={version}") # 提取记忆配置ID并更新终端用户 - memory_config_id, is_legacy_int = self._extract_memory_config_id(release.type, release.config) + memory_config_id, is_legacy_int = self._get_memory_config_id_from_release(release.type, release.config) # 如果检测到旧格式 int 数据,回退到工作空间默认配置 if is_legacy_int and not memory_config_id: @@ -2014,7 +2016,7 @@ class AppService: if memory_config_id: - updated_count = self._update_endusers_memory_config_by_workspace(app.workspace_id, memory_config_id) + updated_count = self._update_endusers_memory_config_by_app(app_id, memory_config_id) logger.info( f"回滚时更新终端用户记忆配置: app_id={app_id}, version={version}, " f"memory_config_id={memory_config_id}, updated_count={updated_count}" diff --git a/api/app/services/conversation_service.py b/api/app/services/conversation_service.py index f8a01a40..bd7f7496 100644 --- a/api/app/services/conversation_service.py +++ b/api/app/services/conversation_service.py @@ -214,7 +214,7 @@ class ConversationService: conversation.message_count += 1 - if conversation.message_count == 1 and role == "user": + if conversation.message_count <= 2 and role == "user": conversation.title = ( content[:50] + ("..." if len(content) > 50 else "") ) @@ -274,7 +274,8 @@ class ConversationService: self, conversation_id: uuid.UUID, max_history: Optional[int] = None, - api_config: Optional[ModelInfo] = None + current_provider: Optional[str] = None, + current_is_omni: Optional[bool] = None ) -> List[dict]: """ Retrieve historical conversation messages formatted as dictionaries. @@ -282,7 +283,8 @@ class ConversationService: Args: conversation_id (uuid.UUID): Conversation UUID. max_history (Optional[int]): Maximum number of messages to retrieve. - api_config (Optional[ModelInfo]): Model API configuration for multimodal processing. + current_provider (Optional[str]): Current provider for file handling. + current_is_omni (Optional[bool]): Current omni flag for file handling. Returns: List[dict]: List of message dictionaries with keys 'role' and 'content'. @@ -292,38 +294,30 @@ class ConversationService: limit=max_history ) - # 转换为字典格式 history = [] for msg in messages: - content = [{"type": "text", "text": msg.content}] - - # 处理 meta_data 中的 files - if msg.meta_data and msg.meta_data.get("files"): - files = msg.meta_data.get("files", []) - if api_config: - # 使用 MultimodalService 处理文件 - from app.services.multimodal_service import MultimodalService - multimodal_service = MultimodalService(self.db, api_config=api_config) - - # 将 files 转换为 FileInput 格式 - file_inputs = [] - for file in files: - from app.schemas.app_schema import FileInput, TransferMethod - file_input = FileInput( - type=file.get("type"), - transfer_method=TransferMethod.REMOTE_URL, - url=file.get("url") - ) - file_inputs.append(file_input) - - processed_files = await multimodal_service.history_process_files(files=file_inputs) - - content.extend(processed_files) - - history.append({ + msg_dict = { "role": msg.role, - "content": content - }) + "content": [{"type": "text", "text": msg.content}] + } + + # 处理用户消息中的多模态文件 + if msg.role == "user" and msg.meta_data: + history_files = msg.meta_data.get("history_files", {}) + + if history_files and current_provider and current_is_omni is not None: + # 检查是否需要重新处理文件 + stored_provider = history_files.get("provider") + stored_is_omni = history_files.get("is_omni") + + # 如果provider或is_omni不匹配,需要重新处理 + if stored_provider != current_provider or stored_is_omni != current_is_omni: + continue + + # provider和is_omni匹配,直接使用存储的内容 + msg_dict["content"].extend(history_files.get("content")) + + history.append(msg_dict) return history @@ -539,6 +533,7 @@ class ConversationService: provider = api_config.provider api_key = api_config.api_key api_base = api_config.api_base + is_omni = api_config.is_omni model_type = config.type llm = RedBearLLM( @@ -546,7 +541,8 @@ class ConversationService: model_name=model_name, provider=provider, api_key=api_key, - base_url=api_base + base_url=api_base, + is_omni=is_omni ), type=ModelType(model_type) ) @@ -554,15 +550,8 @@ class ConversationService: conversation_messages = await self.get_conversation_history( conversation_id=conversation_id, max_history=20, - api_config=ModelInfo( - model_name=model_name, - provider=provider, - api_key=api_key, - api_base=api_base, - capability=api_config.capability, - is_omni=api_config.is_omni, - model_type=model_type - ) + current_provider=provider, + current_is_omni=is_omni ) if len(conversation_messages) == 0: return ConversationOut( diff --git a/api/app/services/draft_run_service.py b/api/app/services/draft_run_service.py index 5989f0f8..c658cf93 100644 --- a/api/app/services/draft_run_service.py +++ b/api/app/services/draft_run_service.py @@ -26,7 +26,7 @@ from app.core.rag.nlp.search import knowledge_retrieval from app.db import get_db_context from app.models import AgentConfig, ModelConfig, ModelType from app.repositories.tool_repository import ToolRepository -from app.schemas.app_schema import FileInput +from app.schemas.app_schema import FileInput, Citation from app.schemas.model_schema import ModelInfo from app.schemas.prompt_schema import PromptMessageRole, render_prompt_message from app.services import task_service @@ -190,13 +190,19 @@ def create_web_search_tool(web_search_config: Dict[str, Any]): return web_search_tool -def create_knowledge_retrieval_tool(kb_config, kb_ids, user_id): +def create_knowledge_retrieval_tool(kb_config, kb_ids, user_id, citations_collector: Optional[List[Citation]] = None): """从知识库中检索相关信息。当用户的问题需要参考知识库、文档或历史记录时,使用此工具进行检索。 Args: kb_config: 知识库配置 kb_ids: 知识库ID列表 user_id: 用户ID + citations_collector: 用于收集引用信息的列表(由外部传入,tool 执行时填充) + 列表元素类型为 Citation,包含字段: + - document_id: 文档唯一标识 + - file_name: 文件名 + - knowledge_id: 知识库 ID + - score: 检索相关性得分 Returns: 检索到的相关知识内容 @@ -229,6 +235,21 @@ def create_knowledge_retrieval_tool(kb_config, kb_ids, user_id): } ) + # 收集引用信息 + if citations_collector is not None: + seen_doc_ids = {c.get("document_id") for c in citations_collector} + for chunk in retrieve_chunks_result: + meta = chunk.metadata or {} + doc_id = meta.get("document_id") or meta.get("doc_id") + if doc_id and doc_id not in seen_doc_ids: + seen_doc_ids.add(doc_id) + citations_collector.append(Citation( + document_id=doc_id, + file_name=meta.get("file_name", ""), + knowledge_id=str(meta.get("knowledge_id", "")), + score=meta.get("score", 0) + )) + return f"检索到以下相关信息:\n\n{context}" else: logger.warning("知识库检索未找到结果") @@ -320,26 +341,26 @@ class AgentRunService: self, knowledge_retrieval_config: dict | None, user_id - ) -> list: + ) -> tuple[list, list]: + """返回 (tools, citations_collector)""" if not knowledge_retrieval_config: - return [] + return [], [] + citations_collector = [] tools = [] knowledge_bases = knowledge_retrieval_config.get("knowledge_bases", []) - kb_ids = bool(knowledge_bases and knowledge_bases[0].get("kb_id")) + kb_ids = [kb["kb_id"] for kb in knowledge_bases if kb.get("kb_id")] if kb_ids: - # 创建知识库检索工具 - kb_tool = create_knowledge_retrieval_tool(knowledge_retrieval_config, kb_ids, user_id) + kb_tool = create_knowledge_retrieval_tool( + knowledge_retrieval_config, kb_ids, user_id, + citations_collector=citations_collector + ) tools.append(kb_tool) - logger.debug( "已添加知识库检索工具", - extra={ - "kb_ids": kb_ids, - "tool_count": len(tools) - } + extra={"kb_ids": kb_ids, "tool_count": len(tools)} ) - return tools + return tools, citations_collector def load_memory_config( self, @@ -424,29 +445,38 @@ class AgentRunService: ) @staticmethod - def _inject_opening_statement( + def _get_opening_statement( features_config: Dict[str, Any], - system_prompt: str, - is_new_conversation: bool - ) -> str: - """首轮对话时将开场白注入 system_prompt""" + is_new_conversation: bool, + variables: Optional[Dict[str, Any]] = None + ) -> tuple[Any, Any]: + """首轮对话时返回开场白文本(支持变量替换),否则返回 None""" if not is_new_conversation: - return system_prompt + return None, None opening = features_config.get("opening_statement", {}) if not (isinstance(opening, dict) and opening.get("enabled") and opening.get("statement")): - return system_prompt + return None, None + statement = opening["statement"] - return f"{system_prompt}\n\n[对话开场白]\n{statement}" + suggested_questions = opening["suggested_questions"] + + # 如果有变量,进行替换(仅支持 {{var_name}} 格式) + if variables: + for var_name, var_value in variables.items(): + placeholder = f"{{{{{var_name}}}}}" + statement = statement.replace(placeholder, str(var_value)) + + return statement, suggested_questions @staticmethod def _filter_citations( features_config: Dict[str, Any], - citations: List[Any] + citations: List[Citation] ) -> List[Any]: """根据 citation 开关决定是否返回引用来源""" citation_cfg = features_config.get("citation", {}) if isinstance(citation_cfg, dict) and citation_cfg.get("enabled"): - return citations + return [cit.model_dump() for cit in citations] return [] async def run( @@ -534,10 +564,6 @@ class AgentRunService: # 3. 处理系统提示词(支持变量替换) system_prompt = system_prompt.get_text_content() or "你是一个专业的AI助手" - # opening_statement:首轮对话注入开场白 - is_new_conversation = not conversation_id - system_prompt = self._inject_opening_statement(features_config, system_prompt, is_new_conversation) - # 4. 准备工具列表 tools = [] @@ -549,7 +575,8 @@ class AgentRunService: tools.extend(skill_tools) if skill_prompts: system_prompt = f"{system_prompt}\n\n{skill_prompts}" - tools.extend(self.load_knowledge_retrieval_config(knowledge_retrieval_config, user_id)) + kb_tools, citations_collector = self.load_knowledge_retrieval_config(knowledge_retrieval_config, user_id) + tools.extend(kb_tools) # 添加长期记忆工具 memory_flag = False if memory: @@ -571,12 +598,18 @@ class AgentRunService: tools=tools, ) - # 5. 处理会话ID(创建或验证) + # 5. 处理会话ID(创建或验证),新会话时写入开场白 + is_new_conversation = not conversation_id + opening, suggested_questions = None, None + if not sub_agent: + opening, suggested_questions = self._get_opening_statement(features_config, is_new_conversation, variables) conversation_id = await self._ensure_conversation( conversation_id=conversation_id, app_id=agent_config.app_id, workspace_id=workspace_id, - user_id=user_id + user_id=user_id, + opening_statement=opening, + suggested_questions=suggested_questions ) model_info = ModelInfo( @@ -589,11 +622,12 @@ class AgentRunService: model_type=model_config.type ) - # 6. 加载历史消息 + # 6. 加载历史消息(包含开场白) history = await self._load_conversation_history( conversation_id=conversation_id, - api_config=model_info, - max_history=10 + max_history=10, + current_provider=api_key_config.get("provider"), + current_is_omni=api_key_config.get("is_omni", False) ) # 6. 处理多模态文件 @@ -602,7 +636,7 @@ class AgentRunService: # 获取 provider 信息 provider = api_key_config.get("provider", "openai") multimodal_service = MultimodalService(self.db, model_info) - processed_files = await multimodal_service.process_files(user_id, files) + processed_files = await multimodal_service.process_files(files) logger.info(f"处理了 {len(processed_files)} 个文件,provider={provider}") # 7. 知识库检索 @@ -645,6 +679,9 @@ class AgentRunService: tenant_id=tenant_id, workspace_id=workspace_id ) if not sub_agent else None + # 过滤 citations(只调用一次) + filtered_citations = self._filter_citations(features_config, citations_collector) + # 10. 保存会话消息 if not sub_agent: await self._save_conversation_message( @@ -661,7 +698,11 @@ class AgentRunService: }) }, files=files, - audio_url=audio_url + processed_files=processed_files, + audio_url=audio_url, + citations=filtered_citations, + provider=api_key_config.get("provider"), + is_omni=api_key_config.get("is_omni", False) ) response = { @@ -676,8 +717,9 @@ class AgentRunService: "suggested_questions": await self._generate_suggested_questions( features_config, result["content"], api_key_config, effective_params ) if not sub_agent else [], - "citations": self._filter_citations(features_config, result.get("citations", [])), + "citations": filtered_citations, "audio_url": audio_url, + "audio_status": "pending" } logger.info( @@ -770,10 +812,6 @@ class AgentRunService: # 3. 处理系统提示词(支持变量替换) system_prompt = system_prompt.get_text_content() or "你是一个专业的AI助手" - # opening_statement:首轮对话注入开场白 - is_new_conversation = not conversation_id - system_prompt = self._inject_opening_statement(features_config, system_prompt, is_new_conversation) - # 4. 准备工具列表 tools = [] @@ -785,7 +823,8 @@ class AgentRunService: tools.extend(skill_tools) if skill_prompts: system_prompt = f"{system_prompt}\n\n{skill_prompts}" - tools.extend(self.load_knowledge_retrieval_config(knowledge_retrieval_config, user_id)) + kb_tools, citations_collector = self.load_knowledge_retrieval_config(knowledge_retrieval_config, user_id) + tools.extend(kb_tools) # 添加长期记忆工具 memory_flag = False @@ -808,13 +847,19 @@ class AgentRunService: streaming=True ) - # 5. 处理会话ID(创建或验证) + # 5. 处理会话ID(创建或验证),新会话时写入开场白 + is_new_conversation = not conversation_id + opening, suggested_questions = None, None + if not sub_agent: + opening, suggested_questions = self._get_opening_statement(features_config, is_new_conversation, variables) conversation_id = await self._ensure_conversation( conversation_id=conversation_id, app_id=agent_config.app_id, workspace_id=workspace_id, user_id=user_id, - sub_agent=sub_agent + sub_agent=sub_agent, + opening_statement=opening, + suggested_questions=suggested_questions ) model_info = ModelInfo( @@ -830,8 +875,9 @@ class AgentRunService: # 6. 加载历史消息 history = await self._load_conversation_history( conversation_id=conversation_id, - api_config=model_info, - max_history=memory_config.get("max_history", 10) + max_history=memory_config.get("max_history", 10), + current_provider=api_key_config.get("provider"), + current_is_omni=api_key_config.get("is_omni", False) ) # 6. 处理多模态文件 @@ -840,7 +886,7 @@ class AgentRunService: # 获取 provider 信息 provider = api_key_config.get("provider", "openai") multimodal_service = MultimodalService(self.db, model_info) - processed_files = await multimodal_service.process_files(user_id, files) + processed_files = await multimodal_service.process_files(files) logger.info(f"处理了 {len(processed_files)} 个文件,provider={provider}") # 7. 知识库检索 @@ -897,6 +943,9 @@ class AgentRunService: if sub_agent: yield self._format_sse_event("sub_usage", {"total_tokens": total_tokens}) + # 过滤 citations(只调用一次) + filtered_citations = self._filter_citations(features_config, citations_collector) + # 11. 保存会话消息 if not sub_agent: await self._save_conversation_message( @@ -909,10 +958,14 @@ class AgentRunService: "usage": {"prompt_tokens": 0, "completion_tokens": 0, "total_tokens": total_tokens} }, files=files, - audio_url=stream_audio_url + processed_files=processed_files, + audio_url=stream_audio_url, + citations=filtered_citations, + provider=api_key_config.get("provider"), + is_omni=api_key_config.get("is_omni", False) ) - # 12. 发送结束事件(包含 suggested_questions 和 tts) + # 12. 发送结束事件(包含 suggested_questions、audio_url 和 audio_status) end_data: Dict[str, Any] = { "conversation_id": conversation_id, "elapsed_time": elapsed_time, @@ -923,7 +976,18 @@ class AgentRunService: features_config, full_content, api_key_config, effective_params ) end_data["audio_url"] = stream_audio_url - end_data["citations"] = self._filter_citations(features_config, []) + # 检查TTS是否已完成(非阻塞,不取消任务) + audio_status = "pending" + if tts_task is not None and tts_task.done(): + # 任务已完成,检查是否有异常 + try: + tts_task.result() + audio_status = "completed" + except Exception as e: + logger.warning(f"TTS任务异常: {e}") + audio_status = "failed" + end_data["audio_status"] = audio_status if stream_audio_url else None + end_data["citations"] = filtered_citations yield self._format_sse_event("end", end_data) logger.info( @@ -1003,7 +1067,9 @@ class AgentRunService: app_id: uuid.UUID, workspace_id: uuid.UUID, user_id: Optional[str], - sub_agent: bool = False + sub_agent: bool = False, + opening_statement: Optional[str] = None, + suggested_questions: Optional[List[str]] = None ) -> str: """确保会话存在(创建或验证) @@ -1012,6 +1078,9 @@ class AgentRunService: app_id: 应用ID workspace_id: 工作空间ID(必须) user_id: 用户ID + sub_agent: 是否为子代理 + opening_statement: 开场白(新会话时作为第一条消息写入) + suggested_questions: 预设问题列表 Returns: str: 会话ID @@ -1049,6 +1118,16 @@ class AgentRunService: self.db.commit() self.db.refresh(new_conversation) + # 如果有开场白,作为第一条 assistant 消息写入数据库 + if opening_statement: + conversation_service.add_message( + conversation_id=uuid.UUID(new_conv_id), + role="assistant", + content=opening_statement, + meta_data={"suggested_questions": suggested_questions} + ) + logger.debug(f"已保存开场白到会话 {new_conv_id}") + logger.info( "创建草稿会话成功", extra={ @@ -1119,14 +1198,17 @@ class AgentRunService: async def _load_conversation_history( self, conversation_id: str, - api_config: ModelInfo | None = None, - max_history: int = 10 + max_history: int = 10, + current_provider: Optional[str] = None, + current_is_omni: Optional[bool] = None ) -> List[Dict[str, str]]: - """加载会话历史消息 + """加载会话历史消息,并根据当前模型配置处理多模态文件 Args: conversation_id: 会话ID max_history: 最大历史消息数量 + current_provider: 当前模型的provider + current_is_omni: 当前模型的is_omni Returns: List[Dict]: 历史消息列表 @@ -1138,7 +1220,8 @@ class AgentRunService: history = await conversation_service.get_conversation_history( conversation_id=uuid.UUID(conversation_id), max_history=max_history, - api_config=api_config + current_provider=current_provider, + current_is_omni=current_is_omni ) logger.debug( @@ -1166,7 +1249,11 @@ class AgentRunService: app_id: Optional[uuid.UUID] = None, user_id: Optional[str] = None, files: Optional[List[FileInput]] = None, - audio_url: Optional[str] = None + processed_files: Optional[List[Dict[str, Any]]] = None, + audio_url: Optional[str] = None, + citations: Optional[List[Any]] = None, + provider: Optional[str] = None, + is_omni: Optional[bool] = None ) -> None: """保存会话消息(会话已通过 _ensure_conversation 确保存在) @@ -1177,6 +1264,12 @@ class AgentRunService: app_id: 应用ID(未使用,保留用于兼容性) user_id: 用户ID(未使用,保留用于兼容性) meta_data: token消耗 + files: 原始文件输入 + processed_files: 处理后的文件 + audio_url: 音频URL + citations: 引用来源列表 + provider: 模型供应商 + is_omni: 是否为全模态模型 """ try: from app.services.conversation_service import ConversationService @@ -1186,15 +1279,24 @@ class AgentRunService: # 保存消息(会话已经存在) human_meta = { - "files": [] + "files": [], + "history_files": {} } if files: for f in files: - # url = await MultimodalService(self.db).get_file_url(f) human_meta["files"].append({ "type": f.type, "url": f.url }) + + # 保存 history_files,包含 provider 和 is_omni 信息 + if processed_files: + human_meta["history_files"] = { + "content": processed_files, + "provider": provider, + "is_omni": is_omni + } + # 保存用户消息 conversation_service.add_message( conversation_id=conv_uuid, @@ -1202,9 +1304,11 @@ class AgentRunService: content=user_message, meta_data=human_meta ) - # 保存助手消息(含 audio_url) + # 保存助手消息(含 audio_url 和 citations) if audio_url: meta_data["audio_url"] = audio_url + if citations: + meta_data["citations"] = citations conversation_service.add_message( conversation_id=conv_uuid, role="assistant", @@ -1420,8 +1524,9 @@ class AgentRunService: workspace_id: Optional[uuid.UUID] = None, ) -> tuple[Optional[str], Optional[asyncio.Task]]: """文本流式输入并行合成音频。 - 返回 (audio_url, task),audio_url 立即可用,task 完成后文件内容就绪。 + 返回 (audio_url, task),audio_url 立即可用(pending状态),task 完成后文件内容就绪。 调用方向 text_queue put 文本 chunk,结束时 put None。 + 前端可通过 GET /storage/files/{file_id}/status 轮询检查音频是否就绪。 """ tts_config = features_config.get("text_to_speech", {}) if not isinstance(tts_config, dict) or not tts_config.get("enabled"): @@ -1808,6 +1913,7 @@ class AgentRunService: ), "cost_estimate": self._estimate_cost(usage, model_info["model_config"]), "audio_url": result.get("audio_url"), + "audio_status": result.get("audio_status"), "citations": result.get("citations", []), "suggested_questions": result.get("suggested_questions", []), "error": None @@ -1885,6 +1991,7 @@ class AgentRunService: "results": [{ **r, "audio_url": r.get("audio_url"), + "audio_status": r.get("audio_status"), "citations": r.get("citations", []), "suggested_questions": r.get("suggested_questions", []), } for r in results], @@ -2016,6 +2123,7 @@ class AgentRunService: full_content = "" returned_conversation_id = model_conversation_id audio_url = None + audio_status = None citations = [] suggested_questions = [] @@ -2074,6 +2182,7 @@ class AgentRunService: # 从 end 事件中提取 features 输出字段 if event_type == "end" and event_data: audio_url = event_data.get("audio_url") + audio_status = event_data.get("audio_status") citations = event_data.get("citations", []) suggested_questions = event_data.get("suggested_questions", []) @@ -2103,6 +2212,7 @@ class AgentRunService: "message": full_content, "elapsed_time": elapsed, "audio_url": audio_url, + "audio_status": audio_status, "citations": citations, "suggested_questions": suggested_questions, "error": None @@ -2117,6 +2227,7 @@ class AgentRunService: "elapsed_time": elapsed, "message_length": len(full_content), "audio_url": audio_url, + "audio_status": audio_status, "citations": citations, "suggested_questions": suggested_questions, "timestamp": time.time() @@ -2253,6 +2364,7 @@ class AgentRunService: "message": r.get("message"), "elapsed_time": r.get("elapsed_time", 0), "audio_url": r.get("audio_url"), + "audio_status": r.get("audio_status"), "citations": r.get("citations", []), "suggested_questions": r.get("suggested_questions", []), "error": r.get("error") diff --git a/api/app/services/file_storage_service.py b/api/app/services/file_storage_service.py index 2ebc5d9a..5897936b 100644 --- a/api/app/services/file_storage_service.py +++ b/api/app/services/file_storage_service.py @@ -325,27 +325,30 @@ class FileStorageService: ) raise - async def get_file_url(self, file_key: str, expires: int = 3600) -> str: + async def get_file_url( + self, + file_key: str, + expires: int = 3600, + file_name: Optional[str] = None, + ) -> str: """ Get an access URL for a file. Args: file_key: The file key. expires: URL validity period in seconds (default: 1 hour). + file_name: If set, adds Content-Disposition: attachment to force download. Returns: URL for accessing the file. """ logger.debug(f"Getting file URL: file_key={file_key}, expires={expires}s") - try: - url = await self.storage.get_url(file_key, expires) + url = await self.storage.get_url(file_key, expires, file_name=file_name) logger.debug(f"File URL generated: file_key={file_key}") return url except Exception as e: - logger.error( - f"Error getting file URL: file_key={file_key}, error={str(e)}" - ) + logger.error(f"Error getting file URL: file_key={file_key}, error={str(e)}") raise diff --git a/api/app/services/generation_service.py b/api/app/services/generation_service.py new file mode 100644 index 00000000..2505793c --- /dev/null +++ b/api/app/services/generation_service.py @@ -0,0 +1,162 @@ +""" +图片和视频生成服务 + +提供统一的生成接口,支持多种 Provider +""" +from typing import Dict, Any, Optional +from sqlalchemy.orm import Session +import uuid + +from app.core.models import RedBearModelConfig, RedBearImageGenerator, RedBearVideoGenerator +from app.core.exceptions import BusinessException +from app.core.error_codes import BizCode +from app.models.models_model import ModelType +from app.repositories.model_repository import ModelConfigRepository, ModelApiKeyRepository +from app.services.model_service import ModelApiKeyService + + +class GenerationService: + """生成服务""" + + def __init__(self, db: Session): + self.db = db + + async def generate_image( + self, + model_config_id: str, + prompt: str, + size: Optional[str] = "2k", + **kwargs + ) -> Dict[str, Any]: + """ + 生成图片 + + Args: + model_config_id: 模型配置ID + prompt: 提示词 + size: 图片尺寸 + **kwargs: 其他参数 + + Returns: + 生成结果 + """ + # 获取模型配置 + model_config = ModelConfigRepository.get_by_id(self.db, uuid.UUID(model_config_id)) + if not model_config: + raise BusinessException("模型配置不存在", code=BizCode.NOT_FOUND) + + if model_config.type != ModelType.IMAGE: + raise BusinessException( + f"模型类型错误,期望 {ModelType.IMAGE},实际 {model_config.type}", + code=BizCode.INVALID_PARAMETER + ) + + # 获取 API Key + api_key_info = ModelApiKeyService.get_available_api_key(self.db, uuid.UUID(model_config_id)) + if not api_key_info: + raise BusinessException("没有可用的 API Key", code=BizCode.NOT_FOUND) + + # 创建配置 + config = RedBearModelConfig( + model_name=api_key_info.model_name, + provider=api_key_info.provider, + api_key=api_key_info.api_key, + base_url=api_key_info.api_base, + extra_params=api_key_info.config or {} + ) + + # 生成图片 + generator = RedBearImageGenerator(config) + result = await generator.agenerate(prompt, size, **kwargs) + + return result + + async def generate_video( + self, + model_config_id: str, + prompt: str, + duration: Optional[int] = None, + **kwargs + ) -> Dict[str, Any]: + """ + 生成视频 + + Args: + model_config_id: 模型配置ID + prompt: 提示词 + duration: 视频时长(秒) + **kwargs: 其他参数 + + Returns: + 生成结果(包含任务ID) + """ + # 获取模型配置 + model_config = ModelConfigRepository.get_by_id(self.db, uuid.UUID(model_config_id)) + if not model_config: + raise BusinessException("模型配置不存在", code=BizCode.NOT_FOUND) + + if model_config.type != ModelType.VIDEO: + raise BusinessException( + f"模型类型错误,期望 {ModelType.VIDEO},实际 {model_config.type}", + code=BizCode.INVALID_PARAMETER + ) + + # 获取 API Key + api_key_info = ModelApiKeyService.get_available_api_key(self.db, uuid.UUID(model_config_id)) + if not api_key_info: + raise BusinessException("没有可用的 API Key", code=BizCode.NOT_FOUND) + + # 创建配置 + config = RedBearModelConfig( + model_name=api_key_info.model_name, + provider=api_key_info.provider, + api_key=api_key_info.api_key, + base_url=api_key_info.api_base, + extra_params=api_key_info.config or {} + ) + + # 生成视频 + generator = RedBearVideoGenerator(config) + result = await generator.agenerate(prompt, duration, **kwargs) + + return result + + async def get_video_task_status( + self, + model_config_id: str, + task_id: str + ) -> Dict[str, Any]: + """ + 查询视频生成任务状态 + + Args: + model_config_id: 模型配置ID + task_id: 任务ID + + Returns: + 任务状态信息 + """ + # 获取模型配置 + model_config = ModelConfigRepository.get_by_id(self.db, uuid.UUID(model_config_id)) + if not model_config: + raise BusinessException("模型配置不存在", code=BizCode.NOT_FOUND) + + # 获取 API Key + api_key_info = ModelApiKeyService.get_available_api_key(self.db, uuid.UUID(model_config_id)) + if not api_key_info: + raise BusinessException("没有可用的 API Key", code=BizCode.NOT_FOUND) + + # 创建配置 + config = RedBearModelConfig( + model_name=api_key_info.model_name, + provider=api_key_info.provider, + api_key=api_key_info.api_key, + base_url=api_key_info.api_base, + extra_params=api_key_info.config or {} + ) + + # 查询任务状态 + generator = RedBearVideoGenerator(config) + result = await generator.aget_task_status(task_id) + + return result diff --git a/api/app/services/home_page_service.py b/api/app/services/home_page_service.py index 8326ad40..4e6bf664 100644 --- a/api/app/services/home_page_service.py +++ b/api/app/services/home_page_service.py @@ -94,29 +94,38 @@ class HomePageService: @staticmethod def load_version_introduction(version: str) -> Dict[str, Any]: """ - 从 JSON 文件加载对应版本的介绍 + 加载对应版本的介绍(优先从数据库读取,fallback 到 JSON 文件) :param version: 系统版本号(如 "0.2.0") :return: 对应版本的详细介绍 """ - # 2. 定义 JSON 文件路径(简化路径处理,保留绝对路径调试特性) + from copy import deepcopy + from app.db import SessionLocal + from app.repositories.home_page_repository import HomePageRepository + + result = deepcopy(HomePageService.DEFAULT_RETURN_DATA) + + try: + db = SessionLocal() + try: + db_result = HomePageRepository.get_version_introduction(db, version) + if db_result: + return db_result + finally: + db.close() + except Exception as e: + pass + json_abs_path = Path(__file__).parent.parent / "version_info.json" json_abs_path = json_abs_path.resolve() - # 3. 初始化返回结果(深拷贝默认模板,避免修改原常量) - from copy import deepcopy - result = deepcopy(HomePageService.DEFAULT_RETURN_DATA) - try: - # 4. 简化文件存在性判断(合并逻辑,减少分支) if not json_abs_path.exists(): result["message"] = f"版本介绍文件不存在:{json_abs_path}" return result - # 5. 读取并解析 JSON 文件(简化文件操作流程) with open(json_abs_path, "r", encoding="utf-8") as f: changelogs = json.load(f) - # 6. 简化版本匹配逻辑,直接返回结果或更新提示信息 if version in changelogs: return changelogs[version] result["message"] = f"暂未查询到 {version} 版本的详细介绍" diff --git a/api/app/services/memory_agent_service.py b/api/app/services/memory_agent_service.py index 1e1d9e45..289fd74c 100644 --- a/api/app/services/memory_agent_service.py +++ b/api/app/services/memory_agent_service.py @@ -19,32 +19,35 @@ from typing import Any, AsyncGenerator, Dict, List, Optional from uuid import UUID import redis -from langchain_core.messages import AIMessage, HumanMessage +from langchain_core.messages import HumanMessage from pydantic import BaseModel, Field from sqlalchemy import func from sqlalchemy.orm import Session +from app.cache import InterestMemoryCache from app.core.config import settings from app.core.logging_config import get_config_logger, get_logger from app.core.memory.agent.langgraph_graph.read_graph import make_read_graph -from app.core.memory.agent.langgraph_graph.write_graph import make_write_graph from app.core.memory.agent.logger_file.log_streamer import LogStreamer from app.core.memory.agent.utils.messages_tools import ( merge_multiple_search_results, reorder_output_results, ) from app.core.memory.agent.utils.type_classifier import status_typle +from app.core.memory.agent.utils.write_tools import write as write_neo4j from app.core.memory.analytics.hot_memory_tags import get_interest_distribution from app.core.memory.utils.llm.llm_utils import MemoryClientFactory from app.db import get_db_context from app.models.knowledge_model import Knowledge, KnowledgeType from app.repositories.neo4j.neo4j_connector import Neo4jConnector +from app.schemas import FileInput from app.schemas.memory_agent_schema import Write_UserInput from app.schemas.memory_config_schema import ConfigurationError from app.services.memory_config_service import MemoryConfigService from app.services.memory_konwledges_server import ( write_rag, ) +from app.services.memory_perceptual_service import MemoryPerceptualService try: from app.core.memory.utils.log.audit_logger import audit_logger @@ -267,8 +270,16 @@ class MemoryAgentService: logger.info("Log streaming completed, cleaning up resources") # LogStreamer uses context manager for file handling, so cleanup is automatic - async def write_memory(self, end_user_id: str, messages: list[dict], config_id: Optional[uuid.UUID] | int, - db: Session, storage_type: str, user_rag_memory_id: str, language: str = "zh") -> str: + async def write_memory( + self, + end_user_id: str, + messages: list[dict], + config_id: Optional[uuid.UUID] | int, + db: Session, + storage_type: str, + user_rag_memory_id: str, + language: str = "zh" + ) -> str: """ Process write operation with config_id @@ -297,8 +308,8 @@ class MemoryAgentService: config_id = connected_config.get("memory_config_id") logger.info(f"Resolved config from end_user: config_id={config_id}, workspace_id={workspace_id}") if config_id is None and workspace_id is None: - raise ValueError( - f"No memory configuration found for end_user {end_user_id}. Please ensure the user has a connected memory configuration.") + raise ValueError(f"No memory configuration found for end_user {end_user_id}. " + f"Please ensure the user has a connected memory configuration.") except Exception as e: if "No memory configuration found" in str(e): raise # Re-raise our specific error @@ -334,48 +345,58 @@ class MemoryAgentService: raise ValueError(error_msg) + perceptual_serivce = MemoryPerceptualService(db) + for message in messages: + message["file_content"] = [] + for file in (message.get("files") or []): + file_object = await perceptual_serivce.generate_perceptual_memory( + end_user_id=end_user_id, + memory_config=memory_config, + file=FileInput(**file) + ) + if file_object is None: + continue + message["file_content"].append((file_object, file["type"])) + logger.info(messages) + + message_text = "\n".join([f"{msg['role']}: {msg['content']}" for msg in messages]) try: if storage_type == "rag": # For RAG storage, convert messages to single string - message_text = "\n".join([f"{msg['role']}: {msg['content']}" for msg in messages]) - result = await write_rag(end_user_id, message_text, user_rag_memory_id) - return result + await write_rag(end_user_id, message_text, user_rag_memory_id) + return "success" else: - async with make_write_graph() as graph: - config = {"configurable": {"thread_id": end_user_id}} - # Convert structured messages to LangChain messages - langchain_messages = [] - for msg in messages: - if msg['role'] == 'user': - langchain_messages.append(HumanMessage(content=msg['content'])) - elif msg['role'] == 'assistant': - langchain_messages.append(AIMessage(content=msg['content'])) - print(100 * '-') - print(langchain_messages) - print(100 * '-') - # 初始状态 - 包含所有必要字段 - initial_state = { - "messages": langchain_messages, - "end_user_id": end_user_id, - "memory_config": memory_config, - "language": language + await write_neo4j( + end_user_id=end_user_id, + messages=messages, + memory_config=memory_config, + ref_id='', + language=language + ) + for lang in ["zh", "en"]: + deleted = await InterestMemoryCache.delete_interest_distribution( + end_user_id, lang + ) + if deleted: + logger.info( + f"Invalidated interest distribution cache: end_user_id={end_user_id}, language={lang}") + for message in messages: + message["file_content"] = [ + perceptual[0].file_path for perceptual in message["file_content"] + ] + return self.writer_messages_deal( + "success", + start_time, + end_user_id, + config_id, + message_text, + { + "status": "success", + "data": messages, + "config_id": memory_config.config_id, + "config_name": memory_config.config_name } - - # 获取节点更新信息 - async for update_event in graph.astream( - initial_state, - stream_mode="updates", - config=config - ): - for node_name, node_data in update_event.items(): - if 'save_neo4j' == node_name: - massages = node_data - massagesstatus = massages.get('write_result')['status'] - contents = massages.get('write_result') - # Convert messages back to string for logging - message_text = "\n".join([f"{msg['role']}: {msg['content']}" for msg in messages]) - return self.writer_messages_deal(massagesstatus, start_time, end_user_id, config_id, message_text, - contents) + ) except Exception as e: # Ensure proper error handling and logging error_msg = f"Write operation failed: {str(e)}" @@ -586,7 +607,7 @@ class MemoryAgentService: retrieved_content.append({query: statements}) # 如果 retrieved_content 为空,设置为空字符串 - if retrieved_content == []: + if not retrieved_content: retrieved_content = '' # 只有当回答不是"信息不足"且不是快速检索时才保存 @@ -1179,7 +1200,7 @@ def get_end_user_connected_config(end_user_id: str, db: Session) -> Dict[str, An app = db.query(App).filter(App.id == app_id).first() if not app: logger.warning(f"App not found: {app_id}") - raise ValueError(f"应用不存在: {app_id}") + # raise ValueError(f"应用不存在: {app_id}") # TODO: temp fix for draft run # if not app.current_release_id: # logger.warning(f"No current release for app: {app_id}") @@ -1252,17 +1273,15 @@ def get_end_user_connected_config(end_user_id: str, db: Session) -> Dict[str, An memory_config_service = MemoryConfigService(db) memory_config = memory_config_service.get_config_with_fallback( memory_config_id=memory_config_id_to_use, - workspace_id=app.workspace_id + workspace_id=end_user.workspace_id ) memory_config_id = str(memory_config.config_id) if memory_config else None result = { "end_user_id": str(end_user_id), - "app_id": str(app_id), - "release_id": str(app.current_release_id) if app.current_release_id else None, "memory_config_id": memory_config_id, - "workspace_id": str(app.workspace_id) + "workspace_id": str(end_user.workspace_id) } logger.info( diff --git a/api/app/services/memory_api_service.py b/api/app/services/memory_api_service.py index f86fbed8..9282fc28 100644 --- a/api/app/services/memory_api_service.py +++ b/api/app/services/memory_api_service.py @@ -28,7 +28,7 @@ class MemoryAPIService: 2. Maps end_user_id to end_user_id for memory operations 3. Delegates to MemoryAgentService for actual memory read/write operations """ - + def __init__(self, db: Session): """Initialize MemoryAPIService. @@ -36,11 +36,11 @@ class MemoryAPIService: db: SQLAlchemy database session """ self.db = db - + def validate_end_user( - self, - end_user_id: str, - workspace_id: uuid.UUID + self, + end_user_id: str, + workspace_id: uuid.UUID ) -> EndUser: """Validate that end_user exists and belongs to the workspace. @@ -56,7 +56,7 @@ class MemoryAPIService: BusinessException: If end_user not in authorized workspace """ logger.info(f"Validating end_user: {end_user_id} for workspace: {workspace_id}") - + # Query end_user by ID try: end_user_uuid = uuid.UUID(end_user_id) @@ -66,7 +66,7 @@ class MemoryAPIService: message=f"Invalid end_user_id format: {end_user_id}", code=BizCode.INVALID_PARAMETER ) - + end_user = self.db.query(EndUser).filter(EndUser.id == end_user_uuid).first() if not end_user: @@ -75,52 +75,74 @@ class MemoryAPIService: resource_type="EndUser", resource_id=end_user_id ) - + # Verify end_user belongs to the workspace via App relationship app = self.db.query(App).filter( App.id == end_user.app_id, App.is_active.is_(True) ).first() - + if not app: logger.warning(f"App not found for end_user: {end_user_id}") - raise ResourceNotFoundException( - resource_type="App", - resource_id=str(end_user.app_id) - ) - - if app.workspace_id != workspace_id: - logger.warning( - f"End user {end_user_id} belongs to workspace {app.workspace_id}, " - f"not authorized workspace {workspace_id}" - ) - raise BusinessException( - message="End user does not belong to authorized workspace", - code=BizCode.FORBIDDEN - ) - + # raise ResourceNotFoundException( + # resource_type="App", + # resource_id=str(end_user.app_id) + # ) + # temporally allow any workspace to access + # if end_user.workspace_id != workspace_id: + # print(f"[DEBUG] end_user.workspace_id={end_user.workspace_id}, api_key.workspace_id={workspace_id}") + # logger.warning( + # f"End user {end_user_id} belongs to workspace {end_user.workspace_id}, " + # f"not authorized workspace {workspace_id}" + # ) + # raise BusinessException( + # message=f"End user does not belong to authorized workspace. end_user.workspace_id={end_user.workspace_id}, api_key.workspace_id={workspace_id}", + # code=BizCode.FORBIDDEN + # ) + logger.info(f"End user {end_user_id} validated successfully") return end_user - + + def _update_end_user_config(self, end_user_id: str, config_id: str) -> None: + """Update the end user's memory_config_id. + + Silently updates the config association. Logs warnings on failure + but does not raise, so it won't block the main read/write operation. + + Args: + end_user_id: End user identifier + config_id: Memory configuration ID to assign + """ + try: + config_uuid = uuid.UUID(config_id) + from app.repositories.end_user_repository import EndUserRepository + end_user_repo = EndUserRepository(self.db) + end_user_repo.update_memory_config_id( + end_user_id=uuid.UUID(end_user_id), + memory_config_id=config_uuid, + ) + except Exception as e: + logger.warning(f"Failed to update memory_config_id for end_user {end_user_id}: {e}") + async def write_memory( - self, - workspace_id: uuid.UUID, - end_user_id: str, - message: str, - config_id: Optional[str] = None, - storage_type: str = "neo4j", - user_rag_memory_id: Optional[str] = None, + self, + workspace_id: uuid.UUID, + end_user_id: str, + message: str, + config_id: str, + storage_type: str = "neo4j", + user_rag_memory_id: Optional[str] = None, ) -> Dict[str, Any]: """Write memory with validation. - Validates end_user exists and belongs to workspace, then delegates - to MemoryAgentService.write_memory. + Validates end_user exists and belongs to workspace, updates the end user's + memory_config_id, then delegates to MemoryAgentService.write_memory. Args: workspace_id: Workspace ID for resource validation end_user_id: End user identifier (used as end_user_id) message: Message content to store - config_id: Optional memory configuration ID + config_id: Memory configuration ID (required) storage_type: Storage backend (neo4j or rag) user_rag_memory_id: Optional RAG memory ID @@ -132,12 +154,13 @@ class MemoryAPIService: BusinessException: If end_user not in authorized workspace or write fails """ logger.info(f"Writing memory for end_user: {end_user_id}, workspace: {workspace_id}") - + # Validate end_user exists and belongs to workspace self.validate_end_user(end_user_id, workspace_id) - - # Use end_user_id as end_user_id for memory operations - + + # Update end user's memory_config_id + self._update_end_user_config(end_user_id, config_id) + try: # Delegate to MemoryAgentService # Convert string message to list[dict] format expected by MemoryAgentService @@ -148,11 +171,11 @@ class MemoryAPIService: config_id=config_id, db=self.db, storage_type=storage_type, - user_rag_memory_id=user_rag_memory_id or "" + user_rag_memory_id=user_rag_memory_id or "", ) - + logger.info(f"Memory write successful for end_user: {end_user_id}") - + # result may be a string "success" or a dict with a "status" key # Preserve the full dict so callers don't silently lose extra fields # (e.g. error codes, metadata) returned by MemoryAgentService. @@ -166,7 +189,7 @@ class MemoryAPIService: "status": result if isinstance(result, str) else "success", "end_user_id": end_user_id, } - + except ConfigurationError as e: logger.error(f"Memory configuration error for end_user {end_user_id}: {e}") raise BusinessException( @@ -181,28 +204,28 @@ class MemoryAPIService: message=f"Memory write failed: {str(e)}", code=BizCode.MEMORY_WRITE_FAILED ) - + async def read_memory( - self, - workspace_id: uuid.UUID, - end_user_id: str, - message: str, - search_switch: str = "0", - config_id: Optional[str] = None, - storage_type: str = "neo4j", - user_rag_memory_id: Optional[str] = None, + self, + workspace_id: uuid.UUID, + end_user_id: str, + message: str, + search_switch: str = "0", + config_id: str = "", + storage_type: str = "neo4j", + user_rag_memory_id: Optional[str] = None, ) -> Dict[str, Any]: """Read memory with validation. - Validates end_user exists and belongs to workspace, then delegates - to MemoryAgentService.read_memory. + Validates end_user exists and belongs to workspace, updates the end user's + memory_config_id, then delegates to MemoryAgentService.read_memory. Args: workspace_id: Workspace ID for resource validation end_user_id: End user identifier (used as end_user_id) message: Query message search_switch: Search mode (0=deep search with verification, 1=deep search, 2=fast search) - config_id: Optional memory configuration ID + config_id: Memory configuration ID (required) storage_type: Storage backend (neo4j or rag) user_rag_memory_id: Optional RAG memory ID @@ -214,13 +237,13 @@ class MemoryAPIService: BusinessException: If end_user not in authorized workspace or read fails """ logger.info(f"Reading memory for end_user: {end_user_id}, workspace: {workspace_id}") - + # Validate end_user exists and belongs to workspace self.validate_end_user(end_user_id, workspace_id) - - # Use end_user_id as end_user_id for memory operations - + # Update end user's memory_config_id + self._update_end_user_config(end_user_id, config_id) + try: # Delegate to MemoryAgentService result = await MemoryAgentService().read_memory( @@ -233,15 +256,15 @@ class MemoryAPIService: storage_type=storage_type, user_rag_memory_id=user_rag_memory_id or "" ) - + logger.info(f"Memory read successful for end_user: {end_user_id}") - + return { "answer": result.get("answer", ""), "intermediate_outputs": result.get("intermediate_outputs", []), "end_user_id": end_user_id } - + except ConfigurationError as e: logger.error(f"Memory configuration error for end_user {end_user_id}: {e}") raise BusinessException( @@ -256,3 +279,50 @@ class MemoryAPIService: message=f"Memory read failed: {str(e)}", code=BizCode.MEMORY_READ_FAILED ) + + def list_memory_configs( + self, + workspace_id: uuid.UUID, + ) -> Dict[str, Any]: + """List all memory configs for a workspace. + + Args: + workspace_id: Workspace ID from API key authorization + + Returns: + Dict with configs list and total count + + Raises: + BusinessException: If listing fails + """ + logger.info(f"Listing memory configs for workspace: {workspace_id}") + + try: + from app.repositories.memory_config_repository import MemoryConfigRepository + + results = MemoryConfigRepository.get_all(self.db, workspace_id=workspace_id) + + configs = [] + for config, scene_name in results: + configs.append({ + "config_id": str(config.config_id), + "config_name": config.config_name, + "config_desc": config.config_desc, + "is_default": config.is_default or False, + "scene_name": scene_name, + "created_at": config.created_at.isoformat() if config.created_at else None, + "updated_at": config.updated_at.isoformat() if config.updated_at else None, + }) + + logger.info(f"Found {len(configs)} memory configs for workspace {workspace_id}") + return { + "configs": configs, + "total": len(configs), + } + + except Exception as e: + logger.error(f"Failed to list memory configs for workspace {workspace_id}: {e}") + raise BusinessException( + message=f"Failed to list memory configs: {str(e)}", + code=BizCode.MEMORY_READ_FAILED + ) diff --git a/api/app/services/memory_config_service.py b/api/app/services/memory_config_service.py index 4d67673f..66c110b1 100644 --- a/api/app/services/memory_config_service.py +++ b/api/app/services/memory_config_service.py @@ -37,7 +37,7 @@ def _validate_config_id(config_id, db: Session = None): """Validate configuration ID format (supports both UUID and integer).""" if isinstance(config_id, uuid.UUID): return config_id - + if config_id is None: raise InvalidConfigError( "Configuration ID cannot be None", @@ -52,26 +52,30 @@ def _validate_config_id(config_id, db: Session = None): field_name="config_id", invalid_value=config_id, ) - # 如果提供了数据库会话,尝试通过 user_id 查询 config_id + # 如果提供了数据库会话,尝试通过 config_id_old 查询 config_id if db is not None: - # 查询 user_id 匹配的记录 - stmt = select(MemoryConfigModel).where(MemoryConfigModel.config_id_old == str(config_id)) + # 查询 config_id_old 匹配的记录 + stmt = select(MemoryConfigModel).where(MemoryConfigModel.config_id_old == config_id) result = db.execute(stmt).scalars().first() if result: - logger.info(f"Found config_id {result.config_id} for user_id {config_id}") + logger.info(f"Found config_id {result.config_id} for config_id_old {config_id}") return result.config_id - return config_id + raise InvalidConfigError( + f"未找到 config_id_old={config_id} 对应的配置", + field_name="config_id", + invalid_value=config_id, + ) if isinstance(config_id, str): config_id_stripped = config_id.strip() - + # Try parsing as UUID first try: return uuid.UUID(config_id_stripped) except ValueError: pass - + # Fall back to integer parsing try: parsed_id = int(config_id_stripped) @@ -81,18 +85,22 @@ def _validate_config_id(config_id, db: Session = None): field_name="config_id", invalid_value=config_id, ) - + # 如果提供了数据库会话,尝试通过 user_id 查询 config_id if db is not None: - # 查询 user_id 匹配的记录 - stmt = select(MemoryConfigModel).where(MemoryConfigModel.user_id == str(parsed_id)) + # 查询 config_id_old 匹配的记录 + stmt = select(MemoryConfigModel).where(MemoryConfigModel.config_id_old == parsed_id) result = db.execute(stmt).scalars().first() - + if result: - logger.info(f"Found config_id {result.config_id} for user_id {parsed_id}") + logger.info(f"Found config_id {result.config_id} for config_id_old {parsed_id}") return result.config_id - return parsed_id + raise InvalidConfigError( + f"未找到 config_id_old={parsed_id} 对应的配置", + field_name="config_id", + invalid_value=config_id, + ) except ValueError: raise InvalidConfigError( f"Invalid configuration ID format: '{config_id}' (must be UUID or positive integer)", @@ -107,28 +115,29 @@ def _validate_config_id(config_id, db: Session = None): ) -def _load_ontology_classes(db: Session, scene_id, pruning_scene: Optional[str]) -> Optional[list]: - """从 ontology_class 表加载场景类型名称列表,用于注入提示词。 +def _load_ontology_class_infos(db: Session, scene_id) -> list: + """从 ontology_class 表加载完整本体类型信息(name + description),用于注入剪枝提示词。 Args: db: 数据库会话 scene_id: 本体场景 UUID - pruning_scene: 语义剪枝场景名称(保留参数,暂未使用) Returns: - class_name 字符串列表,或 None(无数据时) + [{"class_name": ..., "class_description": ...}, ...] 或空列表 """ if not scene_id: - return None + return [] try: from app.repositories.ontology_class_repository import OntologyClassRepository repo = OntologyClassRepository(db) classes = repo.get_classes_by_scene(scene_id) - names = [c.class_name for c in classes if c.class_name] - return names if names else None + return [ + {"class_name": c.class_name, "class_description": c.class_description or ""} + for c in classes if c.class_name + ] except Exception as e: - logger.warning(f"Failed to load ontology classes for scene_id={scene_id}: {e}") - return None + logger.warning(f"Failed to load ontology class infos for scene_id={scene_id}: {e}") + return [] class MemoryConfigService: @@ -153,10 +162,10 @@ class MemoryConfigService: self.db = db def load_memory_config( - self, - config_id: Optional[UUID] = None, - workspace_id: Optional[UUID] = None, - service_name: str = "MemoryConfigService", + self, + config_id: Optional[UUID] = None, + workspace_id: Optional[UUID] = None, + service_name: str = "MemoryConfigService", ) -> MemoryConfig: """ Load memory configuration from database with optional fallback. @@ -193,14 +202,14 @@ class MemoryConfigService: try: # Use get_config_with_fallback if workspace_id is provided memory_config = None + validated_config_id = None if workspace_id: - validated_config_id = None if config_id: try: validated_config_id = _validate_config_id(config_id, self.db) except Exception: validated_config_id = None - + memory_config = self.get_config_with_fallback( memory_config_id=validated_config_id, workspace_id=workspace_id @@ -209,7 +218,7 @@ class MemoryConfigService: validated_config_id = _validate_config_id(config_id, self.db) from app.models.memory_config_model import MemoryConfig as MemoryConfigModel memory_config = self.db.get(MemoryConfigModel, validated_config_id) - + if not memory_config: elapsed_ms = (time.time() - start_time) * 1000 config_logger.error( @@ -232,7 +241,7 @@ class MemoryConfigService: result = MemoryConfigRepository.get_config_with_workspace(self.db, memory_config.config_id) db_query_time = time.time() - db_query_start logger.info(f"[PERF] Config+Workspace query: {db_query_time:.4f}s") - + if not result: raise ConfigurationError( f"Workspace not found for config {memory_config.config_id}" @@ -242,10 +251,10 @@ class MemoryConfigService: # Helper function to validate model with workspace fallback def _validate_model_with_fallback( - model_id: str, - model_type: str, - workspace_default: str, - required: bool = False + model_id: str, + model_type: str, + workspace_default: str, + required: bool = False ) -> tuple: """Validate model ID, falling back to workspace default if invalid. @@ -274,7 +283,7 @@ class MemoryConfigService: logger.warning( f"{model_type} model validation failed, trying workspace default: {e}" ) - + # Fallback to workspace default if workspace_default: try: @@ -296,7 +305,7 @@ class MemoryConfigService: logger.error(f"Workspace default {model_type} model also invalid: {e}") if required: raise - + if required: raise InvalidConfigError( f"{model_type.title()} model is required but not configured", @@ -305,7 +314,7 @@ class MemoryConfigService: config_id=validated_config_id, workspace_id=workspace.id ) - + return None, None # Step 2: Validate embedding model with workspace fallback @@ -342,6 +351,35 @@ class MemoryConfigService: if memory_config.rerank_id or workspace.rerank: logger.info(f"[PERF] Rerank validation: {rerank_time:.4f}s") + vision_uuid, vision_name = validate_and_resolve_model_id( + memory_config.vision_id, + "llm", + self.db, + workspace.tenant_id, + required=False, + config_id=validated_config_id, + workspace_id=workspace.id, + ) + + audio_uuid, audio_name = validate_and_resolve_model_id( + memory_config.audio_id, + "llm", + self.db, + workspace.tenant_id, + required=False, + config_id=validated_config_id, + workspace_id=workspace.id, + ) + + video_uuid, video_name = validate_and_resolve_model_id( + memory_config.video_id, + "llm", + self.db, + workspace.tenant_id, + required=False, + config_id=validated_config_id, + workspace_id=workspace.id, + ) # Create immutable MemoryConfig object config = MemoryConfig( config_id=memory_config.config_id, @@ -355,6 +393,12 @@ class MemoryConfigService: embedding_model_name=embedding_name, rerank_model_id=rerank_uuid, rerank_model_name=rerank_name, + video_model_id=video_uuid, + video_model_name=video_name, + vision_model_id=vision_uuid, + vision_model_name=vision_name, + audio_model_id=audio_uuid, + audio_model_name=audio_name, storage_type=workspace.storage_type or "neo4j", chunker_strategy=memory_config.chunker_strategy or "RecursiveChunker", reflexion_enabled=memory_config.enable_self_reflexion or False, @@ -363,27 +407,34 @@ class MemoryConfigService: reflexion_baseline=memory_config.baseline or "Time", loaded_at=datetime.now(), # Pipeline config: Deduplication - enable_llm_dedup_blockwise=bool(memory_config.enable_llm_dedup_blockwise) if memory_config.enable_llm_dedup_blockwise is not None else False, - enable_llm_disambiguation=bool(memory_config.enable_llm_disambiguation) if memory_config.enable_llm_disambiguation is not None else False, + enable_llm_dedup_blockwise=bool( + memory_config.enable_llm_dedup_blockwise) if memory_config.enable_llm_dedup_blockwise is not None else False, + enable_llm_disambiguation=bool( + memory_config.enable_llm_disambiguation) if memory_config.enable_llm_disambiguation is not None else False, deep_retrieval=bool(memory_config.deep_retrieval) if memory_config.deep_retrieval is not None else True, t_type_strict=float(memory_config.t_type_strict) if memory_config.t_type_strict is not None else 0.8, t_name_strict=float(memory_config.t_name_strict) if memory_config.t_name_strict is not None else 0.8, t_overall=float(memory_config.t_overall) if memory_config.t_overall is not None else 0.8, # Pipeline config: Statement extraction - statement_granularity=int(memory_config.statement_granularity) if memory_config.statement_granularity is not None else 2, - include_dialogue_context=bool(memory_config.include_dialogue_context) if memory_config.include_dialogue_context is not None else False, - max_dialogue_context_chars=int(memory_config.max_context) if memory_config.max_context is not None else 1000, + statement_granularity=int( + memory_config.statement_granularity) if memory_config.statement_granularity is not None else 2, + include_dialogue_context=bool( + memory_config.include_dialogue_context) if memory_config.include_dialogue_context is not None else False, + max_dialogue_context_chars=int( + memory_config.max_context) if memory_config.max_context is not None else 1000, # Pipeline config: Forgetting engine lambda_time=float(memory_config.lambda_time) if memory_config.lambda_time is not None else 0.5, lambda_mem=float(memory_config.lambda_mem) if memory_config.lambda_mem is not None else 0.5, offset=float(memory_config.offset) if memory_config.offset is not None else 0.0, # Pipeline config: Pruning - pruning_enabled=bool(memory_config.pruning_enabled) if memory_config.pruning_enabled is not None else False, + pruning_enabled=bool( + memory_config.pruning_enabled) if memory_config.pruning_enabled is not None else False, pruning_scene=memory_config.pruning_scene or "education", - pruning_threshold=float(memory_config.pruning_threshold) if memory_config.pruning_threshold is not None else 0.5, + pruning_threshold=float( + memory_config.pruning_threshold) if memory_config.pruning_threshold is not None else 0.5, # Ontology scene association scene_id=memory_config.scene_id, - ontology_classes=_load_ontology_classes(self.db, memory_config.scene_id, memory_config.pruning_scene), + ontology_class_infos=_load_ontology_class_infos(self.db, memory_config.scene_id), ) elapsed_ms = (time.time() - start_time) * 1000 @@ -447,9 +498,9 @@ class MemoryConfigService: if not config: logger.warning(f"Model ID {model_id} not found") raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="模型ID不存在") - + api_config: ModelApiKey = config.api_keys[0] - + return { "model_name": api_config.model_name, "provider": api_config.provider, @@ -480,9 +531,9 @@ class MemoryConfigService: if not config: logger.warning(f"Embedding model ID {embedding_id} not found") raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail="嵌入模型ID不存在") - + api_config: ModelApiKey = config.api_keys[0] - + return { "model_name": api_config.model_name, "provider": api_config.provider, @@ -550,11 +601,13 @@ class MemoryConfigService: - pruning_switch: bool - pruning_scene: str - pruning_threshold: float + - ontology_class_infos: list of {class_name, class_description} dicts """ return { "pruning_switch": memory_config.pruning_enabled, "pruning_scene": memory_config.pruning_scene, "pruning_threshold": memory_config.pruning_threshold, + "ontology_class_infos": memory_config.ontology_class_infos or [], } def get_ontology_types(self, memory_config: MemoryConfig): @@ -568,25 +621,25 @@ class MemoryConfigService: """ from app.core.memory.models.ontology_extraction_models import OntologyTypeList from app.repositories.ontology_class_repository import OntologyClassRepository - + if not memory_config.scene_id: logger.debug("No scene_id configured, skipping ontology type fetch") return None - + try: ontology_repo = OntologyClassRepository(self.db) ontology_classes = ontology_repo.get_classes_by_scene(memory_config.scene_id) - + if not ontology_classes: logger.info(f"No ontology classes found for scene_id: {memory_config.scene_id}") return None - + ontology_types = OntologyTypeList.from_db_models(ontology_classes) logger.info( f"Loaded {len(ontology_types.types)} ontology types for scene_id: {memory_config.scene_id}" ) return ontology_types - + except Exception as e: logger.warning( f"Failed to fetch ontology types for scene_id {memory_config.scene_id}: {e}", @@ -595,8 +648,8 @@ class MemoryConfigService: return None def get_workspace_default_config( - self, - workspace_id: UUID + self, + workspace_id: UUID ) -> Optional["MemoryConfigModel"]: """Get workspace default memory config. @@ -610,19 +663,19 @@ class MemoryConfigService: Optional[MemoryConfigModel]: Default config or None if no configs exist """ config = MemoryConfigRepository.get_workspace_default(self.db, workspace_id) - + if not config: logger.warning( "No active memory config found for workspace fallback", extra={"workspace_id": str(workspace_id)} ) - + return config def get_config_with_fallback( - self, - memory_config_id: Optional[UUID], - workspace_id: UUID + self, + memory_config_id: Optional[UUID], + workspace_id: UUID ) -> Optional["MemoryConfigModel"]: """Get memory config with fallback to workspace default. @@ -641,13 +694,13 @@ class MemoryConfigService: "No memory config ID provided, using workspace default", extra={"workspace_id": str(workspace_id)} ) - + config = MemoryConfigRepository.get_with_fallback( self.db, memory_config_id, workspace_id ) - + if not config and memory_config_id: logger.warning( "Memory config not found, falling back to workspace default", @@ -656,13 +709,13 @@ class MemoryConfigService: "workspace_id": str(workspace_id) } ) - + return config def delete_config( - self, - config_id: UUID | int, - force: bool = False + self, + config_id: UUID | int, + force: bool = False ) -> dict: """Delete memory config with protection against in-use configs. @@ -684,7 +737,7 @@ class MemoryConfigService: from app.core.exceptions import ResourceNotFoundException from app.models.memory_config_model import MemoryConfig as MemoryConfigModel from app.repositories.end_user_repository import EndUserRepository - + # 处理旧格式 int 类型的 config_id if isinstance(config_id, int): logger.warning( @@ -696,11 +749,11 @@ class MemoryConfigService: "message": "旧格式配置ID不支持删除操作,请使用新版配置", "legacy_int_id": config_id } - + config = self.db.get(MemoryConfigModel, config_id) if not config: raise ResourceNotFoundException("MemoryConfig", str(config_id)) - + # Check if this is the default config - default configs cannot be deleted if config.is_default: logger.warning( @@ -712,11 +765,11 @@ class MemoryConfigService: "message": "默认配置不允许删除", "is_default": True } - + # Use repository to count connected end users end_user_repo = EndUserRepository(self.db) connected_count = end_user_repo.count_by_memory_config_id(config_id) - + if connected_count > 0 and not force: logger.warning( "Attempted to delete memory config with connected end users", @@ -725,18 +778,18 @@ class MemoryConfigService: "connected_count": connected_count } ) - + return { "status": "warning", "message": f"无法删除记忆配置:{connected_count} 个终端用户正在使用此配置", "connected_count": connected_count, "force_required": True } - + # Force delete: use repository to clear end user references first if connected_count > 0 and force: cleared_count = end_user_repo.clear_memory_config_id(config_id) - + logger.warning( "Force deleting memory config, clearing end user references", extra={ @@ -744,11 +797,11 @@ class MemoryConfigService: "cleared_end_users": cleared_count } ) - + try: self.db.delete(config) self.db.commit() - + logger.info( "Memory config deleted", extra={ @@ -757,16 +810,16 @@ class MemoryConfigService: "affected_users": connected_count } ) - + return { "status": "success", "message": "记忆配置删除成功", "affected_users": connected_count } - + except IntegrityError as e: self.db.rollback() - + # Handle foreign key violation gracefully error_str = str(e.orig) if e.orig else str(e) if "ForeignKeyViolation" in error_str or "foreign key constraint" in error_str.lower(): @@ -782,7 +835,7 @@ class MemoryConfigService: "message": "无法删除记忆配置:仍有终端用户引用此配置,请使用 force=true 强制删除", "force_required": True } - + # Re-raise other integrity errors logger.error( "Delete failed due to integrity error", @@ -797,9 +850,9 @@ class MemoryConfigService: # ==================== 记忆配置提取方法 ==================== def extract_memory_config_id( - self, - app_type: str, - config: dict + self, + app_type: str, + config: dict ) -> tuple[Optional[uuid.UUID], bool]: """从发布配置中提取 memory_config_id(根据应用类型分发) @@ -824,9 +877,26 @@ class MemoryConfigService: logger.warning(f"不支持的应用类型,无法提取记忆配置: app_type={app_type}") return None, False + def _resolve_config_id_old(self, config_id_old: int) -> Optional[uuid.UUID]: + """通过 config_id_old 查询对应的 UUID config_id。 + + Args: + config_id_old: 旧格式的整数配置ID + + Returns: + 对应的 UUID config_id,未找到返回 None + """ + from app.models.memory_config_model import MemoryConfig as MemoryConfigModel + result = self.db.query(MemoryConfigModel).filter( + MemoryConfigModel.config_id_old == config_id_old + ).first() + if result: + return result.config_id + return None + def _extract_memory_config_id_from_agent( - self, - config: dict + self, + config: dict ) -> tuple[Optional[uuid.UUID], bool]: """从 Agent 应用配置中提取 memory_config_id @@ -855,10 +925,11 @@ class MemoryConfigService: elif isinstance(memory_value, str): # Check if it's a numeric string (legacy int format) if memory_value.isdigit(): - logger.warning( - f"Agent 配置中 memory_config_id 为旧格式 int 字符串,将使用工作空间默认配置: " - f"value={memory_value}" - ) + resolved = self._resolve_config_id_old(int(memory_value)) + if resolved: + logger.info(f"Resolved legacy config_id_old={memory_value} to config_id={resolved}") + return resolved, False + logger.warning(f"未找到 config_id_old={memory_value} 对应的配置,将使用工作空间默认配置") return None, True try: return uuid.UUID(memory_value), False @@ -866,11 +937,11 @@ class MemoryConfigService: logger.warning(f"Invalid UUID string: {memory_value}") return None, False elif isinstance(memory_value, int): - # 旧数据存储为 int,需要回退到工作空间默认配置 - logger.warning( - f"Agent 配置中 memory_config_id 为旧格式 int,将使用工作空间默认配置: " - f"value={memory_value}" - ) + resolved = self._resolve_config_id_old(memory_value) + if resolved: + logger.info(f"Resolved legacy config_id_old={memory_value} to config_id={resolved}") + return resolved, False + logger.warning(f"未找到 config_id_old={memory_value} 对应的配置,将使用工作空间默认配置") return None, True else: logger.warning( @@ -885,8 +956,8 @@ class MemoryConfigService: return None, False def _extract_memory_config_id_from_workflow( - self, - config: dict + self, + config: dict ) -> tuple[Optional[uuid.UUID], bool]: """从 Workflow 应用配置中提取 memory_config_id @@ -902,14 +973,14 @@ class MemoryConfigService: - is_legacy_int: 是否检测到旧格式 int 数据 """ nodes = config.get("nodes", []) - + for node in nodes: node_type = node.get("type", "") - + # 检查是否为记忆节点 (support both formats: memory-read/memory-write and MemoryRead/MemoryWrite) if node_type.lower() in ["memoryread", "memorywrite", "memory-read", "memory-write"]: config_id = node.get("config", {}).get("config_id") - + if config_id: try: # 处理字符串、UUID 和 int(旧数据兼容)三种情况 @@ -918,10 +989,16 @@ class MemoryConfigService: elif isinstance(config_id, str): return uuid.UUID(config_id), False elif isinstance(config_id, int): - # 旧数据存储为 int,需要回退到工作空间默认配置 + resolved = self._resolve_config_id_old(config_id) + if resolved: + logger.info( + f"Resolved workflow legacy config_id_old={config_id} to config_id={resolved}: " + f"node_id={node.get('id')}, node_type={node_type}" + ) + return resolved, False logger.warning( - f"工作流记忆节点 config_id 为旧格式 int,将使用工作空间默认配置: " - f"node_id={node.get('id')}, node_type={node_type}, value={config_id}" + f"未找到工作流记忆节点 config_id_old={config_id} 对应的配置,将使用工作空间默认配置: " + f"node_id={node.get('id')}, node_type={node_type}" ) return None, True else: @@ -934,6 +1011,6 @@ class MemoryConfigService: f"工作流记忆节点 config_id 格式无效: node_id={node.get('id')}, " f"node_type={node_type}, error={str(e)}" ) - + logger.debug("工作流配置中未找到记忆节点") return None, False diff --git a/api/app/services/memory_forget_service.py b/api/app/services/memory_forget_service.py index a0bcc1a1..2d91f025 100644 --- a/api/app/services/memory_forget_service.py +++ b/api/app/services/memory_forget_service.py @@ -204,30 +204,35 @@ class MemoryForgetService: end_user_id: str, forgetting_threshold: float, min_days_since_access: int, - limit: int = 20 - ) -> list[Dict[str, Any]]: + page: Optional[int] = None, + pagesize: Optional[int] = None + ) -> Dict[str, Any]: """ 获取待遗忘节点列表 - - 查询满足遗忘条件的节点(激活值低于阈值且最后访问时间超过最小天数) - + + 查询满足遗忘条件的节点(激活值低于阈值且最后访问时间超过最小天数)。支持分页查询。 + Args: connector: Neo4j 连接器 end_user_id: 组ID forgetting_threshold: 遗忘阈值 min_days_since_access: 最小未访问天数 - limit: 返回节点数量限制 - + page: 页码(可选,从1开始) + pagesize: 每页数量(可选) + Returns: - list: 待遗忘节点列表 + dict: 包含待遗忘节点列表和分页信息的字典 + - items: 待遗忘节点列表 + - page: 分页信息(分页时) """ from datetime import timedelta - + # 计算最小访问时间(ISO 8601 格式字符串,使用 UTC 时区) min_access_time = datetime.now(timezone.utc) - timedelta(days=min_days_since_access) min_access_time_str = min_access_time.strftime('%Y-%m-%dT%H:%M:%S.%fZ') - - query = """ + + # 基础查询(用于获取总数) + count_query = """ MATCH (n) WHERE (n:Statement OR n:ExtractedEntity OR n:MemorySummary) AND n.end_user_id = $end_user_id @@ -235,10 +240,22 @@ class MemoryForgetService: AND n.activation_value < $threshold AND n.last_access_time IS NOT NULL AND datetime(n.last_access_time) < datetime($min_access_time_str) - RETURN + RETURN count(n) as total + """ + + # 数据查询 + data_query = """ + MATCH (n) + WHERE (n:Statement OR n:ExtractedEntity OR n:MemorySummary) + AND n.end_user_id = $end_user_id + AND n.activation_value IS NOT NULL + AND n.activation_value < $threshold + AND n.last_access_time IS NOT NULL + AND datetime(n.last_access_time) < datetime($min_access_time_str) + RETURN elementId(n) as node_id, labels(n)[0] as node_type, - CASE + CASE WHEN n:Statement THEN n.statement WHEN n:ExtractedEntity THEN n.name WHEN n:MemorySummary THEN n.content @@ -247,18 +264,32 @@ class MemoryForgetService: n.activation_value as activation_value, n.last_access_time as last_access_time ORDER BY n.activation_value ASC - LIMIT $limit """ - + + # 如果启用分页,添加 SKIP 和 LIMIT + if page is not None and pagesize is not None and page > 0 and pagesize > 0: + data_query += " SKIP $skip LIMIT $limit" + params = { 'end_user_id': end_user_id, 'threshold': forgetting_threshold, - 'min_access_time_str': min_access_time_str, - 'limit': limit + 'min_access_time_str': min_access_time_str } - - results = await connector.execute_query(query, **params) - + + # 获取总数(分页时需要) + total = 0 + if page is not None and pagesize is not None and page > 0 and pagesize > 0: + count_results = await connector.execute_query(count_query, **params) + if count_results: + total = count_results[0]['total'] + + # 添加分页参数 + if page is not None and pagesize is not None and page > 0 and pagesize > 0: + params['skip'] = (page - 1) * pagesize + params['limit'] = pagesize + + results = await connector.execute_query(data_query, **params) + pending_nodes = [] for result in results: # 将节点类型标签转换为小写 @@ -267,7 +298,7 @@ class MemoryForgetService: node_type_label = 'entity' elif node_type_label == 'memorysummary': node_type_label = 'summary' - + # 将 Neo4j DateTime 对象转换为时间戳(毫秒) last_access_time = result['last_access_time'] last_access_dt = convert_neo4j_datetime_to_python(last_access_time) @@ -278,7 +309,7 @@ class MemoryForgetService: last_access_timestamp = int(last_access_dt.timestamp() * 1000) else: last_access_timestamp = 0 - + pending_nodes.append({ 'node_id': str(result['node_id']), 'node_type': node_type_label, @@ -286,8 +317,20 @@ class MemoryForgetService: 'activation_value': result['activation_value'], 'last_access_time': last_access_timestamp }) - - return pending_nodes + + # 构建返回结果 + result: Dict[str, Any] = {'items': pending_nodes} + + # 如果启用分页,添加分页信息 + if page is not None and pagesize is not None and page > 0 and pagesize > 0: + result['page'] = { + 'page': page, + 'pagesize': pagesize, + 'total': total, + 'hasnext': (page * pagesize) < total + } + + return result async def trigger_forgetting_cycle( self, @@ -315,6 +358,12 @@ class MemoryForgetService: # 获取遗忘引擎组件 _, _, forgetting_scheduler, config = await self._get_forgetting_components(db, config_id) + # 如果参数为 None,使用配置中的默认值 + if max_merge_batch_size is None: + max_merge_batch_size = config.get('max_merge_batch_size', 100) + if min_days_since_access is None: + min_days_since_access = config.get('min_days_since_access', 30) + # 记录执行开始时间 execution_time = datetime.now() @@ -630,7 +679,7 @@ class MemoryForgetService: api_logger.error(f"获取历史趋势数据失败: {str(e)}") # 失败时返回空列表,不影响主流程 - # 获取待遗忘节点列表(前20个满足遗忘条件的节点) + # 获取待遗忘节点列表 pending_nodes = [] try: if end_user_id: @@ -646,8 +695,7 @@ class MemoryForgetService: connector=connector, end_user_id=end_user_id, forgetting_threshold=forgetting_threshold, - min_days_since_access=int(min_days), - limit=20 + min_days_since_access=int(min_days) ) api_logger.info(f"成功获取 {len(pending_nodes)} 个待遗忘节点") @@ -655,24 +703,79 @@ class MemoryForgetService: except Exception as e: api_logger.error(f"获取待遗忘节点失败: {str(e)}") # 失败时返回空列表,不影响主流程 - - # 构建统计信息 + + # 构建统计信息(不包含 pending_nodes,已分离到独立接口) stats = { 'activation_metrics': activation_metrics, 'node_distribution': node_distribution, 'recent_trends': recent_trends, - 'pending_nodes': pending_nodes, 'timestamp': int(datetime.now().timestamp() * 1000) } - + api_logger.info( f"成功获取遗忘引擎统计: total_nodes={stats['activation_metrics']['total_nodes']}, " f"low_activation_nodes={stats['activation_metrics']['low_activation_nodes']}, " - f"trend_days={len(recent_trends)}, pending_nodes={len(pending_nodes)}" + f"trend_days={len(recent_trends)}" ) - + return stats - + + async def get_pending_nodes( + self, + db: Session, + end_user_id: str, + config_id: Optional[UUID] = None, + page: int = 1, + pagesize: int = 10 + ) -> Dict[str, Any]: + """ + 获取待遗忘节点列表(独立分页接口) + + 查询满足遗忘条件的节点(激活值低于阈值且最后访问时间超过最小天数)。 + + Args: + db: 数据库会话 + end_user_id: 组ID(必填) + config_id: 配置ID(可选,用于获取遗忘阈值) + page: 页码(从1开始,默认1) + pagesize: 每页数量(默认10) + + Returns: + dict: 包含待遗忘节点列表和分页信息的字典 + - items: 待遗忘节点列表 + - page: 分页信息 + """ + # 获取遗忘引擎组件 + _, _, forgetting_scheduler, config = await self._get_forgetting_components(db, config_id) + + connector = forgetting_scheduler.connector + forgetting_threshold = config['forgetting_threshold'] + + # 验证 min_days_since_access 配置值 + min_days = config.get('min_days_since_access') + if min_days is None or not isinstance(min_days, (int, float)) or min_days < 0: + api_logger.warning( + f"min_days_since_access 配置无效: {min_days}, 使用默认值 7" + ) + min_days = 7 + + # 调用内部方法获取分页数据 + pending_nodes_result = await self._get_pending_forgetting_nodes( + connector=connector, + end_user_id=end_user_id, + forgetting_threshold=forgetting_threshold, + min_days_since_access=int(min_days), + page=page, + pagesize=pagesize + ) + + api_logger.info( + f"成功获取待遗忘节点列表: end_user_id={end_user_id}, " + f"page={page}, pagesize={pagesize}, total={pending_nodes_result.get('page', {}).get('total', 0)}" + ) + + return pending_nodes_result + async def get_forgetting_curve( self, db: Session, diff --git a/api/app/services/memory_konwledges_server.py b/api/app/services/memory_konwledges_server.py index b8961d33..523adadb 100644 --- a/api/app/services/memory_konwledges_server.py +++ b/api/app/services/memory_konwledges_server.py @@ -341,7 +341,7 @@ async def memory_konwledges_up( ) db_document = document_service.create_document(db=db, document=create_document_data, current_user=current_user) - return success(data=document_schema.Document.model_validate(db_document), msg="custom text upload successful") + return db_document async def create_document_chunk( @@ -350,7 +350,7 @@ async def create_document_chunk( create_data: ChunkCreate, db: Session, current_user: User -): +) -> DocumentChunk: """ 创建文档块 @@ -439,10 +439,10 @@ async def create_document_chunk( db_document.chunk_num += 1 db.commit() - return success(data=chunk, msg="文档块创建成功") + return chunk -async def write_rag(end_user_id, message, user_rag_memory_id): +async def write_rag(end_user_id, message, user_rag_memory_id) -> DocumentChunk: """ 将消息写入 RAG 知识库 @@ -482,11 +482,11 @@ async def write_rag(end_user_id, message, user_rag_memory_id): document = find_document_id_by_kb_and_filename(db=db, kb_id=user_rag_memory_id, file_name=f"{end_user_id}.txt") print('======', document) api_logger.info(f"查找文档结果: document_id={document}") + create_chunks = ChunkCreate(content=message) if document is not None: # 文档已存在,直接添加新块 api_logger.info(f"文档已存在,添加新块: document_id={document}") - create_chunks = ChunkCreate(content=message) result = await create_document_chunk( kb_id=kb_uuid, document_id=uuid.UUID(document), @@ -498,13 +498,20 @@ async def write_rag(end_user_id, message, user_rag_memory_id): else: # 文档不存在,创建新文档 api_logger.info(f"文档不存在,创建新文档: end_user_id={end_user_id}") - result = await memory_konwledges_up( + document = await memory_konwledges_up( kb_id=user_rag_memory_id, parent_id=user_rag_memory_id, create_data=create_data, db=db, current_user=current_user ) + result = await create_document_chunk( + kb_id=kb_uuid, + document_id=document.id, + create_data=create_chunks, + db=db, + current_user=current_user + ) # 重新查询刚创建的文档ID new_document_id = find_document_id_by_kb_and_filename( db=db, diff --git a/api/app/services/memory_perceptual_service.py b/api/app/services/memory_perceptual_service.py index 8a7c86e2..3ee238e2 100644 --- a/api/app/services/memory_perceptual_service.py +++ b/api/app/services/memory_perceptual_service.py @@ -12,11 +12,12 @@ from app.core.error_codes import BizCode from app.core.exceptions import BusinessException from app.core.logging_config import get_business_logger from app.core.models import RedBearLLM, RedBearModelConfig -from app.models import FileMetadata +from app.models import FileMetadata, ModelApiKey, ModelType from app.models.memory_perceptual_model import PerceptualType, FileStorageService from app.models.prompt_optimizer_model import RoleType from app.repositories.memory_perceptual_repository import MemoryPerceptualRepository -from app.schemas import FileType +from app.schemas import FileType, FileInput +from app.schemas.memory_config_schema import MemoryConfig from app.schemas.memory_perceptual_schema import ( PerceptualQuerySchema, PerceptualTimelineResponse, @@ -24,6 +25,8 @@ from app.schemas.memory_perceptual_schema import ( AudioModal, Content, VideoModal, TextModal ) from app.schemas.model_schema import ModelInfo +from app.services.model_service import ModelApiKeyService +from app.services.multimodal_service import MultimodalService business_logger = get_business_logger() @@ -195,21 +198,58 @@ class MemoryPerceptualService: business_logger.error(f"Failed to fetch perceptual memory timeline: {str(e)}") raise BusinessException(f"Failed to fetch perceptual memory timeline: {str(e)}", BizCode.DB_ERROR) + def _get_mutlimodal_client( + self, + file_type: FileType, + config: MemoryConfig + ) -> tuple[RedBearLLM | None, ModelApiKey | None]: + model_config = None + if file_type == FileType.AUDIO: + model_config = ModelApiKeyService.get_available_api_key( + self.db, + config.audio_model_id + ) + elif file_type == FileType.VIDEO: + model_config = ModelApiKeyService.get_available_api_key( + self.db, + config.video_model_id + ) + elif file_type == FileType.DOCUMENT: + model_config = ModelApiKeyService.get_available_api_key( + self.db, + config.llm_model_id + ) + elif file_type == FileType.IMAGE: + model_config = ModelApiKeyService.get_available_api_key( + self.db, + config.vision_model_id + ) + llm = None + if model_config: + llm = RedBearLLM( + RedBearModelConfig( + model_name=model_config.model_name, + provider=model_config.provider, + api_key=model_config.api_key, + base_url=model_config.api_base, + is_omni=model_config.is_omni + ) + ) + return llm, model_config + async def generate_perceptual_memory( self, end_user_id: str, - model_config: ModelInfo, - file_type: str, - file_url: str, - file_message: dict, + memory_config: MemoryConfig, + file: FileInput ): - memories = self.repository.get_by_url(file_url) + memories = self.repository.get_by_url(file.url) if memories: - business_logger.info(f"Perceptual memory already exists: {file_url}") + business_logger.info(f"Perceptual memory already exists: {file.url}") if end_user_id not in [memory.end_user_id for memory in memories]: business_logger.info(f"Copy perceptual memory end_user_id: {end_user_id}") memory_cache = memories[0] - self.repository.create_perceptual_memory( + memory = self.repository.create_perceptual_memory( end_user_id=uuid.UUID(end_user_id), perceptual_type=PerceptualType(memory_cache.perceptual_type), file_path=memory_cache.file_path, @@ -219,20 +259,33 @@ class MemoryPerceptualService: meta_data=memory_cache.meta_data ) self.db.commit() - - return - llm = RedBearLLM(RedBearModelConfig( + return memory + else: + for memory in memories: + if memory.end_user_id == uuid.UUID(end_user_id): + return memory + llm, model_config = self._get_mutlimodal_client(file.type, memory_config) + multimodel_service = MultimodalService(self.db, ModelInfo( model_name=model_config.model_name, provider=model_config.provider, api_key=model_config.api_key, - base_url=model_config.api_base, - is_omni=model_config.is_omni - ), type=model_config.model_type) + api_base=model_config.api_base, + is_omni=model_config.is_omni, + capability=model_config.capability, + model_type=ModelType.LLM + )) + file_message = await multimodel_service.process_files( + files=[file] + ) + if not file_message: + business_logger.warning(f"Unsupported file type {file}, model capability: {model_config.capability}") + return None + file_message = file_message[0] try: prompt_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'prompt') with open(os.path.join(prompt_path, 'perceptual_summary_system.jinja2'), 'r', encoding='utf-8') as f: opt_system_prompt = f.read() - rendered_system_message = Template(opt_system_prompt).render(file_type=file_type, language='zh') + rendered_system_message = Template(opt_system_prompt).render(file_type=file.type, language='zh') except FileNotFoundError: raise BusinessException(message="System prompt template not found", code=BizCode.NOT_FOUND) messages = [ @@ -242,8 +295,22 @@ class MemoryPerceptualService: ]} ] result = await llm.ainvoke(messages) - content = json_repair.repair_json(result.content, return_objects=True) - path = urlparse(file_url).path + content = result.content + final_output = "" + if isinstance(content, list): + for msg in content: + if isinstance(msg, dict): + final_output += msg.get("text", "") + elif isinstance(msg, str): + final_output += msg + elif isinstance(content, dict): + final_output += content.get("text", "") + elif isinstance(content, str): + final_output = content + else: + raise ValueError(f"Unexcept Model Output Type: {result.content}") + content = json_repair.repair_json(final_output, return_objects=True) + path = urlparse(file.url).path filename = os.path.basename(path) filename = unquote(filename) file_ext = os.path.splitext(filename)[1] @@ -252,21 +319,21 @@ class MemoryPerceptualService: stmt = select(FileMetadata).where( FileMetadata.id == file_id ) - file = self.db.execute(stmt).scalar_one_or_none() + file_obj = self.db.execute(stmt).scalar_one_or_none() - if file: - filename = file.file_name - file_ext = file.file_ext + if file_obj: + filename = file_obj.file_name + file_ext = file_obj.file_ext except ValueError: business_logger.debug(f"Remote file, file_id={filename}") if not file_ext: - if file_type == FileType.AUDIO: + if file.type == FileType.AUDIO: file_ext = ".mp3" - elif file_type == FileType.VIDEO: + elif file.type == FileType.VIDEO: file_ext = ".mp4" - elif file_type == FileType.DOCUMENT: + elif file.type == FileType.DOCUMENT: file_ext = ".txt" - elif file_type == FileType.IMAGE: + elif file.type == FileType.IMAGE: file_ext = ".jpg" filename += file_ext file_content = { @@ -274,11 +341,11 @@ class MemoryPerceptualService: "topic": content.get("topic"), "domain": content.get("domain") } - if file_type in [FileType.IMAGE, FileType.VIDEO]: + if file.type in [FileType.IMAGE, FileType.VIDEO]: file_modalities = { "scene": content.get("scene", []) } - elif file_type in [FileType.DOCUMENT]: + elif file.type in [FileType.DOCUMENT]: file_modalities = { "section_count": content.get("section_count", 0), "title": content.get("title", ""), @@ -288,10 +355,10 @@ class MemoryPerceptualService: file_modalities = { "speaker_count": content.get("speaker_count", 0) } - self.repository.create_perceptual_memory( + memory = self.repository.create_perceptual_memory( end_user_id=uuid.UUID(end_user_id), - perceptual_type=PerceptualType.trans_from_file_type(file_type), - file_path=file_url, + perceptual_type=PerceptualType.trans_from_file_type(file.type), + file_path=file.url, file_name=filename, file_ext=file_ext, summary=content.get('summary', ""), @@ -301,3 +368,4 @@ class MemoryPerceptualService: } ) self.db.commit() + return memory diff --git a/api/app/services/memory_storage_service.py b/api/app/services/memory_storage_service.py index 6e7c1ad4..58f3e8bd 100644 --- a/api/app/services/memory_storage_service.py +++ b/api/app/services/memory_storage_service.py @@ -11,9 +11,11 @@ import time from datetime import datetime from typing import Any, AsyncGenerator, Dict, List, Optional +from dotenv import load_dotenv +from sqlalchemy.orm import Session + from app.core.logging_config import get_config_logger, get_logger from app.core.memory.analytics.hot_memory_tags import ( - get_hot_memory_tags, get_raw_tags_from_db, filter_tags_with_llm, ) @@ -32,8 +34,6 @@ from app.schemas.memory_storage_schema import ( ) from app.services.memory_config_service import MemoryConfigService from app.utils.sse_utils import format_sse_message -from dotenv import load_dotenv -from sqlalchemy.orm import Session logger = get_logger(__name__) config_logger = get_config_logger() @@ -45,10 +45,10 @@ _neo4j_connector = Neo4jConnector() class MemoryStorageService: """Service for memory storage operations""" - + def __init__(self): logger.info("MemoryStorageService initialized") - + async def get_storage_info(self) -> dict: """ Example wrapper method - retrieves storage information @@ -59,17 +59,17 @@ class MemoryStorageService: Storage information dictionary """ logger.info("Getting storage info ") - + # Empty wrapper - implement your logic here result = { "status": "active", "message": "This is an example wrapper" } - - return result - -class DataConfigService: # 数据配置服务类(PostgreSQL) + return result + + +class DataConfigService: # 数据配置服务类(PostgreSQL) """Service layer for config params CRUD. 使用 SQLAlchemy ORM 进行数据库操作。 @@ -114,7 +114,7 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) return data_list # --- Create --- - def create(self, params: ConfigParamsCreate) -> Dict[str, Any]: # 创建配置参数(仅名称与描述) + def create(self, params: ConfigParamsCreate) -> Dict[str, Any]: # 创建配置参数(仅名称与描述) # 业务层检查同一工作空间下是否已存在同名配置 if params.workspace_id and params.config_name: from app.models.memory_config_model import MemoryConfig @@ -183,20 +183,20 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) return None # --- Delete --- - def delete(self, key: ConfigParamsDelete) -> Dict[str, Any]: # 删除配置参数(按配置ID) + def delete(self, key: ConfigParamsDelete) -> Dict[str, Any]: # 删除配置参数(按配置ID) success = MemoryConfigRepository.delete(self.db, key.config_id) if not success: raise ValueError("未找到配置") return {"affected": 1} # --- Update --- - def update(self, update: ConfigUpdate) -> Dict[str, Any]: # 部分更新配置参数 + def update(self, update: ConfigUpdate) -> Dict[str, Any]: # 部分更新配置参数 config = MemoryConfigRepository.update(self.db, update) if not config: raise ValueError("未找到配置") return {"affected": 1} - def update_extracted(self, update: ConfigUpdateExtracted) -> Dict[str, Any]: # 更新记忆萃取引擎配置参数 + def update_extracted(self, update: ConfigUpdateExtracted) -> Dict[str, Any]: # 更新记忆萃取引擎配置参数 config = MemoryConfigRepository.update_extracted(self.db, update) if not config: raise ValueError("未找到配置") @@ -207,14 +207,14 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) # 使用新方法: MemoryForgetService.read_forgetting_config() 和 MemoryForgetService.update_forgetting_config() # --- Read --- - def get_extracted(self, key: ConfigKey) -> Dict[str, Any]: # 获取萃取配置参数 + def get_extracted(self, key: ConfigKey) -> Dict[str, Any]: # 获取萃取配置参数 result = MemoryConfigRepository.get_extracted_config(self.db, key.config_id) if not result: raise ValueError("未找到配置") return result # --- Read All --- - def get_all(self, workspace_id = None) -> List[Dict[str, Any]]: # 获取所有配置参数 + def get_all(self, workspace_id=None) -> List[Dict[str, Any]]: # 获取所有配置参数 results = MemoryConfigRepository.get_all(self.db, workspace_id) # 检查并修正 pruning_scene 与 scene_name 不一致的记录 @@ -241,13 +241,8 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) except (ValueError, TypeError): config_id_old = None - - if config_id_old: - memory_config=config_id_old - else: - memory_config=config.config_id config_dict = { - "config_id": memory_config, + "config_id": str(config.config_id), "config_name": config.config_name, "config_desc": config.config_desc, "workspace_id": str(config.workspace_id) if config.workspace_id else None, @@ -289,7 +284,6 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) # 将 created_at 和 updated_at 转换为 YYYYMMDDHHmmss 格式 return self._convert_timestamps_to_format(data_list) - async def pilot_run_stream(self, payload: ConfigPilotRun, language: str = "zh") -> AsyncGenerator[str, None]: """ 流式执行试运行,产生 SSE 格式的进度事件 @@ -311,14 +305,14 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) """ from pathlib import Path project_root = str(Path(__file__).resolve().parents[2]) - + try: # 发出初始进度事件 yield format_sse_message("starting", { "message": "开始试运行...", "time": int(time.time() * 1000) }) - + # 步骤 1: 配置加载和验证(数据库优先) payload_cid = str(getattr(payload, "config_id", "") or "").strip() cid: Optional[str] = payload_cid if payload_cid else None @@ -344,27 +338,28 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) # 关联了本体场景,优先使用 custom_text if hasattr(payload, 'custom_text') and payload.custom_text: dialogue_text = payload.custom_text.strip() - logger.info(f"[PILOT_RUN_STREAM] Using custom_text for scene_id={memory_config.scene_id}, length: {len(dialogue_text)}") + logger.info( + f"[PILOT_RUN_STREAM] Using custom_text for scene_id={memory_config.scene_id}, length: {len(dialogue_text)}") else: # 如果没有提供 custom_text,回退到 dialogue_text dialogue_text = payload.dialogue_text.strip() if payload.dialogue_text else "" - logger.info(f"[PILOT_RUN_STREAM] No custom_text provided, using dialogue_text for scene_id={memory_config.scene_id}") + logger.info( + f"[PILOT_RUN_STREAM] No custom_text provided, using dialogue_text for scene_id={memory_config.scene_id}") else: # 没有关联本体场景,使用 dialogue_text dialogue_text = payload.dialogue_text.strip() if payload.dialogue_text else "" logger.info(f"[PILOT_RUN_STREAM] No scene_id, using dialogue_text, length: {len(dialogue_text)}") - + # 验证最终使用的文本不为空 if not dialogue_text: raise ValueError("试运行模式必须提供有效的文本内容(dialogue_text 或 custom_text)") - - logger.info(f"[PILOT_RUN_STREAM] Final text preview: {dialogue_text[:100]}") + logger.info(f"[PILOT_RUN_STREAM] Final text preview: {dialogue_text[:100]}") # 步骤 2: 创建进度回调函数捕获管线进度 # 使用队列在回调和生成器之间传递进度事件 progress_queue: asyncio.Queue = asyncio.Queue() - + async def progress_callback(stage: str, message: str, data: Optional[Dict[str, Any]] = None) -> None: """ 进度回调函数,将进度事件放入队列 @@ -375,14 +370,15 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) data: 可选的结果数据(用于传递节点执行结果) """ await progress_queue.put((stage, message, data)) - + # 步骤 3: 在后台任务中执行管线 async def run_pipeline(): """在后台执行管线并捕获异常""" try: from app.services.pilot_run_service import run_pilot_extraction - - logger.info(f"[PILOT_RUN_STREAM] Calling run_pilot_extraction with dialogue_text length: {len(dialogue_text)}") + + logger.info( + f"[PILOT_RUN_STREAM] Calling run_pilot_extraction with dialogue_text length: {len(dialogue_text)}") await run_pilot_extraction( memory_config=memory_config, dialogue_text=dialogue_text, @@ -391,60 +387,60 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) language=language, ) logger.info("[PILOT_RUN_STREAM] pipeline_main completed") - + # 标记管线完成 await progress_queue.put(("__PIPELINE_COMPLETE__", "", None)) except Exception as e: # 将异常放入队列 await progress_queue.put(("__PIPELINE_ERROR__", str(e), None)) - + # 启动后台任务 pipeline_task = asyncio.create_task(run_pipeline()) - + # 步骤 4: 从队列中读取进度事件并发出 while True: try: # 等待进度事件,设置超时以检测客户端断开 stage, message, data = await asyncio.wait_for( - progress_queue.get(), + progress_queue.get(), timeout=0.5 ) - + # 检查特殊标记 if stage == "__PIPELINE_COMPLETE__": break elif stage == "__PIPELINE_ERROR__": raise RuntimeError(message) - + # 构建进度事件数据 progress_data = { "message": message, "time": int(time.time() * 1000) } - + # 如果有结果数据,添加到事件中 if data: progress_data["data"] = data - + # 发出进度事件,使用 stage 作为事件类型 yield format_sse_message(stage, progress_data) - + except TimeoutError: # 超时,继续等待(这允许检测客户端断开) continue - + # 等待管线任务完成 await pipeline_task - + # 步骤 5: 读取提取结果 from app.core.config import settings result_path = settings.get_memory_output_path("extracted_result.json") if not os.path.isfile(result_path): raise FileNotFoundError(f"试运行完成,但未找到提取结果文件: {result_path}") - + with open(result_path, "r", encoding="utf-8") as rf: extracted_result = json.load(rf) - + # 步骤 6: 计算本体覆盖率并合并到结果中 result_data = { "config_id": cid, @@ -460,15 +456,15 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) result_data["ontology_coverage"] = ontology_coverage except Exception as cov_err: logger.warning(f"[PILOT_RUN_STREAM] Ontology coverage computation failed: {cov_err}", exc_info=True) - + yield format_sse_message("result", result_data) - + # 步骤 7: 发出完成事件 yield format_sse_message("done", { "message": "试运行完成", "time": int(time.time() * 1000) }) - + except asyncio.CancelledError: # 客户端断开连接 logger.info("[PILOT_RUN_STREAM] Client disconnected during streaming") @@ -483,11 +479,10 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) "time": int(time.time() * 1000) }) - async def _compute_ontology_coverage( - self, - extracted_result: Dict[str, Any], - memory_config, + self, + extracted_result: Dict[str, Any], + memory_config, ) -> Optional[Dict[str, Any]]: """根据提取结果中的实体类型,与场景/通用本体类型做互斥分类统计。 @@ -580,8 +575,6 @@ class DataConfigService: # 数据配置服务类(PostgreSQL) # -------------------- Neo4j Search & Analytics (fused from data_search_service.py) -------------------- # Ensure env for connector (e.g., NEO4J_PASSWORD) -load_dotenv() -_neo4j_connector = Neo4jConnector() async def search_dialogue(end_user_id: Optional[str] = None) -> Dict[str, Any]: @@ -664,7 +657,7 @@ async def kb_type_distribution(end_user_id: Optional[str] = None) -> Dict[str, A # 检查结果是否为空或长度不足 if not result or len(result) < 4: data = { - "total": 0, + "total": 0, "distribution": [ {"type": "dialogue", "count": 0}, {"type": "chunk", "count": 0}, @@ -701,10 +694,11 @@ async def search_edges(end_user_id: Optional[str] = None) -> List[Dict[str, Any] ) return result + async def analytics_hot_memory_tags( - db: Session, - current_user: User, - limit: int = 10 + db: Session, + current_user: User, + limit: int = 10 ) -> List[Dict[str, Any]]: """ 获取热门记忆标签,按数量排序并返回前N个 @@ -721,27 +715,27 @@ async def analytics_hot_memory_tags( from app.services.memory_dashboard_service import get_workspace_end_users # 使用 asyncio.to_thread 避免阻塞事件循环 end_users = await asyncio.to_thread(get_workspace_end_users, db, workspace_id, current_user) - + if not end_users: return [] - + # 步骤1: 收集所有用户的原始标签(不调用LLM) connector = Neo4jConnector() try: all_raw_tags = [] for end_user in end_users: raw_tags = await get_raw_tags_from_db( - connector, - str(end_user.id), - limit=raw_limit, + connector, + str(end_user.id), + limit=raw_limit, by_user=False ) if raw_tags: all_raw_tags.extend(raw_tags) - + if not all_raw_tags: return [] - + # 步骤2: 聚合相同标签的频率 tag_frequency_map = {} for tag_name, frequency in all_raw_tags: @@ -749,36 +743,36 @@ async def analytics_hot_memory_tags( tag_frequency_map[tag_name] += frequency else: tag_frequency_map[tag_name] = frequency - + # 步骤3: 按频率降序排序,取前raw_limit个 sorted_tags = sorted( - tag_frequency_map.items(), - key=lambda x: x[1], + tag_frequency_map.items(), + key=lambda x: x[1], reverse=True )[:raw_limit] - + if not sorted_tags: return [] - + # 步骤4: 只调用一次LLM进行筛选 tag_names = [tag for tag, _ in sorted_tags] - + # 使用第一个用户的end_user_id来获取LLM配置 # 因为同一工作空间下的用户应该使用相同的配置 first_end_user_id = str(end_users[0].id) filtered_tag_names = await filter_tags_with_llm(tag_names, first_end_user_id) - + # 步骤5: 根据LLM筛选结果构建最终列表(保留频率) final_tags = [] for tag, freq in sorted_tags: if tag in filtered_tag_names: final_tags.append((tag, freq)) - + # 步骤6: 只返回前limit个 top_tags = final_tags[:limit] - + return [{"name": t, "frequency": f} for t, f in top_tags] - + finally: await connector.close() @@ -815,11 +809,11 @@ async def analytics_recent_activity_stats(workspace_id: Optional[str] = None) -> source = "log" total = ( - stats.get("chunk_count", 0) - + stats.get("statements_count", 0) - + stats.get("triplet_entities_count", 0) - + stats.get("triplet_relations_count", 0) - + stats.get("temporal_count", 0) + stats.get("chunk_count", 0) + + stats.get("statements_count", 0) + + stats.get("triplet_entities_count", 0) + + stats.get("triplet_relations_count", 0) + + stats.get("temporal_count", 0) ) # 计算"最新一次活动多久前"(仅日志来源时有效) @@ -845,5 +839,3 @@ async def analytics_recent_activity_stats(workspace_id: Optional[str] = None) -> data = {"total": total, "stats": stats, "latest_relative": latest_relative, "source": source} return data - - diff --git a/api/app/services/model_service.py b/api/app/services/model_service.py index a7398504..b98674ba 100644 --- a/api/app/services/model_service.py +++ b/api/app/services/model_service.py @@ -154,10 +154,17 @@ class ModelConfigService: } elif model_type_lower == "embedding": - # Embedding 模型验证(在线程中运行同步方法) + # Embedding 模型验证 + # 统一使用 RedBearEmbeddings(自动支持火山引擎多模态) embedding = RedBearEmbeddings(model_config) test_texts = [test_message, "测试文本"] - vectors = await asyncio.to_thread(embedding.embed_documents, test_texts) + + # 火山引擎使用 embed_batch,其他使用 embed_documents + if provider.lower() == "volcano": + vectors = await asyncio.to_thread(embedding.embed_batch, test_texts) + else: + vectors = await asyncio.to_thread(embedding.embed_documents, test_texts) + elapsed_time = time.time() - start_time return { @@ -193,6 +200,56 @@ class ModelConfigService: }, "error": None } + + elif model_type_lower == "image": + # 图片生成模型验证 + from app.core.models.generation import RedBearImageGenerator + + generator = RedBearImageGenerator(model_config) + result = await generator.agenerate( + prompt="a cute panda", + size="2K" + ) + elapsed_time = time.time() - start_time + logger.info(f"成功生成图片,结果: {result}") + + return { + "valid": True, + "message": "图片生成模型配置验证成功", + "response": f"成功生成图片,结果: {result}", + "elapsed_time": elapsed_time, + "usage": { + "prompt_length": len("a cute panda"), + "image_count": 1 + }, + "error": None + } + + elif model_type_lower == "video": + # 视频生成模型验证 + from app.core.models.generation import RedBearVideoGenerator + + generator = RedBearVideoGenerator(model_config) + result = await generator.agenerate( + prompt="a cute panda playing in bamboo forest", + duration=5 + ) + elapsed_time = time.time() - start_time + + # 视频生成是异步任务,返回任务ID + task_id = result.get("task_id") if isinstance(result, dict) else None + + return { + "valid": True, + "message": "视频生成模型配置验证成功", + "response": f"成功创建视频生成任务,任务ID: {task_id}", + "elapsed_time": elapsed_time, + "usage": { + "prompt_length": len("a cute panda playing in bamboo forest"), + "task_id": task_id + }, + "error": None + } else: return { diff --git a/api/app/services/multimodal_service.py b/api/app/services/multimodal_service.py index 6cb0a7f0..f854e987 100644 --- a/api/app/services/multimodal_service.py +++ b/api/app/services/multimodal_service.py @@ -9,17 +9,18 @@ - OpenAI: 支持 URL 和 base64 格式 """ import base64 +import csv import io -import uuid +import json +import re +import olefile +import struct import zipfile -import chardet from abc import ABC, abstractmethod from typing import List, Dict, Any, Optional -import csv -import json - import PyPDF2 +import chardet import httpx import magic import openpyxl @@ -35,7 +36,6 @@ from app.models.file_metadata_model import FileMetadata from app.schemas.app_schema import FileInput, FileType, TransferMethod from app.schemas.model_schema import ModelInfo from app.services.audio_transcription_service import AudioTranscriptionService -from app.tasks import write_perceptual_memory logger = get_business_logger() @@ -297,6 +297,7 @@ PROVIDER_STRATEGIES = { "bedrock": BedrockFormatStrategy, "anthropic": BedrockFormatStrategy, "openai": OpenAIFormatStrategy, + "volcano": OpenAIFormatStrategy, } @@ -342,92 +343,14 @@ class MultimodalService: async def process_files( self, - end_user_id: uuid.UUID | str, files: Optional[List[FileInput]], - ) -> List[Dict[str, Any]]: """ 处理文件列表,返回 LLM 可用的格式 Args: - end_user_id: 用户ID files: 文件输入列表 - Returns: - List[Dict]: LLM 可用的内容格式列表(根据 provider 返回不同格式) - """ - if not files: - return [] - if isinstance(end_user_id, uuid.UUID): - end_user_id = str(end_user_id) - - # 获取对应的策略 - # dashscope 的 omni 模型使用 OpenAI 兼容格式 - if self.provider == "dashscope" and self.is_omni: - strategy_class = OpenAIFormatStrategy - else: - strategy_class = PROVIDER_STRATEGIES.get(self.provider) - if not strategy_class: - logger.warning(f"未找到 provider '{self.provider}' 的策略,使用默认策略") - strategy_class = DashScopeFormatStrategy - - result = [] - for idx, file in enumerate(files): - strategy = strategy_class(file) - if not file.url: - file.url = await self.get_file_url(file) - try: - if file.type == FileType.IMAGE and "vision" in self.capability: - is_support, content = await self._process_image(file, strategy) - result.append(content) - if is_support: - self.write_perceptual_memory(end_user_id, file.type, file.url, content) - elif file.type == FileType.DOCUMENT: - is_support, content = await self._process_document(file, strategy) - result.append(content) - if is_support: - self.write_perceptual_memory(end_user_id, file.type, file.url, content) - elif file.type == FileType.AUDIO and "audio" in self.capability: - is_support, content = await self._process_audio(file, strategy) - result.append(content) - if is_support: - self.write_perceptual_memory(end_user_id, file.type, file.url, content) - elif file.type == FileType.VIDEO and "video" in self.capability: - is_support, content = await self._process_video(file, strategy) - result.append(content) - if is_support: - self.write_perceptual_memory(end_user_id, file.type, file.url, content) - else: - logger.warning(f"不支持的文件类型: {file.type}") - except Exception as e: - logger.error( - f"处理文件失败", - extra={ - "file_index": idx, - "file_type": file.type, - "error": str(e) - }, - exc_info=True - ) - # 继续处理其他文件,不中断整个流程 - result.append({ - "type": "text", - "text": f"[文件处理失败: {str(e)}]" - }) - - logger.info(f"成功处理 {len(result)}/{len(files)} 个文件,provider={self.provider}") - return result - - async def history_process_files( - self, - files: Optional[List[FileInput]], - ) -> List[Dict[str, Any]]: - """ - 处理文件列表,返回 LLM 可用的格式 - - Args: - files: 文件输入列表 - Returns: List[Dict]: LLM 可用的内容格式列表(根据 provider 返回不同格式) """ @@ -483,17 +406,6 @@ class MultimodalService: logger.info(f"成功处理 {len(result)}/{len(files)} 个文件,provider={self.provider}") return result - def write_perceptual_memory( - self, - end_user_id: str, - file_type: str, - file_url: str, - file_message: dict - ): - """写入感知记忆""" - if end_user_id and self.api_config: - write_perceptual_memory.delay(end_user_id, self.api_config.model_dump(), file_type, file_url, file_message) - async def _process_image(self, file: FileInput, strategy) -> tuple[bool, Dict[str, Any]]: """ 处理图片文件 @@ -693,31 +605,75 @@ class MultimodalService: try: word_file = io.BytesIO(file_content) doc = Document(word_file) - return '\n'.join(p.text for p in doc.paragraphs) + text_lines = [] + for p in doc.paragraphs: + text = p.text.strip() + if text: + text_lines.append(text) + + for table in doc.tables: + for row in table.rows: + for cell in row.cells: + text = cell.text.strip() + if text: + text_lines.append(text) + + full_text = "\n".join(text_lines) + return full_text.strip() or "[docx 文件无文本内容]" except Exception as e: - logger.error(f"提取 docx 文本失败: {e}") + logger.error(f"提取 docx 文本失败: {str(e)}", exc_info=True) return f"[docx 提取失败: {str(e)}]" - # 旧版 .doc(OLE2 格式) + # 旧版 .doc(OLE2/CFB 格式),按 Word Binary Format 规范解析 piece table try: - import olefile ole = olefile.OleFileIO(io.BytesIO(file_content)) - if not ole.exists('WordDocument'): - return "[doc 提取失败: 未找到 WordDocument 流]" - # 读取 WordDocument 流,提取可见 ASCII/Unicode 文本 - stream = ole.openstream('WordDocument').read() - # Word Binary Format: 文本在流中以 UTF-16-LE 编码存储 - # 简单提取:过滤出可打印字符段 - try: - text = stream.decode('utf-16-le', errors='ignore') - except Exception: - text = stream.decode('latin-1', errors='ignore') - # 过滤控制字符,保留可打印内容 - import re - text = re.sub(r'[\x00-\x08\x0b\x0c\x0e-\x1f\x7f]', '', text) - text = re.sub(r' +', ' ', text).strip() + word_stream = ole.openstream('WordDocument').read() + + # FIB offset 0xA bit9 决定使用 0Table 还是 1Table + fib_flags = struct.unpack_from(' List[UserModel]: """获取租户下的用户列表""" @@ -155,6 +156,7 @@ class TenantService: skip=skip, limit=limit, is_active=is_active, + is_superuser=is_superuser, search=search ) @@ -162,12 +164,14 @@ class TenantService: self, tenant_id: uuid.UUID, is_active: Optional[bool] = None, + is_superuser: Optional[bool] = None, search: Optional[str] = None ) -> int: """统计租户下的用户数量""" return self.user_repo.count_users_by_tenant( tenant_id=tenant_id, is_active=is_active, + is_superuser=is_superuser, search=search ) diff --git a/api/app/services/user_memory_service.py b/api/app/services/user_memory_service.py index 12e0c324..ab51d922 100644 --- a/api/app/services/user_memory_service.py +++ b/api/app/services/user_memory_service.py @@ -361,83 +361,58 @@ class UserMemoryService: if hasattr(original_value, 'timestamp'): data[key] = UserMemoryService._datetime_to_timestamp(original_value) return data - - def update_end_user_profile( + # ======================== 用户别名及信息 ======================== + def get_end_user_info( self, db: Session, - end_user_id: str, - profile_update: Any + end_user_id: str ) -> Dict[str, Any]: """ - 更新终端用户的基本信息 + 查询单个终端用户信息记录 Args: db: 数据库会话 end_user_id: 终端用户ID (UUID) - profile_update: 包含更新字段的 Pydantic 模型 Returns: { "success": bool, - "data": dict, # 更新后的用户档案数据 + "data": dict, "error": Optional[str] } """ try: - # 转换为UUID并查询用户 - user_uuid = uuid.UUID(end_user_id) - repo = EndUserRepository(db) - end_user = repo.get_by_id(user_uuid) + from app.repositories.end_user_info_repository import EndUserInfoRepository + from app.core.api_key_utils import datetime_to_timestamp - if not end_user: - logger.warning(f"终端用户不存在: end_user_id={end_user_id}") + # 转换为UUID并查询 + user_uuid = uuid.UUID(end_user_id) + end_user_info_record = EndUserInfoRepository(db).get_by_end_user_id(user_uuid) + + if not end_user_info_record: + logger.warning(f"终端用户信息记录不存在: end_user_id={end_user_id}") return { "success": False, "data": None, - "error": "终端用户不存在" + "error": "终端用户信息记录不存在" } - # 获取更新数据(排除 end_user_id 字段) - update_data = profile_update.model_dump(exclude_unset=True, exclude={'end_user_id'}) + # 构建响应数据(转换时间为毫秒时间戳) + response_data = { + "end_user_info_id": str(end_user_info_record.id), + "end_user_id": str(end_user_info_record.end_user_id), + "other_name": end_user_info_record.other_name, + "aliases": end_user_info_record.aliases, + "meta_data": end_user_info_record.meta_data, + "created_at": datetime_to_timestamp(end_user_info_record.created_at), + "updated_at": datetime_to_timestamp(end_user_info_record.updated_at) + } - # 特殊处理 hire_date:如果提供了时间戳,转换为 DateTime - if 'hire_date' in update_data: - hire_date_timestamp = update_data['hire_date'] - if hire_date_timestamp is not None: - from app.core.api_key_utils import timestamp_to_datetime - update_data['hire_date'] = timestamp_to_datetime(hire_date_timestamp) - # 如果是 None,保持 None(允许清空) - - # 更新字段 - for field, value in update_data.items(): - setattr(end_user, field, value) - - # 更新时间戳 - end_user.updated_at = datetime.now() - end_user.updatetime_profile = datetime.now() - - # 提交更改 - db.commit() - db.refresh(end_user) - - # 构建响应数据 - from app.schemas.end_user_schema import EndUserProfileResponse - profile_data = EndUserProfileResponse( - id=end_user.id, - other_name=end_user.other_name, - position=end_user.position, - department=end_user.department, - contact=end_user.contact, - phone=end_user.phone, - hire_date=end_user.hire_date, - updatetime_profile=end_user.updatetime_profile - ) - - logger.info(f"成功更新用户信息: end_user_id={end_user_id}, updated_fields={list(update_data.keys())}") + logger.info(f"成功查询终端用户信息记录: end_user_id={end_user_id}") return { "success": True, - "data": self.convert_profile_to_dict_with_timestamp(profile_data), + "data": response_data, "error": None } @@ -446,17 +421,181 @@ class UserMemoryService: return { "success": False, "data": None, - "error": "无效的用户ID格式" + "error": "无效的终端用户ID格式" } except Exception as e: - db.rollback() - logger.error(f"用户信息更新失败: end_user_id={end_user_id}, error={str(e)}") + logger.error(f"查询终端用户信息记录失败: end_user_id={end_user_id}, error={str(e)}") return { "success": False, "data": None, "error": str(e) } + def update_end_user_info( + self, + db: Session, + end_user_id: str, + update_data: Dict[str, Any] + ) -> Dict[str, Any]: + """ + 更新终端用户信息记录 + + Args: + db: 数据库会话 + end_user_id: 终端用户ID (UUID) + update_data: 更新数据字典 + + Returns: + { + "success": bool, + "data": dict, + "error": Optional[str] + } + """ + try: + from app.repositories.end_user_info_repository import EndUserInfoRepository + from app.repositories.end_user_repository import EndUserRepository + from app.core.api_key_utils import datetime_to_timestamp + + # 转换为UUID并查询 + user_uuid = uuid.UUID(end_user_id) + end_user_info_record = EndUserInfoRepository(db).get_by_end_user_id(user_uuid) + + if not end_user_info_record: + logger.warning(f"终端用户信息记录不存在: end_user_id={end_user_id}") + return { + "success": False, + "data": None, + "error": "终端用户信息记录不存在" + } + + # 定义允许更新的字段白名单 + allowed_fields = {'other_name', 'aliases', 'meta_data'} + + # 用户占位名称黑名单,不允许作为 other_name 或出现在 aliases 中 + _user_placeholder_names = {'用户', '我', 'User', 'I'} + + # 过滤 other_name:不允许设置为占位名称 + if 'other_name' in update_data and update_data['other_name'] and update_data['other_name'].strip() in _user_placeholder_names: + logger.warning(f"拒绝将占位名称 '{update_data['other_name']}' 设置为 other_name") + del update_data['other_name'] + + # 过滤 aliases:移除占位名称和非字符串值 + if 'aliases' in update_data and update_data['aliases']: + update_data['aliases'] = [ + a for a in update_data['aliases'] + if isinstance(a, str) and a.strip() and a.strip() not in _user_placeholder_names + ] + + # 检查是否更新了 aliases 字段 + aliases_updated = 'aliases' in update_data and update_data['aliases'] != end_user_info_record.aliases + + # 检查是否更新了 other_name 字段 + other_name_updated = 'other_name' in update_data and update_data['other_name'] != end_user_info_record.other_name + + # 更新字段(仅允许白名单中的字段) + for field, value in update_data.items(): + if field in allowed_fields: + setattr(end_user_info_record, field, value) + + # 更新时间戳 + end_user_info_record.updated_at = datetime.now() + + # 如果 other_name 被更新,同步更新 end_user 表 + if other_name_updated: + end_user_record = EndUserRepository(db).get_by_id(user_uuid) + if end_user_record: + end_user_record.other_name = update_data['other_name'] + end_user_record.updated_at = datetime.now() + logger.info(f"同步更新 end_user 表的 other_name: end_user_id={end_user_id}, other_name={update_data['other_name']}") + else: + logger.warning(f"未找到对应的 end_user 记录: end_user_id={end_user_id}") + + # 提交更改 + db.commit() + db.refresh(end_user_info_record) + + # 如果 aliases 被更新,同步到 Neo4j + if aliases_updated: + try: + import asyncio + asyncio.run(self._sync_aliases_to_neo4j(end_user_id, update_data['aliases'])) + logger.info(f"已触发 aliases 同步到 Neo4j: end_user_id={end_user_id}, aliases={update_data['aliases']}") + except Exception as sync_error: + logger.error(f"触发同步 aliases 到 Neo4j 失败: {sync_error}", exc_info=True) + # 不影响主流程,只记录错误 + + # 构建响应数据(转换时间为毫秒时间戳) + response_data = { + "end_user_info_id": str(end_user_info_record.id), + "end_user_id": str(end_user_info_record.end_user_id), + "other_name": end_user_info_record.other_name, + "aliases": end_user_info_record.aliases, + "meta_data": end_user_info_record.meta_data, + "created_at": datetime_to_timestamp(end_user_info_record.created_at), + "updated_at": datetime_to_timestamp(end_user_info_record.updated_at) + } + + logger.info(f"成功更新终端用户信息记录: end_user_id={end_user_id}, updated_fields={list(update_data.keys())}") + + return { + "success": True, + "data": response_data, + "error": None + } + + except ValueError: + logger.error(f"无效的 end_user_id 格式: {end_user_id}") + return { + "success": False, + "data": None, + "error": "无效的终端用户ID格式" + } + except Exception as e: + db.rollback() + logger.error(f"更新终端用户信息记录失败: end_user_id={end_user_id}, error={str(e)}") + return { + "success": False, + "data": None, + "error": str(e) + } + + async def _sync_aliases_to_neo4j(self, end_user_id: str, aliases: List[str]) -> None: + """ + 将 aliases 同步到 Neo4j 中的用户实体 + + Args: + end_user_id: 终端用户ID + aliases: 别名列表 + """ + from app.repositories.neo4j.neo4j_connector import Neo4jConnector + + # Cypher 查询:更新用户实体的 aliases + cypher_query = """ + MATCH (e:ExtractedEntity) + WHERE e.end_user_id = $end_user_id + AND e.name IN ['用户', '我', 'User', 'I'] + SET e.aliases = $aliases + RETURN e.id AS entity_id, e.name AS entity_name, e.aliases AS updated_aliases + """ + + connector = Neo4jConnector() + try: + result = await connector.execute_query( + cypher_query, + end_user_id=end_user_id, + aliases=aliases + ) + + if result: + logger.info(f"成功同步 aliases 到 Neo4j: end_user_id={end_user_id}, 更新了 {len(result)} 个实体节点") + else: + logger.warning(f"未找到需要更新的用户实体节点: end_user_id={end_user_id}") + + except Exception as e: + logger.error(f"同步 aliases 到 Neo4j 失败: {e}", exc_info=True) + raise + async def get_cached_memory_insight( self, db: Session, @@ -1887,7 +2026,8 @@ async def _extract_node_properties(label: str, properties: Dict[str, Any],node_ "Chunk": ["content", "created_at"], "Statement": ["temporal_info", "stmt_type", "statement", "valid_at", "created_at", "caption","emotion_keywords","emotion_type","emotion_subject"], "ExtractedEntity": ["description", "name", "entity_type", "created_at", "caption","aliases","connect_strength"], - "MemorySummary": ["summary", "content", "created_at", "caption"] # 添加 content 字段 + "MemorySummary": ["summary", "content", "created_at", "caption"], # 添加 content 字段 + "Perceptual": ["file_name", "file_path", "file_type", "domain", "topic", "keywords", "summary"] } # 获取该节点类型的白名单字段 diff --git a/api/app/services/user_service.py b/api/app/services/user_service.py index e23b1ac3..3122d282 100644 --- a/api/app/services/user_service.py +++ b/api/app/services/user_service.py @@ -78,18 +78,7 @@ def create_user(db: Session, user: UserCreate) -> User: business_logger.info(f"创建用户: {user.username}, email: {user.email}") try: - # 检查用户名是否已存在 - business_logger.debug(f"检查用户名是否已存在: {user.username}") - db_user_by_username = user_repository.get_user_by_username(db, username=user.username) - if db_user_by_username: - business_logger.warning(f"用户名已存在: {user.username}") - raise BusinessException( - "用户名已存在", - code=BizCode.DUPLICATE_NAME, - context={"username": user.username, "email": user.email} - ) - - # 检查邮箱是否已注册 + # 检查邮箱是否已注册(邮箱保持唯一) business_logger.debug(f"检查邮箱是否已注册: {user.email}") db_user_by_email = user_repository.get_user_by_email(db, email=user.email) if db_user_by_email: @@ -164,22 +153,7 @@ def create_superuser(db: Session, user: UserCreate, current_user: User) -> User: ) try: - # 检查用户名是否已存在 - business_logger.debug(f"检查用户名是否已存在: {user.username}") - db_user_by_username = user_repository.get_user_by_username(db, username=user.username) - if db_user_by_username: - business_logger.warning(f"用户名已存在: {user.username}") - raise BusinessException( - "用户名已存在", - code=BizCode.DUPLICATE_NAME, - context={ - "username": user.username, - "email": user.email, - "created_by": str(current_user.id) - } - ) - - # 检查邮箱是否已注册 + # 检查邮箱是否已注册(邮箱保持唯一) business_logger.debug(f"检查邮箱是否已注册: {user.email}") db_user_by_email = user_repository.get_user_by_email(db, email=user.email) if db_user_by_email: @@ -276,6 +250,20 @@ def deactivate_user(db: Session, user_id_to_deactivate: uuid.UUID, current_user: } ) + # 检查是否为租户联系人 + from app.models.tenant_model import Tenants + tenant = db.query(Tenants).filter(Tenants.id == db_user.tenant_id).first() + if tenant and tenant.contact_email and tenant.contact_email == db_user.email: + business_logger.warning(f"尝试停用租户联系人: {db_user.email}, tenant_id={db_user.tenant_id}") + raise BusinessException( + "该管理员是租户联系人,请先在租户信息中更换联系邮箱,再禁用此管理员", + code=BizCode.FORBIDDEN, + context={ + "user_id": str(user_id_to_deactivate), + "tenant_id": str(db_user.tenant_id) + } + ) + # 停用用户 business_logger.debug(f"执行用户停用: {db_user.username} (ID: {user_id_to_deactivate})") db_user.is_active = False diff --git a/api/app/services/workflow_import_service.py b/api/app/services/workflow_import_service.py index 2b36c5ea..fd8f25f3 100644 --- a/api/app/services/workflow_import_service.py +++ b/api/app/services/workflow_import_service.py @@ -12,7 +12,7 @@ from app.aioRedis import aio_redis_set, aio_redis_get from app.core.config import settings from app.core.exceptions import BusinessException from app.core.workflow.adapters.base_adapter import WorkflowImportResult, WorkflowParserResult -from app.core.workflow.adapters.errors import UnsupportPlatform, InvalidConfiguration +from app.core.workflow.adapters.errors import UnsupportedPlatform, InvalidConfiguration from app.core.workflow.adapters.registry import PlatformAdapterRegistry from app.schemas import AppCreate from app.schemas.workflow_schema import WorkflowConfigCreate @@ -46,7 +46,7 @@ class WorkflowImportService: success=False, temp_id=None, workflow_id=None, - errors=[UnsupportPlatform(platform=platform)] + errors=[UnsupportedPlatform(platform=platform)] ) adapter = self.registry.get_adapter(platform, config) diff --git a/api/app/services/workflow_service.py b/api/app/services/workflow_service.py index 0e1306b7..c7d7f2b1 100644 --- a/api/app/services/workflow_service.py +++ b/api/app/services/workflow_service.py @@ -20,15 +20,17 @@ from app.core.workflow.variable.base_variable import FileObject from app.db import get_db from app.models import App from app.models.workflow_model import WorkflowConfig, WorkflowExecution +from app.repositories import knowledge_repository from app.repositories.workflow_repository import ( WorkflowConfigRepository, WorkflowExecutionRepository, WorkflowNodeExecutionRepository ) -from app.schemas import DraftRunRequest, FileInput, FileType +from app.schemas import DraftRunRequest, FileInput from app.services.conversation_service import ConversationService from app.services.multi_agent_service import convert_uuids_to_str from app.services.multimodal_service import MultimodalService +from app.services.workspace_service import get_workspace_storage_type_without_auth logger = logging.getLogger(__name__) @@ -540,6 +542,25 @@ class WorkflowService: mapped = internal_event return mapped + def _get_memory_store_info(self, workspace_id: uuid.UUID) -> tuple[str, str]: + storage_type = get_workspace_storage_type_without_auth(self.db, workspace_id) + user_rag_memory_id = "" + if storage_type == "rag": + knowledge = knowledge_repository.get_knowledge_by_name( + db=self.db, + name="USER_RAG_MERORY", + workspace_id=workspace_id + ) + if knowledge: + user_rag_memory_id = str(knowledge.id) + else: + logger.warning( + f"No knowledge base named 'USER_RAG_MEMORY' found, " + f"workspace_id: {workspace_id}, will use neo4j storage" + ) + storage_type = 'neo4j' + return storage_type, user_rag_memory_id + # ==================== 工作流执行 ==================== async def run( @@ -607,6 +628,7 @@ class WorkflowService: try: files = await self._handle_file_input(payload.files) + storage_type, user_rag_memory_id = self._get_memory_store_info(workspace_id) input_data["files"] = files message_id = uuid.uuid4() # 更新状态为运行中 @@ -631,7 +653,9 @@ class WorkflowService: input_data=input_data, execution_id=execution.execution_id, workspace_id=str(workspace_id), - user_id=payload.user_id + user_id=payload.user_id, + memory_storage_type=storage_type, + user_rag_memory_id=user_rag_memory_id ) # 更新执行结果 if result.get("status") == "completed": @@ -780,6 +804,7 @@ class WorkflowService: try: files = await self._handle_file_input(payload.files) + storage_type, user_rag_memory_id = self._get_memory_store_info(workspace_id) input_data["files"] = files self.update_execution_status(execution.execution_id, "running") executions = self.execution_repo.get_by_conversation_id(conversation_id=conversation_id_uuid) @@ -801,6 +826,8 @@ class WorkflowService: execution_id=execution.execution_id, workspace_id=str(workspace_id), user_id=payload.user_id, + memory_storage_type=storage_type, + user_rag_memory_id=user_rag_memory_id ): if event.get("event") == "workflow_end": status = event.get("data", {}).get("status") diff --git a/api/app/services/workspace_service.py b/api/app/services/workspace_service.py index cefb8380..90b5cf65 100644 --- a/api/app/services/workspace_service.py +++ b/api/app/services/workspace_service.py @@ -863,7 +863,7 @@ def get_workspace_storage_type( def get_workspace_storage_type_without_auth( db: Session, workspace_id: uuid.UUID, -) -> Optional[str]: +) -> str: """获取工作空间的存储类型(无需权限验证,用于公开分享等场景) Args: diff --git a/api/app/tasks.py b/api/app/tasks.py index 3a237d82..72421a5f 100644 --- a/api/app/tasks.py +++ b/api/app/tasks.py @@ -1,5 +1,4 @@ import asyncio -import hashlib import os import re import shutil @@ -36,12 +35,12 @@ from app.core.rag.vdb.elasticsearch.elasticsearch_vector import ( ) from app.db import get_db, get_db_context from app.models import Document, File, Knowledge +from app.models.end_user_model import EndUser from app.schemas import document_schema, file_schema -from app.schemas.model_schema import ModelInfo -from app.services.memory_agent_service import MemoryAgentService -from app.services.memory_perceptual_service import MemoryPerceptualService +from app.services.memory_agent_service import MemoryAgentService, get_end_user_connected_config +from app.services.memory_forget_service import MemoryForgetService from app.utils.config_utils import resolve_config_id -from app.utils.redis_lock import RedisLock +from app.utils.redis_lock import RedisFairLock logger = get_logger(__name__) @@ -102,7 +101,12 @@ def get_sync_redis_client() -> Optional[redis.StrictRedis]: def set_asyncio_event_loop(): - """Set the asyncio event loop for the current thread.""" + """Ensure an open asyncio event loop exists for the current thread. + + Reuses the existing event loop if one is available and still open. + Creates and installs a new event loop only when the current one is + closed or missing (e.g. after ``_shutdown_loop_gracefully``). + """ try: loop = asyncio.get_event_loop() if loop.is_closed(): @@ -114,6 +118,30 @@ def set_asyncio_event_loop(): return loop +def _shutdown_loop_gracefully(loop: asyncio.AbstractEventLoop): + """Gracefully shutdown pending async generators and tasks on the event loop. + + This prevents 'RuntimeError: Event loop is closed' from httpx.AsyncClient.__del__ + by giving pending aclose() coroutines a chance to run before the loop is discarded. + + Note: This only tears down the given loop. Callers that need a fresh event + loop afterwards should use ``set_asyncio_event_loop()`` explicitly. + """ + try: + # Cancel and collect all remaining tasks + all_tasks = asyncio.all_tasks(loop) + if all_tasks: + for task in all_tasks: + task.cancel() + loop.run_until_complete(asyncio.gather(*all_tasks, return_exceptions=True)) + # Shutdown async generators (triggers __aclose__ on httpx clients etc.) + loop.run_until_complete(loop.shutdown_asyncgens()) + except Exception: + pass + finally: + loop.close() + + @celery_app.task(name="tasks.process_item") def process_item(item: dict): """ @@ -1073,9 +1101,15 @@ def read_message_task(self, end_user_id: str, message: str, history: List[Dict[s @celery_app.task(name="app.core.memory.agent.write_message", bind=True) -def write_message_task(self, end_user_id: str, message: list[dict], config_id: str | int, storage_type: str, - user_rag_memory_id: str, - language: str = "zh") -> Dict[str, Any]: +def write_message_task( + self, + end_user_id: str, + message: list[dict], + config_id: str | int, + storage_type: str, + user_rag_memory_id: str, + language: str = "zh" +) -> Dict[str, Any]: """Celery task to process a write message via MemoryAgentService. Args: end_user_id: Group ID for the memory agent (also used as end_user_id) @@ -1091,7 +1125,6 @@ def write_message_task(self, end_user_id: str, message: list[dict], config_id: s Raises: Exception on failure """ - logger.info( f"[CELERY WRITE] Starting write task - end_user_id={end_user_id}, " f"config_id={config_id} (type: {type(config_id).__name__}), " @@ -1105,14 +1138,11 @@ def write_message_task(self, end_user_id: str, message: list[dict], config_id: s try: with get_db_context() as db: actual_config_id = resolve_config_id(config_id, db) - print(100 * '-') - print(actual_config_id) - print(100 * '-') - logger.info( - f"[CELERY WRITE] Converted config_id to UUID: {actual_config_id} (type: {type(actual_config_id).__name__})") + logger.info(f"[CELERY WRITE] Converted config_id to UUID: {actual_config_id} " + f"(type: {type(actual_config_id).__name__})") except (ValueError, AttributeError) as e: - logger.error( - f"[CELERY WRITE] Invalid config_id format: {config_id} (type: {type(config_id).__name__}), error: {e}") + logger.error(f"[CELERY WRITE] Invalid config_id format: {config_id} " + f"(type: {type(config_id).__name__}), error: {e}") return { "status": "FAILURE", "error": f"Invalid config_id format: {config_id} - {str(e)}", @@ -1144,17 +1174,36 @@ def write_message_task(self, end_user_id: str, message: list[dict], config_id: s logger.info(f"[CELERY WRITE] Write completed successfully: {result}") return result + redis_client = get_sync_redis_client() + lock = None + if redis_client is not None: + lock = RedisFairLock( + key=f"memory_write:{end_user_id}", + redis_client=redis_client, + expire=600, + timeout=3600, + auto_renewal=True, + ) + if not lock.acquire(): + logger.warning(f"[CELERY WRITE] 获取锁超时,跳过本次写入: end_user_id={end_user_id}") + return { + "status": "SKIPPED", + "error": "acquire lock timeout", + "end_user_id": end_user_id, + "config_id": str(config_id), + "elapsed_time": time.time() - start_time, + "task_id": self.request.id, + } + try: - # 尝试获取现有事件循环,如果不存在则创建新的 loop = set_asyncio_event_loop() result = loop.run_until_complete(_run()) elapsed_time = time.time() - start_time - logger.info( - f"[CELERY WRITE] Task completed successfully - elapsed_time={elapsed_time:.2f}s, task_id={self.request.id}") + logger.info(f"[CELERY WRITE] Task completed successfully " + f"- elapsed_time={elapsed_time:.2f}s, task_id={self.request.id}") - # 记录该用户最后一次 write_message 成功的时间,供时间轴筛选使用 try: _r = get_sync_redis_client() if _r is not None: @@ -1167,7 +1216,6 @@ def write_message_task(self, end_user_id: str, message: list[dict], config_id: s ) except Exception as _e: logger.warning(f"[CELERY WRITE] 写入 last_done 时间戳失败(不影响主流程): {_e}") - return { "status": "SUCCESS", "result": result, @@ -1196,6 +1244,15 @@ def write_message_task(self, end_user_id: str, message: list[dict], config_id: s "elapsed_time": elapsed_time, "task_id": self.request.id } + finally: + if lock is not None: + try: + lock.release() + except Exception as e: + logger.warning(f"[CELERY WRITE] 释放锁失败: {e}") + # Gracefully shutdown the event loop to prevent + # 'RuntimeError: Event loop is closed' from httpx.AsyncClient.__del__ + _shutdown_loop_gracefully(loop) # unused task @@ -1859,7 +1916,7 @@ def workspace_reflection_task(self) -> Dict[str, Any]: @celery_app.task( name="app.tasks.run_forgetting_cycle_task", bind=True, - ignore_result=True, + ignore_result=False, # 改为 False 以便在 Flower 中查看结果 max_retries=0, acks_late=False, time_limit=7200, @@ -1867,68 +1924,77 @@ def workspace_reflection_task(self) -> Dict[str, Any]: ) def run_forgetting_cycle_task(self, config_id: Optional[uuid.UUID] = None) -> Dict[str, Any]: """定时任务:运行遗忘周期 - - 定期执行遗忘周期,识别并融合低激活值的知识节点。 - - Args: - config_id: 配置ID(可选,如果为None则使用默认配置) - - Returns: - 包含任务执行结果的字典 + + 遍历所有终端用户,执行遗忘周期。 """ start_time = time.time() - async def _run() -> Dict[str, Any]: - from app.services.memory_forget_service import MemoryForgetService - + async def _process_users() -> Dict[str, Any]: with get_db_context() as db: - try: - logger.info(f"开始执行遗忘周期定时任务,config_id: {config_id}") + end_users = db.query(EndUser).all() + if not end_users: + logger.info("没有终端用户,跳过遗忘周期") + return {"status": "SUCCESS", "message": "没有终端用户", + "report": {"merged_count": 0, "failed_count": 0, "processed_users": 0}, + "duration_seconds": time.time() - start_time} - forget_service = MemoryForgetService() + logger.info(f"开始处理 {len(end_users)} 个终端用户的遗忘周期") + forget_service = MemoryForgetService() + total_merged = total_failed = processed_users = 0 + failed_users = [] - # 运行遗忘周期 - # FIXME: MemeoryForgetService - report = await forget_service.trigger_forgetting( - db=db, - end_user_id=None, # 处理所有组 - config_id=config_id - ) + for end_user in end_users: + try: + # 获取用户配置(自动回退到工作空间默认配置) + connected_config = get_end_user_connected_config(str(end_user.id), db) + user_config_id = resolve_config_id(connected_config.get("memory_config_id"), db) + + if not user_config_id: + failed_users.append({"end_user_id": str(end_user.id), "error": "无法获取配置"}) + continue - duration = time.time() - start_time + # 执行遗忘周期 + report = await forget_service.trigger_forgetting_cycle( + db=db, end_user_id=str(end_user.id), config_id=user_config_id + ) + + total_merged += report.get('merged_count', 0) + total_failed += report.get('failed_count', 0) + processed_users += 1 + + logger.info(f"用户 {end_user.id}: 融合 {report.get('merged_count', 0)} 对节点") + + except Exception as e: + logger.error(f"处理用户 {end_user.id} 失败: {e}", exc_info=True) + failed_users.append({"end_user_id": str(end_user.id), "error": str(e)}) - logger.info( - f"遗忘周期定时任务完成: " - f"融合 {report['merged_count']} 对节点, " - f"失败 {report['failed_count']} 对, " - f"耗时 {duration:.2f} 秒" - ) + duration = time.time() - start_time + logger.info(f"遗忘周期完成: {processed_users}/{len(end_users)} 用户, " + f"融合 {total_merged} 对, 耗时 {duration:.2f}s") - return { - "status": "SUCCESS", - "message": "遗忘周期执行成功", - "report": report, - "duration_seconds": duration - } - - except Exception as e: - duration = time.time() - start_time - logger.error(f"遗忘周期定时任务失败: {str(e)}", exc_info=True) - - return { - "status": "FAILED", - "message": f"遗忘周期执行失败: {str(e)}", - "duration_seconds": duration - } + return { + "status": "SUCCESS", + "message": f"处理 {processed_users} 个用户", + "report": { + "merged_count": total_merged, + "failed_count": total_failed, + "processed_users": processed_users, + "total_users": len(end_users), + "failed_users": failed_users + }, + "duration_seconds": duration + } # 运行异步函数 - loop = asyncio.new_event_loop() - asyncio.set_event_loop(loop) try: - result = loop.run_until_complete(_run()) - return result - finally: - loop.close() + return asyncio.run(_process_users()) + except Exception as e: + logger.error(f"遗忘周期任务失败: {e}", exc_info=True) + return { + "status": "FAILED", + "message": f"任务失败: {str(e)}", + "duration_seconds": time.time() - start_time + } # ============================================================================= @@ -2611,68 +2677,111 @@ def init_interest_distribution_for_users(self, end_user_ids: List[str]) -> Dict[ } -@celery_app.task( - name="app.tasks.write_perceptual_memory", - bind=True, - ignore_result=True, - max_retries=0, - acks_late=False, - time_limit=3600, - soft_time_limit=3300, -) -def write_perceptual_memory( - self, - end_user_id: str, - model_api_config: dict, - file_type: str, - file_url: str, - file_message: dict -): - """ - Write perceptual memory for a user into PostgreSQL and Neo4j. - - This task generates or updates the user's perceptual memory - in the backend databases. It is intended to be executed asynchronously - via Celery. - - Args: - end_user_id (uuid.UUID): The unique identifier of the end user. - model_api_config (ModelInfo): API configuration for the model - used to generate perceptual memory. - file_type (str): The file type - file_url (url): The url of file - file_message (dict): The file message containing details about the file - to be processed. - - Returns: - None - """ - file_url_md5 = hashlib.md5(file_url.encode("utf-8")).hexdigest() - set_asyncio_event_loop() - with RedisLock(f"perceptual:{file_url_md5}", redis_client=get_sync_redis_client()): - model_info = ModelInfo(**model_api_config) - with get_db_context() as db: - memory_perceptual_service = MemoryPerceptualService(db) - return asyncio.run(memory_perceptual_service.generate_perceptual_memory( - end_user_id, - model_info, - file_type, - file_url, - file_message, - )) - - # ============================================================================= # 社区聚类补全任务(触发型) # ============================================================================= +@celery_app.task( + name="app.tasks.run_incremental_clustering", + bind=True, + ignore_result=False, + max_retries=2, + acks_late=True, + time_limit=1800, # 30分钟硬超时 + soft_time_limit=1700, +) +def run_incremental_clustering( + self, + end_user_id: str, + new_entity_ids: List[str], + llm_model_id: Optional[str] = None, + embedding_model_id: Optional[str] = None, +) -> Dict[str, Any]: + """增量聚类任务:处理新增实体的社区分配和元数据生成。 + + 此任务在后台异步执行,不阻塞 write_message 主流程。 + + Args: + end_user_id: 用户 ID + new_entity_ids: 新增实体 ID 列表 + llm_model_id: LLM 模型 ID(可选) + embedding_model_id: Embedding 模型 ID(可选) + + Returns: + 包含任务执行结果的字典 + """ + start_time = time.time() + + async def _run() -> Dict[str, Any]: + from app.core.logging_config import get_logger + from app.repositories.neo4j.neo4j_connector import Neo4jConnector + from app.core.memory.storage_services.clustering_engine.label_propagation import LabelPropagationEngine + + logger = get_logger(__name__) + logger.info( + f"[IncrementalClustering] 开始增量聚类任务 - end_user_id={end_user_id}, " + f"实体数={len(new_entity_ids)}, llm_model_id={llm_model_id}" + ) + + connector = Neo4jConnector() + try: + engine = LabelPropagationEngine( + connector=connector, + llm_model_id=llm_model_id, + embedding_model_id=embedding_model_id, + ) + + # 执行增量聚类 + await engine.run(end_user_id=end_user_id, new_entity_ids=new_entity_ids) + + logger.info(f"[IncrementalClustering] 增量聚类完成 - end_user_id={end_user_id}") + + return { + "status": "SUCCESS", + "end_user_id": end_user_id, + "entity_count": len(new_entity_ids), + } + except Exception as e: + logger.error(f"[IncrementalClustering] 增量聚类失败: {e}", exc_info=True) + raise + finally: + await connector.close() + + try: + loop = set_asyncio_event_loop() + result = loop.run_until_complete(_run()) + result["elapsed_time"] = time.time() - start_time + result["task_id"] = self.request.id + + logger.info( + f"[IncrementalClustering] 任务完成 - task_id={self.request.id}, " + f"elapsed_time={result['elapsed_time']:.2f}s" + ) + + return result + except Exception as e: + elapsed_time = time.time() - start_time + logger.error( + f"[IncrementalClustering] 任务失败 - task_id={self.request.id}, " + f"elapsed_time={elapsed_time:.2f}s, error={str(e)}", + exc_info=True + ) + return { + "status": "FAILURE", + "error": str(e), + "end_user_id": end_user_id, + "elapsed_time": elapsed_time, + "task_id": self.request.id, + } + + @celery_app.task( name="app.tasks.init_community_clustering_for_users", bind=True, ignore_result=False, max_retries=0, acks_late=False, - time_limit=7200, # 2小时硬超时 + time_limit=7200, # 2小时硬超时 soft_time_limit=6900, ) def init_community_clustering_for_users(self, end_user_ids: List[str], workspace_id: Optional[str] = None) -> Dict[str, Any]: @@ -2760,7 +2869,7 @@ def init_community_clustering_for_users(self, end_user_ids: List[str], workspace patch_fail = 0 for cid in incomplete_ids: try: - await engine._generate_community_metadata(cid, end_user_id) + await engine._generate_community_metadata([cid], end_user_id) patch_ok += 1 except Exception as patch_err: patch_fail += 1 @@ -2787,7 +2896,8 @@ def init_community_clustering_for_users(self, end_user_ids: List[str], workspace embedding_model_id=embedding_model_id, ) - logger.info(f"[CommunityCluster] 用户 {end_user_id} 有 {len(entities)} 个实体,开始全量聚类,llm_model_id={llm_model_id}") + logger.info( + f"[CommunityCluster] 用户 {end_user_id} 有 {len(entities)} 个实体,开始全量聚类,llm_model_id={llm_model_id}") await engine.full_clustering(end_user_id) initialized += 1 logger.info(f"[CommunityCluster] 用户 {end_user_id} 聚类完成") @@ -2810,12 +2920,6 @@ def init_community_clustering_for_users(self, end_user_ids: List[str], workspace } try: - try: - import nest_asyncio - nest_asyncio.apply() - except ImportError: - pass - loop = set_asyncio_event_loop() result = loop.run_until_complete(_run()) result["elapsed_time"] = time.time() - start_time @@ -2829,3 +2933,6 @@ def init_community_clustering_for_users(self, end_user_ids: List[str], workspace "elapsed_time": time.time() - start_time, "task_id": self.request.id, } + + +# unused task \ No newline at end of file diff --git a/api/app/utils/redis_lock.py b/api/app/utils/redis_lock.py index 99f62d84..a86ba46e 100644 --- a/api/app/utils/redis_lock.py +++ b/api/app/utils/redis_lock.py @@ -1,6 +1,7 @@ import redis import uuid import time +import threading UNLOCK_SCRIPT = """ if redis.call("get", KEYS[1]) == ARGV[1] then @@ -10,45 +11,136 @@ else end """ +RENEW_SCRIPT = """ +if redis.call("get", KEYS[1]) == ARGV[1] then + return redis.call("expire", KEYS[1], ARGV[2]) +else + return 0 +end +""" -class RedisLock: +CLEANUP_DEAD_HEAD_SCRIPT = """ +local queue_key = KEYS[1] +local lock_key = KEYS[2] + +local first = redis.call("lindex", queue_key, 0) +if not first then + return 0 +end + +if redis.call("exists", lock_key) == 1 then + return 0 +end + +redis.call("lpop", queue_key) +return 1 +""" + +SAFE_RELEASE_QUEUE_SCRIPT = """ +local queue_key = KEYS[1] +local value = ARGV[1] + +local first = redis.call("lindex", queue_key, 0) +if first == value then + redis.call("lpop", queue_key) + return 1 +end +return 0 +""" + + +def _ensure_str(val): + """统一将 Redis 返回值转为 str,兼容 decode_responses=True/False""" + if val is None: + return None + if isinstance(val, bytes): + return val.decode("utf-8") + return str(val) + + +class RedisFairLock: def __init__( self, key: str, redis_client: redis.StrictRedis, - expire: int = 60, - retry_interval: float = 0.1, - timeout: float = 30 - + expire: int = 30, + retry_interval: float = 0.05, + timeout: float = 600, + auto_renewal: bool = True ): self.key = key - self.expire = expire + self.queue_key = f"{key}:queue" self.value = str(uuid.uuid4()) - self._locked = False + self.expire = expire self.retry_interval = retry_interval self.timeout = timeout - self.redis_client = redis_client + self.redis = redis_client + self._locked = False + self.auto_renewal = auto_renewal + self._renew_thread = None + self._stop_renew = threading.Event() - def acquire(self) -> bool: + def acquire(self): start = time.time() + + self.redis.rpush(self.queue_key, self.value) + while True: - ok = self.redis_client.set(self.key, self.value, ex=self.expire, nx=True) - if ok: - self._locked = True - return True - if time.time() - start >= self.timeout: + first = _ensure_str(self.redis.lindex(self.queue_key, 0)) + + if first == self.value: + ok = self.redis.set(self.key, self.value, nx=True, ex=self.expire) + if ok: + self._locked = True + + if self.auto_renewal: + self._start_renewal() + return True + + if first: + self.redis.eval(CLEANUP_DEAD_HEAD_SCRIPT, 2, self.queue_key, self.key) + + if time.time() - start > self.timeout: + self.redis.lrem(self.queue_key, 0, self.value) return False + time.sleep(self.retry_interval) + def _renewal_loop(self): + while not self._stop_renew.is_set(): + time.sleep(self.expire / 3) + if self._stop_renew.is_set(): + break + + self.redis.eval( + RENEW_SCRIPT, + 1, + self.key, + self.value, + str(self.expire) + ) + + def _start_renewal(self): + self._stop_renew = threading.Event() + self._renew_thread = threading.Thread(target=self._renewal_loop, daemon=True) + self._renew_thread.start() + + def _stop_renewal(self): + self._stop_renew.set() + if self._renew_thread: + self._renew_thread.join(timeout=1) + def release(self): if not self._locked: return - self.redis_client.eval( - UNLOCK_SCRIPT, - 1, - self.key, - self.value - ) + + if self.auto_renewal: + self._stop_renewal() + + self.redis.eval(UNLOCK_SCRIPT, 1, self.key, self.value) + + self.redis.eval(SAFE_RELEASE_QUEUE_SCRIPT, 1, self.queue_key, self.value) + self._locked = False def __enter__(self): @@ -59,3 +151,4 @@ class RedisLock: def __exit__(self, exc_type, exc_val, exc_tb): self.release() + diff --git a/api/migrations/versions/05a681a6ca93_202603231611.py b/api/migrations/versions/05a681a6ca93_202603231611.py new file mode 100644 index 00000000..5ab9c4de --- /dev/null +++ b/api/migrations/versions/05a681a6ca93_202603231611.py @@ -0,0 +1,32 @@ +"""202603231611 + +Revision ID: 05a681a6ca93 +Revises: 74b51dfece29 +Create Date: 2026-03-23 16:12:44.110292 + +""" +from typing import Sequence, Union + +from alembic import op +import sqlalchemy as sa + + +# revision identifiers, used by Alembic. +revision: str = '05a681a6ca93' +down_revision: Union[str, None] = '74b51dfece29' +branch_labels: Union[str, Sequence[str], None] = None +depends_on: Union[str, Sequence[str], None] = None + + +def upgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.drop_index(op.f('ix_users_username'), table_name='users') + op.create_index(op.f('ix_users_username'), 'users', ['username'], unique=False) + # ### end Alembic commands ### + + +def downgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.drop_index(op.f('ix_users_username'), table_name='users') + op.create_index(op.f('ix_users_username'), 'users', ['username'], unique=True) + # ### end Alembic commands ### diff --git a/api/migrations/versions/1480a7d680fb_202603261815.py b/api/migrations/versions/1480a7d680fb_202603261815.py new file mode 100644 index 00000000..4c6f8c9c --- /dev/null +++ b/api/migrations/versions/1480a7d680fb_202603261815.py @@ -0,0 +1,59 @@ +"""202603261815 + +Revision ID: 1480a7d680fb +Revises: adaefcbe2aa1 +Create Date: 2026-03-26 18:16:07.886033 + +""" +from typing import Sequence, Union + +from alembic import op +import sqlalchemy as sa +from sqlalchemy.dialects import postgresql + +# revision identifiers, used by Alembic. +revision: str = '1480a7d680fb' +down_revision: Union[str, None] = 'adaefcbe2aa1' +branch_labels: Union[str, Sequence[str], None] = None +depends_on: Union[str, Sequence[str], None] = None + + +def upgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.create_table('end_user_info', + sa.Column('id', sa.UUID(), nullable=False), + sa.Column('end_user_id', sa.UUID(), nullable=False, comment='关联的终端用户ID'), + sa.Column('other_name', sa.String(), nullable=False, comment='关联的用户名称'), + sa.Column('aliases', sa.ARRAY(sa.String()), nullable=True, comment='用户别名列表(字符串数组)'), + sa.Column('meta_data', postgresql.JSONB(astext_type=sa.Text()), nullable=True, comment='用户相关的扩展信息(JSON格式)'), + sa.Column('created_at', sa.DateTime(), nullable=True, comment='创建时间'), + sa.Column('updated_at', sa.DateTime(), nullable=True, comment='更新时间'), + sa.ForeignKeyConstraint(['end_user_id'], ['end_users.id'], ), + sa.PrimaryKeyConstraint('id') + ) + op.create_index(op.f('ix_end_user_info_end_user_id'), 'end_user_info', ['end_user_id'], unique=False) + op.create_index(op.f('ix_end_user_info_id'), 'end_user_info', ['id'], unique=False) + + connection = op.get_bind() + connection.execute(sa.text(""" + INSERT INTO end_user_info (id, end_user_id, other_name, aliases, meta_data, created_at, updated_at) + SELECT + gen_random_uuid() as id, + id as end_user_id, + other_name, + '{}'::TEXT[] as aliases, + NULL as meta_data, + NOW() as created_at, + NOW() as updated_at + FROM end_users + """)) + + # ### end Alembic commands ### + + +def downgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.drop_index(op.f('ix_end_user_info_id'), table_name='end_user_info') + op.drop_index(op.f('ix_end_user_info_end_user_id'), table_name='end_user_info') + op.drop_table('end_user_info') + # ### end Alembic commands ### diff --git a/api/migrations/versions/1ea8fe97b5b7_202603252115.py b/api/migrations/versions/1ea8fe97b5b7_202603252115.py new file mode 100644 index 00000000..1f0df3e7 --- /dev/null +++ b/api/migrations/versions/1ea8fe97b5b7_202603252115.py @@ -0,0 +1,42 @@ +"""202603252115 + +Revision ID: 1ea8fe97b5b7 +Revises: e28bcc212da5 +Create Date: 2026-03-25 21:14:41.825048 + +""" +from typing import Sequence, Union + +from alembic import op +import sqlalchemy as sa + + +# revision identifiers, used by Alembic. +revision: str = '1ea8fe97b5b7' +down_revision: Union[str, None] = 'e28bcc212da5' +branch_labels: Union[str, Sequence[str], None] = None +depends_on: Union[str, Sequence[str], None] = None + + +def upgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.add_column('tenants', sa.Column('contact_name', sa.String(length=100), nullable=True)) + op.add_column('tenants', sa.Column('contact_email', sa.String(length=255), nullable=True)) + op.add_column('tenants', sa.Column('contact_phone', sa.String(length=50), nullable=True)) + op.add_column('tenants', sa.Column('plan', sa.String(length=50), nullable=True)) + op.add_column('tenants', sa.Column('plan_expired_at', sa.DateTime(), nullable=True)) + op.add_column('tenants', sa.Column('api_ops_rate_limit', sa.String(length=100), nullable=True)) + op.add_column('tenants', sa.Column('status', sa.String(length=50), nullable=True)) + # ### end Alembic commands ### + + +def downgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.drop_column('tenants', 'status') + op.drop_column('tenants', 'api_ops_rate_limit') + op.drop_column('tenants', 'plan_expired_at') + op.drop_column('tenants', 'plan') + op.drop_column('tenants', 'contact_phone') + op.drop_column('tenants', 'contact_email') + op.drop_column('tenants', 'contact_name') + # ### end Alembic commands ### diff --git a/api/migrations/versions/4e89970f9e7c_202603271515.py b/api/migrations/versions/4e89970f9e7c_202603271515.py new file mode 100644 index 00000000..f37c4b27 --- /dev/null +++ b/api/migrations/versions/4e89970f9e7c_202603271515.py @@ -0,0 +1,30 @@ +"""202603271515 + +Revision ID: 4e89970f9e7c +Revises: 6b8a461148ff +Create Date: 2026-03-27 15:12:27.518344 + +""" +from typing import Sequence, Union + +from alembic import op +import sqlalchemy as sa + + +# revision identifiers, used by Alembic. +revision: str = '4e89970f9e7c' +down_revision: Union[str, None] = '6b8a461148ff' +branch_labels: Union[str, Sequence[str], None] = None +depends_on: Union[str, Sequence[str], None] = None + + +def upgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.add_column('users', sa.Column('phone', sa.String(length=50), nullable=True)) + # ### end Alembic commands ### + + +def downgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.drop_column('users', 'phone') + # ### end Alembic commands ### diff --git a/api/migrations/versions/6b8a461148ff_202603261955.py b/api/migrations/versions/6b8a461148ff_202603261955.py new file mode 100644 index 00000000..a0bdac87 --- /dev/null +++ b/api/migrations/versions/6b8a461148ff_202603261955.py @@ -0,0 +1,32 @@ +"""202603261955 + +Revision ID: 6b8a461148ff +Revises: 1480a7d680fb +Create Date: 2026-03-26 19:55:24.041039 + +""" +from typing import Sequence, Union + +from alembic import op +import sqlalchemy as sa + + +# revision identifiers, used by Alembic. +revision: str = '6b8a461148ff' +down_revision: Union[str, None] = '1480a7d680fb' +branch_labels: Union[str, Sequence[str], None] = None +depends_on: Union[str, Sequence[str], None] = None + + +def upgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.drop_column('tenants', 'feature_user_management') + op.drop_column('tenants', 'feature_billing') + # ### end Alembic commands ### + + +def downgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.add_column('tenants', sa.Column('feature_billing', sa.BOOLEAN(), server_default=sa.text('false'), autoincrement=False, nullable=False, comment='是否启用收费管理菜单')) + op.add_column('tenants', sa.Column('feature_user_management', sa.BOOLEAN(), server_default=sa.text('false'), autoincrement=False, nullable=False, comment='是否启用用户管理菜单')) + # ### end Alembic commands ### diff --git a/api/migrations/versions/74b51dfece29_20260311000.py b/api/migrations/versions/74b51dfece29_20260311000.py new file mode 100644 index 00000000..aa9feab1 --- /dev/null +++ b/api/migrations/versions/74b51dfece29_20260311000.py @@ -0,0 +1,156 @@ +"""20260311000 + +Revision ID: 74b51dfece29 +Revises: f017efe4831c +Create Date: 2026-03-19 10:15:42.488027 + +""" +from typing import Sequence, Union + +from alembic import op +import sqlalchemy as sa + + +# revision identifiers, used by Alembic. +revision: str = '74b51dfece29' +down_revision: Union[str, None] = 'f017efe4831c' +branch_labels: Union[str, Sequence[str], None] = None +depends_on: Union[str, Sequence[str], None] = None + + +def upgrade() -> None: + # 先删除旧的触发器(如果存在) + op.execute("DROP TRIGGER IF EXISTS tr_documents_update_stats ON documents;") + + # 创建或更新 knowledges 统计信息的函数 + op.execute(""" +CREATE OR REPLACE FUNCTION update_knowledge_stats() +RETURNS TRIGGER AS $$ +DECLARE + -- 声明变量用于存储当前处理的知识库ID + current_kb_id UUID; + -- 声明变量用于存储文件夹知识库ID(如果存在) + folder_kb_id UUID; + -- 声明变量用于存储递归查询结果 + folder_ids UUID[]; +BEGIN + -- 处理 documents 表的插入、更新或删除 + IF TG_TABLE_NAME = 'documents' THEN + -- 1. 更新 knowledges 表的 doc_num + UPDATE knowledges SET doc_num = ( + SELECT COUNT(*) FROM documents + WHERE kb_id = knowledges.id AND status = 1 + ) + WHERE id = NEW.kb_id OR id = OLD.kb_id; + + -- 2. 更新 knowledges 表的 chunk_num + UPDATE knowledges SET chunk_num = ( + SELECT COALESCE(SUM(chunk_num), 0) FROM documents + WHERE kb_id = knowledges.id AND status = 1 + ) + WHERE id = NEW.kb_id OR id = OLD.kb_id; + + -- 通过 knowledge_shares 表同步统计信息 + -- 1. 使用 source_kb_id 的 doc_num 更新 target_kb_id 的 doc_num + UPDATE knowledges AS target + SET doc_num = source.doc_num + FROM knowledge_shares ks + JOIN knowledges AS source ON source.id = ks.source_kb_id + WHERE ks.target_kb_id = target.id + AND (source.id = NEW.kb_id OR source.id = OLD.kb_id); + + -- 2. 使用 source_kb_id 的 chunk_num 更新 target_kb_id 的 chunk_num + UPDATE knowledges AS target + SET chunk_num = source.chunk_num + FROM knowledge_shares ks + JOIN knowledges AS source ON source.id = ks.source_kb_id + WHERE ks.target_kb_id = target.id + AND (source.id = NEW.kb_id OR source.id = OLD.kb_id); + + -- 处理文件夹知识库的统计更新 + -- 获取当前处理的知识库ID(可能是NEW或OLD中的kb_id) + IF NEW.kb_id IS NOT NULL THEN + current_kb_id := NEW.kb_id; + ELSIF OLD.kb_id IS NOT NULL THEN + current_kb_id := OLD.kb_id; + ELSE + RETURN NULL; + END IF; + + -- 查找当前知识库的父文件夹(如果有) + SELECT id INTO folder_kb_id FROM knowledges + WHERE id IN ( + SELECT parent_id FROM knowledges WHERE id = current_kb_id + ) AND type = 'Folder'; + + -- 如果存在父文件夹,递归处理所有父文件夹 + IF folder_kb_id IS NOT NULL THEN + -- 使用递归CTE获取所有父文件夹ID(包括多级嵌套) + WITH RECURSIVE folder_hierarchy AS ( + -- 基础查询:获取直接父文件夹 + SELECT id FROM knowledges + WHERE id = folder_kb_id AND type = 'Folder' + UNION ALL + -- 递归查询:获取父文件夹的父文件夹 + SELECT k.id FROM knowledges k + JOIN folder_hierarchy fh ON k.id = k.parent_id + WHERE k.type = 'Folder' + ) + -- 将结果存入数组以便处理 + SELECT array_agg(id) INTO folder_ids FROM folder_hierarchy; + + -- 遍历所有父文件夹并更新统计信息 + FOR i IN 1..array_length(folder_ids, 1) LOOP + -- 更新文件夹的doc_num(汇总所有子知识库的doc_num) + UPDATE knowledges SET doc_num = ( + -- 汇总直接子知识库的doc_num + SELECT COALESCE(SUM(child.doc_num), 0) + FROM knowledges child + WHERE child.parent_id = folder_ids[i] AND child.status = 1 + -- 加上直接属于该文件夹的文档数(如果有) + UNION ALL + SELECT COALESCE(COUNT(*), 0) + FROM documents + WHERE kb_id = folder_ids[i] AND status = 1 + LIMIT 1 + ) + WHERE id = folder_ids[i]; + + -- 更新文件夹的chunk_num(汇总所有子知识库的chunk_num) + UPDATE knowledges SET chunk_num = ( + -- 汇总直接子知识库的chunk_num + SELECT COALESCE(SUM(child.chunk_num), 0) + FROM knowledges child + WHERE child.parent_id = folder_ids[i] AND child.status = 1 + -- 加上直接属于该文件夹的文档的chunk_num(如果有) + UNION ALL + SELECT COALESCE(SUM(d.chunk_num), 0) + FROM documents d + WHERE d.kb_id = folder_ids[i] AND d.status = 1 + LIMIT 1 + ) + WHERE id = folder_ids[i]; + END LOOP; + END IF; + END IF; + + RETURN NULL; +END; +$$ LANGUAGE plpgsql; + """) + + # documents 表上的触发器(插入、更新、删除后) + op.execute(""" +CREATE TRIGGER tr_documents_update_stats + AFTER INSERT OR UPDATE OR DELETE ON documents + FOR EACH ROW + EXECUTE FUNCTION update_knowledge_stats(); + """) + + +def downgrade() -> None: + # 删除触发器 + op.execute("DROP TRIGGER IF EXISTS tr_documents_update_stats ON documents;") + # 删除函数 + op.execute("DROP FUNCTION IF EXISTS update_knowledge_stats();") + diff --git a/api/migrations/versions/adaefcbe2aa1_202603261630.py b/api/migrations/versions/adaefcbe2aa1_202603261630.py new file mode 100644 index 00000000..b8235dd7 --- /dev/null +++ b/api/migrations/versions/adaefcbe2aa1_202603261630.py @@ -0,0 +1,32 @@ +"""202603261630 + +Revision ID: adaefcbe2aa1 +Revises: 1ea8fe97b5b7 +Create Date: 2026-03-26 16:27:17.590077 + +""" +from typing import Sequence, Union + +from alembic import op +import sqlalchemy as sa + + +# revision identifiers, used by Alembic. +revision: str = 'adaefcbe2aa1' +down_revision: Union[str, None] = '1ea8fe97b5b7' +branch_labels: Union[str, Sequence[str], None] = None +depends_on: Union[str, Sequence[str], None] = None + + +def upgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.add_column('tenants', sa.Column('feature_billing', sa.Boolean(), server_default='false', nullable=False, comment='是否启用收费管理菜单')) + op.add_column('tenants', sa.Column('feature_user_management', sa.Boolean(), server_default='false', nullable=False, comment='是否启用用户管理菜单')) + # ### end Alembic commands ### + + +def downgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.drop_column('tenants', 'feature_user_management') + op.drop_column('tenants', 'feature_billing') + # ### end Alembic commands ### diff --git a/api/migrations/versions/e28bcc212da5_202603241530.py b/api/migrations/versions/e28bcc212da5_202603241530.py new file mode 100644 index 00000000..00173522 --- /dev/null +++ b/api/migrations/versions/e28bcc212da5_202603241530.py @@ -0,0 +1,34 @@ +"""202603241530 + +Revision ID: e28bcc212da5 +Revises: 05a681a6ca93 +Create Date: 2026-03-24 15:32:14.461480 + +""" +from typing import Sequence, Union + +from alembic import op +import sqlalchemy as sa + + +# revision identifiers, used by Alembic. +revision: str = 'e28bcc212da5' +down_revision: Union[str, None] = '05a681a6ca93' +branch_labels: Union[str, Sequence[str], None] = None +depends_on: Union[str, Sequence[str], None] = None + + +def upgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.add_column('memory_config', sa.Column('vision_id', sa.String(), nullable=True, comment='视觉模型配置ID')) + op.add_column('memory_config', sa.Column('audio_id', sa.String(), nullable=True, comment='语音模型配置ID')) + op.add_column('memory_config', sa.Column('video_id', sa.String(), nullable=True, comment='视频模型配置ID')) + # ### end Alembic commands ### + + +def downgrade() -> None: + # ### commands auto generated by Alembic - please adjust! ### + op.drop_column('memory_config', 'video_id') + op.drop_column('memory_config', 'audio_id') + op.drop_column('memory_config', 'vision_id') + # ### end Alembic commands ### diff --git a/api/pyproject.toml b/api/pyproject.toml index e6fddea8..8ced574c 100644 --- a/api/pyproject.toml +++ b/api/pyproject.toml @@ -147,6 +147,7 @@ dependencies = [ "modelscope>=1.34.0", "python-magic>=0.4.14; sys_platform == 'linux' or sys_platform == 'darwin'", "python-magic-bin>=0.4.14; sys_platform=='win32'", + "volcengine-python-sdk[ark]==5.0.19" ] [tool.pytest.ini_options] diff --git a/api/tests/workflow/executor/test_vairable_pool.py b/api/tests/workflow/executor/test_vairable_pool.py index 3404eb79..0ba4d259 100644 --- a/api/tests/workflow/executor/test_vairable_pool.py +++ b/api/tests/workflow/executor/test_vairable_pool.py @@ -303,7 +303,7 @@ async def test_get_node_output_not_exist_with_default(): """测试获取不存在的节点输出(使用默认值)""" pool = VariablePool() - result = pool.get_node_output("nonexistent_node", defalut=None, strict=False) + result = pool.get_node_output("nonexistent_node", default=None, strict=False) assert result is None diff --git a/redbear-mem-benchmark b/redbear-mem-benchmark index c3bbc693..e853d99f 160000 --- a/redbear-mem-benchmark +++ b/redbear-mem-benchmark @@ -1 +1 @@ -Subproject commit c3bbc6931c570e6fac88c0b00658b4f08dc2ac77 +Subproject commit e853d99ff0d42ee81333db0fe0b6927536e4aa0e diff --git a/web/package.json b/web/package.json index db6a8408..0284f397 100644 --- a/web/package.json +++ b/web/package.json @@ -30,7 +30,7 @@ "@lexical/list": "^0.39.0", "@lexical/react": "^0.39.0", "@lexical/rich-text": "^0.39.0", - "antd": "^5.27.4", + "antd": "^5.29.2", "axios": "^1.12.2", "clsx": "^2.1.1", "codemirror": "^6.0.2", diff --git a/web/src/App.tsx b/web/src/App.tsx index 1d298358..a10f9409 100644 --- a/web/src/App.tsx +++ b/web/src/App.tsx @@ -21,7 +21,6 @@ import { useTranslation } from 'react-i18next'; import { lightTheme } from './styles/antdThemeConfig.ts' import router from './routes'; import { useI18n } from '@/store/locale' -import LayoutBg from '@/components/Layout/LayoutBg' import dayjs from 'dayjs' import 'dayjs/locale/en' import 'dayjs/locale/zh-cn' @@ -61,7 +60,6 @@ function App() { theme={lightTheme} > - }> { } }) } +// Get workspace API call statistics +export const getWorkspaceApiStatistics = (data: { start_date: number; end_date: number; }) => { + return request.get(`/apps/workspace/api-statistics`, data) +} // Export application export const appExport = (app_id: string, appName: string, data?: { release_id: string }) => { return request.getDownloadFile(`/apps/${app_id}/export`, `${appName}.yml`, data) @@ -165,4 +169,9 @@ export const cancelShare = (app_id: string, target_workspace_id?: string) => { export const cancelSpaceShare = (target_workspace_id?: string) => { return request.delete(`/apps/share/${target_workspace_id}`) } - +// Application conversation logs +export const getAppLogsUrl = (app_id: string) => `/apps/${app_id}/logs` +// Get full conversation message history +export const getAppLogDetail = (app_id: string, conversation_id: string) => { + return request.get(`/apps/${app_id}/logs/${conversation_id}`) +} \ No newline at end of file diff --git a/web/src/api/fileStorage.ts b/web/src/api/fileStorage.ts index ce133565..83f5b212 100644 --- a/web/src/api/fileStorage.ts +++ b/web/src/api/fileStorage.ts @@ -2,7 +2,7 @@ * @Author: ZhaoYing * @Date: 2026-02-03 13:59:56 * @Last Modified by: ZhaoYing - * @Last Modified time: 2026-02-09 16:24:05 + * @Last Modified time: 2026-03-23 18:05:43 */ import { request, API_PREFIX } from '@/utils/request' @@ -32,4 +32,13 @@ export const deleteFile = (fileId: string) => { } export const shareFileUploadUrlWithoutApiPrefix = `/storage/share/files` -export const shareFileUploadUrl = `${API_PREFIX}${shareFileUploadUrlWithoutApiPrefix}` \ No newline at end of file +export const shareFileUploadUrl = `${API_PREFIX}${shareFileUploadUrlWithoutApiPrefix}` + +// Get file info +export const getFileInfoByUrl = (url: string) => { + return request.get('/storage/files/info-by-url', {url}) +} +// Get file status +export const getFileStatusById = (file_id: string) => { + return request.get(`/storage/files/${file_id}/status`) +} \ No newline at end of file diff --git a/web/src/api/knowledgeBase.ts b/web/src/api/knowledgeBase.ts index 63ec80ae..05200221 100644 --- a/web/src/api/knowledgeBase.ts +++ b/web/src/api/knowledgeBase.ts @@ -68,8 +68,8 @@ export const getModelTypeList = async () => { return response as any[]; }; // 获取模型列表 -export const getModelList = async (pageInfo: PageRequest) => { - const response = await request.get(`${apiPrefix}/models`, { ...pageInfo, is_active: true }); +export const getModelList = async (types: string[], pageInfo: PageRequest) => { + const response = await request.get(`${apiPrefix}/models`, { ...pageInfo, type: types?.join(','), is_active: true }); return response as any; }; //获取模型提供者 diff --git a/web/src/api/memory.ts b/web/src/api/memory.ts index 9a464893..4467b649 100644 --- a/web/src/api/memory.ts +++ b/web/src/api/memory.ts @@ -2,7 +2,7 @@ * @Author: ZhaoYing * @Date: 2026-02-03 14:00:06 * @Last Modified by: ZhaoYing - * @Last Modified time: 2026-03-19 18:35:10 + * @Last Modified time: 2026-03-24 17:48:01 */ import { request } from '@/utils/request' import type { AxiosRequestConfig } from 'axios' @@ -87,12 +87,13 @@ export const getUserSummary = (end_user_id: string) => { export const getNodeStatistics = (end_user_id: string) => { return request.get(`/memory-storage/analytics/node_statistics`, { end_user_id }) } -// Basic information -export const getEndUserProfile = (end_user_id: string) => { - return request.get(`/memory-storage/read_end_user/profile`, { end_user_id }) +// 查询用户别名及信息 +export const getEndUserInfo = (end_user_id: string) => { + return request.get(`/memory-storage/end_user_info`, { end_user_id }) } -export const updatedEndUserProfile = (values: EndUser) => { - return request.post(`/memory-storage/updated_end_user/profile`, values) +// 更新用户别名及信息 +export const updatedEndUserInfo = (values: EndUser) => { + return request.post(`/memory-storage/end_user_info/updated`, values) } // User Memory - Relationship network export const getMemorySearchEdges = (end_user_id: string, config?: AxiosRequestConfig) => { @@ -153,6 +154,8 @@ export const analyticsRefresh = (end_user_id: string) => { export const getForgetStats = (end_user_id: string) => { return request.get(`/memory/forget-memory/stats`, { end_user_id }) } +// 获取带遗忘节点列表 +export const getForgetPendingNodesUrl = '/memory/forget-memory/pending-nodes' // Implicit Memory - Preferences export const getImplicitPreferences = (end_user_id: string) => { return request.get(`/memory/implicit-memory/preferences/${end_user_id}`) diff --git a/web/src/assets/font/MiSans/MiSans-Bold.woff2 b/web/src/assets/font/MiSans/MiSans-Bold.woff2 new file mode 100644 index 00000000..e4a21bee Binary files /dev/null and b/web/src/assets/font/MiSans/MiSans-Bold.woff2 differ diff --git a/web/src/assets/font/MiSans/MiSans-Demibold.woff2 b/web/src/assets/font/MiSans/MiSans-Demibold.woff2 new file mode 100644 index 00000000..70205afb Binary files /dev/null and b/web/src/assets/font/MiSans/MiSans-Demibold.woff2 differ diff --git a/web/src/assets/font/MiSans/MiSans-ExtraLight.woff2 b/web/src/assets/font/MiSans/MiSans-ExtraLight.woff2 new file mode 100644 index 00000000..45d16c98 Binary files /dev/null and b/web/src/assets/font/MiSans/MiSans-ExtraLight.woff2 differ diff --git a/web/src/assets/font/MiSans/MiSans-Heavy.woff2 b/web/src/assets/font/MiSans/MiSans-Heavy.woff2 new file mode 100644 index 00000000..09ee22e3 Binary files /dev/null and b/web/src/assets/font/MiSans/MiSans-Heavy.woff2 differ diff --git a/web/src/assets/font/MiSans/MiSans-Light.woff2 b/web/src/assets/font/MiSans/MiSans-Light.woff2 new file mode 100644 index 00000000..a2bb950b Binary files /dev/null and b/web/src/assets/font/MiSans/MiSans-Light.woff2 differ diff --git a/web/src/assets/font/MiSans/MiSans-Medium.woff2 b/web/src/assets/font/MiSans/MiSans-Medium.woff2 new file mode 100644 index 00000000..617f7407 Binary files /dev/null and b/web/src/assets/font/MiSans/MiSans-Medium.woff2 differ diff --git a/web/src/assets/font/MiSans/MiSans-Normal.woff2 b/web/src/assets/font/MiSans/MiSans-Normal.woff2 new file mode 100644 index 00000000..d24e89dd Binary files /dev/null and b/web/src/assets/font/MiSans/MiSans-Normal.woff2 differ diff --git a/web/src/assets/font/MiSans/MiSans-Regular.woff2 b/web/src/assets/font/MiSans/MiSans-Regular.woff2 new file mode 100644 index 00000000..6a699b50 Binary files /dev/null and b/web/src/assets/font/MiSans/MiSans-Regular.woff2 differ diff --git a/web/src/assets/font/MiSans/MiSans-Semibold.woff2 b/web/src/assets/font/MiSans/MiSans-Semibold.woff2 new file mode 100644 index 00000000..34f43f7c Binary files /dev/null and b/web/src/assets/font/MiSans/MiSans-Semibold.woff2 differ diff --git a/web/src/assets/font/MiSans/MiSans-Thin.woff2 b/web/src/assets/font/MiSans/MiSans-Thin.woff2 new file mode 100644 index 00000000..ec8a3b55 Binary files /dev/null and b/web/src/assets/font/MiSans/MiSans-Thin.woff2 differ diff --git a/web/src/assets/font/MiSans/index.ts b/web/src/assets/font/MiSans/index.ts new file mode 100644 index 00000000..e69de29b diff --git a/web/src/assets/images/CloudUploadOutlined.svg b/web/src/assets/images/CloudUploadOutlined.svg new file mode 100644 index 00000000..86fdf286 --- /dev/null +++ b/web/src/assets/images/CloudUploadOutlined.svg @@ -0,0 +1,26 @@ + + + 编组 12 + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/application/arrow_right.svg b/web/src/assets/images/application/arrow_right.svg new file mode 100644 index 00000000..06400efc --- /dev/null +++ b/web/src/assets/images/application/arrow_right.svg @@ -0,0 +1,17 @@ + + + 编组 25 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/application/clean.svg b/web/src/assets/images/application/clean.svg index 5d134404..a728abaa 100644 --- a/web/src/assets/images/application/clean.svg +++ b/web/src/assets/images/application/clean.svg @@ -1,13 +1,15 @@ 编组 11 - - - - - - - + + + + + + + + + diff --git a/web/src/assets/images/application/copy.svg b/web/src/assets/images/application/copy.svg new file mode 100644 index 00000000..1bd47c0b --- /dev/null +++ b/web/src/assets/images/application/copy.svg @@ -0,0 +1,18 @@ + + + 复制 + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/application/debuggingEmpty.png b/web/src/assets/images/application/debuggingEmpty.png index 0879d4e3..f5d4ef0d 100644 Binary files a/web/src/assets/images/application/debuggingEmpty.png and b/web/src/assets/images/application/debuggingEmpty.png differ diff --git a/web/src/assets/images/application/model.svg b/web/src/assets/images/application/model.svg index 4d482df5..a93f5771 100644 --- a/web/src/assets/images/application/model.svg +++ b/web/src/assets/images/application/model.svg @@ -1,12 +1,12 @@ -_模型预测 - - - - - - + + + + + + diff --git a/web/src/assets/images/application/model_hover.svg b/web/src/assets/images/application/model_hover.svg deleted file mode 100644 index 04e25219..00000000 --- a/web/src/assets/images/application/model_hover.svg +++ /dev/null @@ -1,15 +0,0 @@ - - - -_模型预测 - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/application/save.svg b/web/src/assets/images/application/save.svg new file mode 100644 index 00000000..02dbf635 --- /dev/null +++ b/web/src/assets/images/application/save.svg @@ -0,0 +1,19 @@ + + + 保存 + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/application/set.svg b/web/src/assets/images/application/set.svg new file mode 100644 index 00000000..797d2bad --- /dev/null +++ b/web/src/assets/images/application/set.svg @@ -0,0 +1,15 @@ + + + 设置-灰 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/close.svg b/web/src/assets/images/close.svg index cba672fc..d6e1a9b4 100644 --- a/web/src/assets/images/close.svg +++ b/web/src/assets/images/close.svg @@ -1,11 +1,11 @@ 关闭 - - - - - + + + + + diff --git a/web/src/assets/images/common/arrow_right_dark.svg b/web/src/assets/images/common/arrow_right_dark.svg new file mode 100644 index 00000000..b20a440c --- /dev/null +++ b/web/src/assets/images/common/arrow_right_dark.svg @@ -0,0 +1,18 @@ + + + 编组 5 + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/arrow_up.svg b/web/src/assets/images/common/arrow_up.svg new file mode 100644 index 00000000..a5105d46 --- /dev/null +++ b/web/src/assets/images/common/arrow_up.svg @@ -0,0 +1,14 @@ + + + 下拉 + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/caret_right_outlined.svg b/web/src/assets/images/common/caret_right_outlined.svg new file mode 100644 index 00000000..fcb3c68c --- /dev/null +++ b/web/src/assets/images/common/caret_right_outlined.svg @@ -0,0 +1,16 @@ + + + 编组 38 + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/check_green.svg b/web/src/assets/images/common/check_green.svg new file mode 100644 index 00000000..a16b1ee2 --- /dev/null +++ b/web/src/assets/images/common/check_green.svg @@ -0,0 +1,20 @@ + + + 完成 + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/copy_dark.svg b/web/src/assets/images/common/copy_dark.svg new file mode 100644 index 00000000..faa6fca1 --- /dev/null +++ b/web/src/assets/images/common/copy_dark.svg @@ -0,0 +1,14 @@ + + + 复制 + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/dash.svg b/web/src/assets/images/common/dash.svg new file mode 100644 index 00000000..cf9efb7d --- /dev/null +++ b/web/src/assets/images/common/dash.svg @@ -0,0 +1,15 @@ + + + 编组 27@3x + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/delete.svg b/web/src/assets/images/common/delete.svg new file mode 100644 index 00000000..4eb610ed --- /dev/null +++ b/web/src/assets/images/common/delete.svg @@ -0,0 +1,30 @@ + + + 编组 33 + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/delete_dark.svg b/web/src/assets/images/common/delete_dark.svg new file mode 100644 index 00000000..cf93cfd6 --- /dev/null +++ b/web/src/assets/images/common/delete_dark.svg @@ -0,0 +1,16 @@ + + + 删除 + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/delete_hover.svg b/web/src/assets/images/common/delete_hover.svg new file mode 100644 index 00000000..bf38179b --- /dev/null +++ b/web/src/assets/images/common/delete_hover.svg @@ -0,0 +1,20 @@ + + + 编组 33 + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/delete_red.svg b/web/src/assets/images/common/delete_red.svg new file mode 100644 index 00000000..58ad4d41 --- /dev/null +++ b/web/src/assets/images/common/delete_red.svg @@ -0,0 +1,30 @@ + + + 编组 33 + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/delete_red_big.svg b/web/src/assets/images/common/delete_red_big.svg new file mode 100644 index 00000000..7751b4e1 --- /dev/null +++ b/web/src/assets/images/common/delete_red_big.svg @@ -0,0 +1,19 @@ + + + 编组 33 + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/edit.svg b/web/src/assets/images/common/edit.svg new file mode 100644 index 00000000..cf00d703 --- /dev/null +++ b/web/src/assets/images/common/edit.svg @@ -0,0 +1,27 @@ + + + 编辑 + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/edit_bg.svg b/web/src/assets/images/common/edit_bg.svg new file mode 100644 index 00000000..4711afa4 --- /dev/null +++ b/web/src/assets/images/common/edit_bg.svg @@ -0,0 +1,17 @@ + + + 编辑 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/edit_bold.svg b/web/src/assets/images/common/edit_bold.svg new file mode 100644 index 00000000..c41984b2 --- /dev/null +++ b/web/src/assets/images/common/edit_bold.svg @@ -0,0 +1,16 @@ + + + 编辑 + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/eye.svg b/web/src/assets/images/common/eye.svg new file mode 100644 index 00000000..df2af1cf --- /dev/null +++ b/web/src/assets/images/common/eye.svg @@ -0,0 +1,16 @@ + + + 编辑 + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/eye_bg.svg b/web/src/assets/images/common/eye_bg.svg new file mode 100644 index 00000000..275c13c2 --- /dev/null +++ b/web/src/assets/images/common/eye_bg.svg @@ -0,0 +1,17 @@ + + + 编辑 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/global_outline.svg b/web/src/assets/images/common/global_outline.svg new file mode 100644 index 00000000..86301a0e --- /dev/null +++ b/web/src/assets/images/common/global_outline.svg @@ -0,0 +1,20 @@ + + + 互联网 + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/link.svg b/web/src/assets/images/common/link.svg new file mode 100644 index 00000000..5773d546 --- /dev/null +++ b/web/src/assets/images/common/link.svg @@ -0,0 +1,13 @@ + + + link-outlined + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/more.svg b/web/src/assets/images/common/more.svg new file mode 100644 index 00000000..6c24cf52 --- /dev/null +++ b/web/src/assets/images/common/more.svg @@ -0,0 +1,13 @@ + + + 卡片1@3x + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/more_hover.svg b/web/src/assets/images/common/more_hover.svg new file mode 100644 index 00000000..d08ba08f --- /dev/null +++ b/web/src/assets/images/common/more_hover.svg @@ -0,0 +1,25 @@ + + + 更多@3x + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/plus.svg b/web/src/assets/images/common/plus.svg new file mode 100644 index 00000000..5a2d7b83 --- /dev/null +++ b/web/src/assets/images/common/plus.svg @@ -0,0 +1,11 @@ + + + 形状结合@2x + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/plus_dark.svg b/web/src/assets/images/common/plus_dark.svg new file mode 100644 index 00000000..b0882a02 --- /dev/null +++ b/web/src/assets/images/common/plus_dark.svg @@ -0,0 +1,15 @@ + + + 编组 5 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/plus_grey.svg b/web/src/assets/images/common/plus_grey.svg new file mode 100644 index 00000000..05fb64e3 --- /dev/null +++ b/web/src/assets/images/common/plus_grey.svg @@ -0,0 +1,13 @@ + + + 形状结合@2x + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/question.svg b/web/src/assets/images/common/question.svg new file mode 100644 index 00000000..f8b0fee4 --- /dev/null +++ b/web/src/assets/images/common/question.svg @@ -0,0 +1,15 @@ + + + 问号小 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/return.svg b/web/src/assets/images/common/return.svg new file mode 100644 index 00000000..cb8166c0 --- /dev/null +++ b/web/src/assets/images/common/return.svg @@ -0,0 +1,17 @@ + + + 退出 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/common/save.svg b/web/src/assets/images/common/save.svg new file mode 100644 index 00000000..5970236d --- /dev/null +++ b/web/src/assets/images/common/save.svg @@ -0,0 +1,19 @@ + + + 保存 + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/conversation/analysisEmpty.png b/web/src/assets/images/conversation/analysisEmpty.png index 6d497f31..50adbd82 100644 Binary files a/web/src/assets/images/conversation/analysisEmpty.png and b/web/src/assets/images/conversation/analysisEmpty.png differ diff --git a/web/src/assets/images/conversation/audio.svg b/web/src/assets/images/conversation/audio.svg index 57a5ca49..c7c4e1fd 100644 --- a/web/src/assets/images/conversation/audio.svg +++ b/web/src/assets/images/conversation/audio.svg @@ -1,17 +1,28 @@ - - 编组 15 - - - - - - - - - - - + + 语音 + + + + + + + + + + + + + + + + + + + + + + diff --git a/web/src/assets/images/conversation/audio_ing.svg b/web/src/assets/images/conversation/audio_ing.svg deleted file mode 100644 index 280a1bd9..00000000 --- a/web/src/assets/images/conversation/audio_ing.svg +++ /dev/null @@ -1,21 +0,0 @@ - - - 编组 15 - - - - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/conversation/conversation.svg b/web/src/assets/images/conversation/conversation.svg index 2ebc02fb..a21f34bc 100644 --- a/web/src/assets/images/conversation/conversation.svg +++ b/web/src/assets/images/conversation/conversation.svg @@ -1,11 +1,12 @@ - + 对话 - - - - - + + + + + + diff --git a/web/src/assets/images/conversation/conversationEmpty.svg b/web/src/assets/images/conversation/conversationEmpty.svg index 2b642355..8320fd75 100644 --- a/web/src/assets/images/conversation/conversationEmpty.svg +++ b/web/src/assets/images/conversation/conversationEmpty.svg @@ -1,21 +1,23 @@ 编组 14 - - - - - - - - - - - - - - - + + + + + + + + + + + + + + + + + diff --git a/web/src/assets/images/conversation/deepThinking.svg b/web/src/assets/images/conversation/deepThinking.svg index b7658bf4..58ad411f 100644 --- a/web/src/assets/images/conversation/deepThinking.svg +++ b/web/src/assets/images/conversation/deepThinking.svg @@ -1,16 +1,29 @@ 深度思考 - - - - - - - - - - + + + + + + + + + + + + + + + + + + + + + + + diff --git a/web/src/assets/images/conversation/delete.svg b/web/src/assets/images/conversation/delete.svg index 27f1c15f..b46dea12 100644 --- a/web/src/assets/images/conversation/delete.svg +++ b/web/src/assets/images/conversation/delete.svg @@ -5,7 +5,7 @@ - + diff --git a/web/src/assets/images/conversation/exclamation_circle.svg b/web/src/assets/images/conversation/exclamation_circle.svg new file mode 100644 index 00000000..9b96bbce --- /dev/null +++ b/web/src/assets/images/conversation/exclamation_circle.svg @@ -0,0 +1,15 @@ + + + 告警实心 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/conversation/link.svg b/web/src/assets/images/conversation/link.svg index 18031b71..17298c10 100644 --- a/web/src/assets/images/conversation/link.svg +++ b/web/src/assets/images/conversation/link.svg @@ -1,18 +1,26 @@ - - 链接 - - - - - - - - - - - - + + 编组 6 + + + + + + + + + + + + + + + + + + + + diff --git a/web/src/assets/images/conversation/loading.svg b/web/src/assets/images/conversation/loading.svg index 01adc786..7bed9e7f 100644 --- a/web/src/assets/images/conversation/loading.svg +++ b/web/src/assets/images/conversation/loading.svg @@ -1,13 +1,24 @@ - - 编组 5 - - - - - - - + + 编组 14 + + + + + + + + + + + + + + + + + + diff --git a/web/src/assets/images/conversation/memoryFunction.svg b/web/src/assets/images/conversation/memoryFunction.svg index f63dc231..d0f3daf8 100644 --- a/web/src/assets/images/conversation/memoryFunction.svg +++ b/web/src/assets/images/conversation/memoryFunction.svg @@ -1,15 +1,26 @@ - brain-2-line - - - - - - - - - + 1 + + + + + + + + + + + + + + + + + + + + diff --git a/web/src/assets/images/conversation/memoryFunctionChecked.svg b/web/src/assets/images/conversation/memoryFunctionChecked.svg index db12f037..cf136428 100644 --- a/web/src/assets/images/conversation/memoryFunctionChecked.svg +++ b/web/src/assets/images/conversation/memoryFunctionChecked.svg @@ -1,14 +1,27 @@ - brain-2-line - - - - - - - - + 1 + + + + + + + + + + + + + + + + + + + + + diff --git a/web/src/assets/images/conversation/normalReply.svg b/web/src/assets/images/conversation/normalReply.svg new file mode 100644 index 00000000..19b8c28d --- /dev/null +++ b/web/src/assets/images/conversation/normalReply.svg @@ -0,0 +1,28 @@ + + + 正常 + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/conversation/online.svg b/web/src/assets/images/conversation/online.svg index 0ae567ca..c9c5812b 100644 --- a/web/src/assets/images/conversation/online.svg +++ b/web/src/assets/images/conversation/online.svg @@ -1,17 +1,28 @@ - 互联网 - - - - - - - - - - - + 联网 + + + + + + + + + + + + + + + + + + + + + + diff --git a/web/src/assets/images/conversation/onlineChecked.svg b/web/src/assets/images/conversation/onlineChecked.svg index 89fd61c4..fdd2b4b2 100644 --- a/web/src/assets/images/conversation/onlineChecked.svg +++ b/web/src/assets/images/conversation/onlineChecked.svg @@ -1,16 +1,29 @@ - 互联网 - - - - - - - - - - + 联网 + + + + + + + + + + + + + + + + + + + + + + + diff --git a/web/src/assets/images/conversation/quickReply.svg b/web/src/assets/images/conversation/quickReply.svg new file mode 100644 index 00000000..9a90ef1c --- /dev/null +++ b/web/src/assets/images/conversation/quickReply.svg @@ -0,0 +1,28 @@ + + + 快速回复 + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/conversation/redbear.png b/web/src/assets/images/conversation/redbear.png new file mode 100644 index 00000000..8fd5e2f6 Binary files /dev/null and b/web/src/assets/images/conversation/redbear.png differ diff --git a/web/src/assets/images/conversation/send.svg b/web/src/assets/images/conversation/send.svg index a44dcc40..5e9f5a21 100644 --- a/web/src/assets/images/conversation/send.svg +++ b/web/src/assets/images/conversation/send.svg @@ -1,15 +1,27 @@ - - 发送 - - - - - - - - - + + 发送-2@2x + + + + + + + + + + + + + + + + + + + + + diff --git a/web/src/assets/images/conversation/sendDisabled.svg b/web/src/assets/images/conversation/sendDisabled.svg index bf774bfd..7eb01380 100644 --- a/web/src/assets/images/conversation/sendDisabled.svg +++ b/web/src/assets/images/conversation/sendDisabled.svg @@ -1,16 +1,26 @@ - - 发送-2 - - - - - - - - - - + + 发送-2@2x + + + + + + + + + + + + + + + + + + + + diff --git a/web/src/assets/images/conversation/variables.svg b/web/src/assets/images/conversation/variables.svg new file mode 100644 index 00000000..e95c6922 --- /dev/null +++ b/web/src/assets/images/conversation/variables.svg @@ -0,0 +1,31 @@ + + + 变量 (1) + + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/copy_active.svg b/web/src/assets/images/copy_active.svg index 29a0f520..27f3c265 100644 --- a/web/src/assets/images/copy_active.svg +++ b/web/src/assets/images/copy_active.svg @@ -2,7 +2,7 @@ 复制 - + diff --git a/web/src/assets/images/deleteBg.svg b/web/src/assets/images/deleteBg.svg index 47deed9a..90409bdb 100644 --- a/web/src/assets/images/deleteBg.svg +++ b/web/src/assets/images/deleteBg.svg @@ -1,13 +1,13 @@ 编组 8 - - - - - - - + + + + + + + diff --git a/web/src/assets/images/deleteBorder.svg b/web/src/assets/images/deleteBorder.svg index 6e90bf4a..62b7bf96 100644 --- a/web/src/assets/images/deleteBorder.svg +++ b/web/src/assets/images/deleteBorder.svg @@ -1,12 +1,12 @@ 编组 8 - - - - - - + + + + + + diff --git a/web/src/assets/images/edit.svg b/web/src/assets/images/edit.svg index 67b90d2b..f503f005 100644 --- a/web/src/assets/images/edit.svg +++ b/web/src/assets/images/edit.svg @@ -2,7 +2,7 @@ 编辑 - + diff --git a/web/src/assets/images/editBg.svg b/web/src/assets/images/editBg.svg index 54ce218f..cfdaceef 100644 --- a/web/src/assets/images/editBg.svg +++ b/web/src/assets/images/editBg.svg @@ -1,13 +1,13 @@ 编组 13 - - - - - - - + + + + + + + diff --git a/web/src/assets/images/editBorder.svg b/web/src/assets/images/editBorder.svg index 6a0bd89f..4f6b0762 100644 --- a/web/src/assets/images/editBorder.svg +++ b/web/src/assets/images/editBorder.svg @@ -1,12 +1,12 @@ 编组 13 - - - - - - + + + + + + diff --git a/web/src/assets/images/edit_active.svg b/web/src/assets/images/edit_active.svg new file mode 100644 index 00000000..7beb376b --- /dev/null +++ b/web/src/assets/images/edit_active.svg @@ -0,0 +1,14 @@ + + + 编辑 + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/edit_hover.svg b/web/src/assets/images/edit_hover.svg index 6cb4e043..b69ed65a 100644 --- a/web/src/assets/images/edit_hover.svg +++ b/web/src/assets/images/edit_hover.svg @@ -2,7 +2,7 @@ 编辑 - + diff --git a/web/src/assets/images/empty/noData.png b/web/src/assets/images/empty/noData.png new file mode 100644 index 00000000..5258d466 Binary files /dev/null and b/web/src/assets/images/empty/noData.png differ diff --git a/web/src/assets/images/home/application.svg b/web/src/assets/images/home/application.svg new file mode 100644 index 00000000..65ce7ccd --- /dev/null +++ b/web/src/assets/images/home/application.svg @@ -0,0 +1,20 @@ + + + icon_应用管理 + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/home/arrow_top_right.svg b/web/src/assets/images/home/arrow_top_right.svg deleted file mode 100644 index fe969a19..00000000 --- a/web/src/assets/images/home/arrow_top_right.svg +++ /dev/null @@ -1,16 +0,0 @@ - - - 编组 16 - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/home/arrow_top_right_hover.svg b/web/src/assets/images/home/arrow_top_right_hover.svg deleted file mode 100644 index 903f9618..00000000 --- a/web/src/assets/images/home/arrow_top_right_hover.svg +++ /dev/null @@ -1,16 +0,0 @@ - - - 编组 16 - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/home/arrow_up.svg b/web/src/assets/images/home/arrow_up.svg new file mode 100644 index 00000000..914cb156 --- /dev/null +++ b/web/src/assets/images/home/arrow_up.svg @@ -0,0 +1,15 @@ + + + 箭头_向上 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/home/chunk_count.svg b/web/src/assets/images/home/chunk_count.svg index 830dac67..544f3cc6 100644 --- a/web/src/assets/images/home/chunk_count.svg +++ b/web/src/assets/images/home/chunk_count.svg @@ -1,22 +1,16 @@ 编组 32 - - - - - - - - - - - - - - - - + + + + + + + + + + diff --git a/web/src/assets/images/home/knowledge.svg b/web/src/assets/images/home/knowledge.svg new file mode 100644 index 00000000..91624510 --- /dev/null +++ b/web/src/assets/images/home/knowledge.svg @@ -0,0 +1,19 @@ + + + 知识库 + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/home/memoryConversation.svg b/web/src/assets/images/home/memoryConversation.svg new file mode 100644 index 00000000..59f74de2 --- /dev/null +++ b/web/src/assets/images/home/memoryConversation.svg @@ -0,0 +1,19 @@ + + + 编组 10 + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/home/statements_count.svg b/web/src/assets/images/home/statements_count.svg index a20666d1..6d545356 100644 --- a/web/src/assets/images/home/statements_count.svg +++ b/web/src/assets/images/home/statements_count.svg @@ -1,15 +1,18 @@ 编组 38 - - - - - - - - - + + + + + + + + + + + + diff --git a/web/src/assets/images/home/temporal_count.svg b/web/src/assets/images/home/temporal_count.svg index 050697bc..739acb30 100644 --- a/web/src/assets/images/home/temporal_count.svg +++ b/web/src/assets/images/home/temporal_count.svg @@ -1,17 +1,20 @@ 编组 39 - - - - - - - - - - - + + + + + + + + + + + + + + diff --git a/web/src/assets/images/home/totalMemoryCapacity.png b/web/src/assets/images/home/totalMemoryCapacity.png new file mode 100644 index 00000000..4a58cfad Binary files /dev/null and b/web/src/assets/images/home/totalMemoryCapacity.png differ diff --git a/web/src/assets/images/home/triplet_count.svg b/web/src/assets/images/home/triplet_count.svg index ebcfd0aa..603ede84 100644 --- a/web/src/assets/images/home/triplet_count.svg +++ b/web/src/assets/images/home/triplet_count.svg @@ -1,15 +1,14 @@ 编组 37 - - - - - - - - - + + + + + + + + diff --git a/web/src/assets/images/index/apps.svg b/web/src/assets/images/index/apps.svg index 58907fd6..b49bda51 100644 --- a/web/src/assets/images/index/apps.svg +++ b/web/src/assets/images/index/apps.svg @@ -1,14 +1,14 @@ 编组 34 - - - - + + + + - + diff --git a/web/src/assets/images/index/arrow_down.svg b/web/src/assets/images/index/arrow_down.svg index b77a3f8a..366e5848 100644 --- a/web/src/assets/images/index/arrow_down.svg +++ b/web/src/assets/images/index/arrow_down.svg @@ -1,10 +1,10 @@ 箭头_向上 - - - - + + + + diff --git a/web/src/assets/images/index/arrow_down_d.svg b/web/src/assets/images/index/arrow_down_d.svg index 40e5d94b..7393ca80 100644 --- a/web/src/assets/images/index/arrow_down_d.svg +++ b/web/src/assets/images/index/arrow_down_d.svg @@ -1,10 +1,10 @@ 编组 30 - - - - + + + + diff --git a/web/src/assets/images/index/arrow_up.svg b/web/src/assets/images/index/arrow_up.svg index 62aeee96..8a8bae53 100644 --- a/web/src/assets/images/index/arrow_up.svg +++ b/web/src/assets/images/index/arrow_up.svg @@ -1,10 +1,10 @@ 箭头_向上 - - - - + + + + diff --git a/web/src/assets/images/index/arrow_up_d.svg b/web/src/assets/images/index/arrow_up_d.svg index 3c19fef3..3529a291 100644 --- a/web/src/assets/images/index/arrow_up_d.svg +++ b/web/src/assets/images/index/arrow_up_d.svg @@ -1,10 +1,10 @@ 编组 30 - - - - + + + + diff --git a/web/src/assets/images/index/guide_bg@2x.png b/web/src/assets/images/index/guide_bg@2x.png index 3b7490fb..fbf452e6 100644 Binary files a/web/src/assets/images/index/guide_bg@2x.png and b/web/src/assets/images/index/guide_bg@2x.png differ diff --git a/web/src/assets/images/index/help_center.svg b/web/src/assets/images/index/help_center.svg index 6d272121..28595b0a 100644 --- a/web/src/assets/images/index/help_center.svg +++ b/web/src/assets/images/index/help_center.svg @@ -1,13 +1,13 @@ 编组 17 - - - - - - - + + + + + + + diff --git a/web/src/assets/images/index/index_bg@2x.png b/web/src/assets/images/index/index_bg@2x.png index d20ee4d3..fbf30083 100644 Binary files a/web/src/assets/images/index/index_bg@2x.png and b/web/src/assets/images/index/index_bg@2x.png differ diff --git a/web/src/assets/images/index/model_mgt.svg b/web/src/assets/images/index/model_mgt.svg index 89e13ec3..536f8877 100644 --- a/web/src/assets/images/index/model_mgt.svg +++ b/web/src/assets/images/index/model_mgt.svg @@ -1,26 +1,26 @@ 编组 25 - - - - - - + + + + + + - - - - - - - - - - - - - + + + + + + + + + + + + + diff --git a/web/src/assets/images/index/models.svg b/web/src/assets/images/index/models.svg index 890f240a..60863681 100644 --- a/web/src/assets/images/index/models.svg +++ b/web/src/assets/images/index/models.svg @@ -1,9 +1,9 @@ 编组 14 - - - + + + diff --git a/web/src/assets/images/index/space_mgt.svg b/web/src/assets/images/index/space_mgt.svg index af1db66c..a71f7431 100644 --- a/web/src/assets/images/index/space_mgt.svg +++ b/web/src/assets/images/index/space_mgt.svg @@ -1,13 +1,13 @@ 编组 26 - - - - - - - + + + + + + + diff --git a/web/src/assets/images/index/spaces.svg b/web/src/assets/images/index/spaces.svg index 1c61bc6b..e79eb113 100644 --- a/web/src/assets/images/index/spaces.svg +++ b/web/src/assets/images/index/spaces.svg @@ -7,10 +7,10 @@ - - - - + + + + diff --git a/web/src/assets/images/index/user_mgt.svg b/web/src/assets/images/index/user_mgt.svg index d53a97b9..4ec237aa 100644 --- a/web/src/assets/images/index/user_mgt.svg +++ b/web/src/assets/images/index/user_mgt.svg @@ -1,13 +1,13 @@ 编组 24 - - - - - - - + + + + + + + diff --git a/web/src/assets/images/index/users.svg b/web/src/assets/images/index/users.svg index 545d9636..bfb37872 100644 --- a/web/src/assets/images/index/users.svg +++ b/web/src/assets/images/index/users.svg @@ -1,10 +1,10 @@ 编组 33 - - - - + + + + diff --git a/web/src/assets/images/memory/arrow_right.svg b/web/src/assets/images/memory/arrow_right.svg index 0d17ec3b..090330e9 100644 --- a/web/src/assets/images/memory/arrow_right.svg +++ b/web/src/assets/images/memory/arrow_right.svg @@ -1,12 +1,12 @@ - 下拉备份 - - - - - - + 下拉 + + + + + + diff --git a/web/src/assets/images/memory/clock_orange.svg b/web/src/assets/images/memory/clock_orange.svg new file mode 100644 index 00000000..5c2b58cf --- /dev/null +++ b/web/src/assets/images/memory/clock_orange.svg @@ -0,0 +1,18 @@ + + + 时间戳 + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/memory/debug.svg b/web/src/assets/images/memory/debug.svg new file mode 100644 index 00000000..325a355a --- /dev/null +++ b/web/src/assets/images/memory/debug.svg @@ -0,0 +1,15 @@ + + + 配置管理 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menu/apiKey.png b/web/src/assets/images/menu/apiKey.png deleted file mode 100644 index 53d19428..00000000 Binary files a/web/src/assets/images/menu/apiKey.png and /dev/null differ diff --git a/web/src/assets/images/menu/apiKey_active.png b/web/src/assets/images/menu/apiKey_active.png deleted file mode 100644 index 4f8d1cfa..00000000 Binary files a/web/src/assets/images/menu/apiKey_active.png and /dev/null differ diff --git a/web/src/assets/images/menu/dashboard.svg b/web/src/assets/images/menu/dashboard.svg deleted file mode 100644 index 43e05b3a..00000000 --- a/web/src/assets/images/menu/dashboard.svg +++ /dev/null @@ -1,18 +0,0 @@ - - - 编组 27 - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/dashboard_active.svg b/web/src/assets/images/menu/dashboard_active.svg deleted file mode 100644 index 3f1bc65c..00000000 --- a/web/src/assets/images/menu/dashboard_active.svg +++ /dev/null @@ -1,18 +0,0 @@ - - - 编组 27 - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/knowledge.svg b/web/src/assets/images/menu/knowledge.svg deleted file mode 100644 index 3fc1ec0f..00000000 --- a/web/src/assets/images/menu/knowledge.svg +++ /dev/null @@ -1,19 +0,0 @@ - - - 知识库 - - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/knowledge_active.svg b/web/src/assets/images/menu/knowledge_active.svg deleted file mode 100644 index 9b09bbf4..00000000 --- a/web/src/assets/images/menu/knowledge_active.svg +++ /dev/null @@ -1,19 +0,0 @@ - - - 知识库 - - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/member.svg b/web/src/assets/images/menu/member.svg deleted file mode 100644 index 56cca8c1..00000000 --- a/web/src/assets/images/menu/member.svg +++ /dev/null @@ -1,18 +0,0 @@ - - - 用户总数总计 - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/member_active.svg b/web/src/assets/images/menu/member_active.svg deleted file mode 100644 index 30cf9261..00000000 --- a/web/src/assets/images/menu/member_active.svg +++ /dev/null @@ -1,18 +0,0 @@ - - - 用户总数总计 - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/memory.svg b/web/src/assets/images/menu/memory.svg deleted file mode 100644 index 71696861..00000000 --- a/web/src/assets/images/menu/memory.svg +++ /dev/null @@ -1,16 +0,0 @@ - - - brain-2-line - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/memoryConversation.svg b/web/src/assets/images/menu/memoryConversation.svg deleted file mode 100644 index 369cbc5a..00000000 --- a/web/src/assets/images/menu/memoryConversation.svg +++ /dev/null @@ -1,19 +0,0 @@ - - - 编组 10 - - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/memoryConversation_active.svg b/web/src/assets/images/menu/memoryConversation_active.svg deleted file mode 100644 index c79a75f6..00000000 --- a/web/src/assets/images/menu/memoryConversation_active.svg +++ /dev/null @@ -1,19 +0,0 @@ - - - 编组 10 - - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/memory_active.svg b/web/src/assets/images/menu/memory_active.svg deleted file mode 100644 index eabe9221..00000000 --- a/web/src/assets/images/menu/memory_active.svg +++ /dev/null @@ -1,16 +0,0 @@ - - - brain-2-line - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/model.svg b/web/src/assets/images/menu/model.svg deleted file mode 100644 index bbb7e103..00000000 --- a/web/src/assets/images/menu/model.svg +++ /dev/null @@ -1,13 +0,0 @@ - - - -_模型预测 - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/model_active.svg b/web/src/assets/images/menu/model_active.svg deleted file mode 100644 index 274b146e..00000000 --- a/web/src/assets/images/menu/model_active.svg +++ /dev/null @@ -1,13 +0,0 @@ - - - -_模型预测 - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/ontology.svg b/web/src/assets/images/menu/ontology.svg deleted file mode 100644 index 9bfda42b..00000000 --- a/web/src/assets/images/menu/ontology.svg +++ /dev/null @@ -1,11 +0,0 @@ - - - 本体管理备份 - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/ontology_active.svg b/web/src/assets/images/menu/ontology_active.svg deleted file mode 100644 index 1271c2c3..00000000 --- a/web/src/assets/images/menu/ontology_active.svg +++ /dev/null @@ -1,11 +0,0 @@ - - - 本体管理 - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/pricing.svg b/web/src/assets/images/menu/pricing.svg deleted file mode 100644 index 5510ba23..00000000 --- a/web/src/assets/images/menu/pricing.svg +++ /dev/null @@ -1,22 +0,0 @@ - - - 菜单-收费管理 - - - - - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/pricing_active.svg b/web/src/assets/images/menu/pricing_active.svg deleted file mode 100644 index f708877d..00000000 --- a/web/src/assets/images/menu/pricing_active.svg +++ /dev/null @@ -1,22 +0,0 @@ - - - 菜单-收费管理 - - - - - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/prompt.svg b/web/src/assets/images/menu/prompt.svg deleted file mode 100644 index ffef9a34..00000000 --- a/web/src/assets/images/menu/prompt.svg +++ /dev/null @@ -1,15 +0,0 @@ - - - 提示词备份 - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/prompt_active.svg b/web/src/assets/images/menu/prompt_active.svg deleted file mode 100644 index ac45e13c..00000000 --- a/web/src/assets/images/menu/prompt_active.svg +++ /dev/null @@ -1,15 +0,0 @@ - - - 提示词 - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/skills.svg b/web/src/assets/images/menu/skills.svg deleted file mode 100644 index ac121d1e..00000000 --- a/web/src/assets/images/menu/skills.svg +++ /dev/null @@ -1,14 +0,0 @@ - - - 技能点 - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/skills_active.svg b/web/src/assets/images/menu/skills_active.svg deleted file mode 100644 index 789b5586..00000000 --- a/web/src/assets/images/menu/skills_active.svg +++ /dev/null @@ -1,14 +0,0 @@ - - - 技能点备份 - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/space.svg b/web/src/assets/images/menu/space.svg deleted file mode 100644 index c82c7922..00000000 --- a/web/src/assets/images/menu/space.svg +++ /dev/null @@ -1,15 +0,0 @@ - - - 模型管理 - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/spaceConfig.svg b/web/src/assets/images/menu/spaceConfig.svg deleted file mode 100644 index bcfeae12..00000000 --- a/web/src/assets/images/menu/spaceConfig.svg +++ /dev/null @@ -1,17 +0,0 @@ - - - 模型 (1) - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/spaceConfig_active.svg b/web/src/assets/images/menu/spaceConfig_active.svg deleted file mode 100644 index 41b25689..00000000 --- a/web/src/assets/images/menu/spaceConfig_active.svg +++ /dev/null @@ -1,17 +0,0 @@ - - - 模型 (1) - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/space_active.svg b/web/src/assets/images/menu/space_active.svg deleted file mode 100644 index 69b1629c..00000000 --- a/web/src/assets/images/menu/space_active.svg +++ /dev/null @@ -1,15 +0,0 @@ - - - 模型管理 - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/tool.png b/web/src/assets/images/menu/tool.png deleted file mode 100644 index 669238e8..00000000 Binary files a/web/src/assets/images/menu/tool.png and /dev/null differ diff --git a/web/src/assets/images/menu/tool_active.png b/web/src/assets/images/menu/tool_active.png deleted file mode 100644 index 252cd702..00000000 Binary files a/web/src/assets/images/menu/tool_active.png and /dev/null differ diff --git a/web/src/assets/images/menu/user.svg b/web/src/assets/images/menu/user.svg deleted file mode 100644 index b1eaf5b9..00000000 --- a/web/src/assets/images/menu/user.svg +++ /dev/null @@ -1,13 +0,0 @@ - - - 138设置、系统设置、功能设置、属性 - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/userMemory1.svg b/web/src/assets/images/menu/userMemory1.svg deleted file mode 100644 index c4b9cd51..00000000 --- a/web/src/assets/images/menu/userMemory1.svg +++ /dev/null @@ -1,18 +0,0 @@ - - - 编组 29 - - - - - - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menu/user_active.svg b/web/src/assets/images/menu/user_active.svg deleted file mode 100644 index 38de2069..00000000 --- a/web/src/assets/images/menu/user_active.svg +++ /dev/null @@ -1,13 +0,0 @@ - - - 138设置、系统设置、功能设置、属性 - - - - - - - - - - \ No newline at end of file diff --git a/web/src/assets/images/menuNew/apiKey.svg b/web/src/assets/images/menuNew/apiKey.svg new file mode 100644 index 00000000..c31e2d5c --- /dev/null +++ b/web/src/assets/images/menuNew/apiKey.svg @@ -0,0 +1,13 @@ + + + api + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/apiKey_active.svg b/web/src/assets/images/menuNew/apiKey_active.svg new file mode 100644 index 00000000..7520cb86 --- /dev/null +++ b/web/src/assets/images/menuNew/apiKey_active.svg @@ -0,0 +1,13 @@ + + + api + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menu/application.svg b/web/src/assets/images/menuNew/application.svg similarity index 65% rename from web/src/assets/images/menu/application.svg rename to web/src/assets/images/menuNew/application.svg index 37967d3a..a8fe8fc0 100644 --- a/web/src/assets/images/menu/application.svg +++ b/web/src/assets/images/menuNew/application.svg @@ -1,11 +1,11 @@ 应用管理 - - - - - + + + + + diff --git a/web/src/assets/images/menu/application_active.svg b/web/src/assets/images/menuNew/application_active.svg similarity index 65% rename from web/src/assets/images/menu/application_active.svg rename to web/src/assets/images/menuNew/application_active.svg index 3fe48200..0d8f91f9 100644 --- a/web/src/assets/images/menu/application_active.svg +++ b/web/src/assets/images/menuNew/application_active.svg @@ -1,11 +1,11 @@ 应用管理 - - - - - + + + + + diff --git a/web/src/assets/images/menuNew/dashboard.svg b/web/src/assets/images/menuNew/dashboard.svg new file mode 100644 index 00000000..d35e35fb --- /dev/null +++ b/web/src/assets/images/menuNew/dashboard.svg @@ -0,0 +1,18 @@ + + + 编组 27 + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/dashboard_active.svg b/web/src/assets/images/menuNew/dashboard_active.svg new file mode 100644 index 00000000..4a0f57b6 --- /dev/null +++ b/web/src/assets/images/menuNew/dashboard_active.svg @@ -0,0 +1,18 @@ + + + 编组 27 + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/knowledge.svg b/web/src/assets/images/menuNew/knowledge.svg new file mode 100644 index 00000000..2d7a28de --- /dev/null +++ b/web/src/assets/images/menuNew/knowledge.svg @@ -0,0 +1,20 @@ + + + 知识库 + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/knowledge_active.svg b/web/src/assets/images/menuNew/knowledge_active.svg new file mode 100644 index 00000000..0a2fba96 --- /dev/null +++ b/web/src/assets/images/menuNew/knowledge_active.svg @@ -0,0 +1,20 @@ + + + 知识库 + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/member.svg b/web/src/assets/images/menuNew/member.svg new file mode 100644 index 00000000..35edbe1a --- /dev/null +++ b/web/src/assets/images/menuNew/member.svg @@ -0,0 +1,18 @@ + + + 用户总数总计 + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/member_active.svg b/web/src/assets/images/menuNew/member_active.svg new file mode 100644 index 00000000..96269cd5 --- /dev/null +++ b/web/src/assets/images/menuNew/member_active.svg @@ -0,0 +1,18 @@ + + + 用户总数总计 + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/memory.svg b/web/src/assets/images/menuNew/memory.svg new file mode 100644 index 00000000..17c8368b --- /dev/null +++ b/web/src/assets/images/menuNew/memory.svg @@ -0,0 +1,16 @@ + + + brain-2-line + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/memoryConversation.svg b/web/src/assets/images/menuNew/memoryConversation.svg new file mode 100644 index 00000000..f74146b0 --- /dev/null +++ b/web/src/assets/images/menuNew/memoryConversation.svg @@ -0,0 +1,13 @@ + + + 对话 + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/memoryConversation_active.svg b/web/src/assets/images/menuNew/memoryConversation_active.svg new file mode 100644 index 00000000..c2c4aae3 --- /dev/null +++ b/web/src/assets/images/menuNew/memoryConversation_active.svg @@ -0,0 +1,13 @@ + + + 对话 + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/memory_active.svg b/web/src/assets/images/menuNew/memory_active.svg new file mode 100644 index 00000000..3aa5ff94 --- /dev/null +++ b/web/src/assets/images/menuNew/memory_active.svg @@ -0,0 +1,16 @@ + + + brain-2-line + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/menuFold.svg b/web/src/assets/images/menuNew/menuFold.svg new file mode 100644 index 00000000..3350cfc4 --- /dev/null +++ b/web/src/assets/images/menuNew/menuFold.svg @@ -0,0 +1,15 @@ + + + 收起 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/model.svg b/web/src/assets/images/menuNew/model.svg new file mode 100644 index 00000000..8fdc015a --- /dev/null +++ b/web/src/assets/images/menuNew/model.svg @@ -0,0 +1,13 @@ + + + -_模型预测 + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/model_active.svg b/web/src/assets/images/menuNew/model_active.svg new file mode 100644 index 00000000..6145f360 --- /dev/null +++ b/web/src/assets/images/menuNew/model_active.svg @@ -0,0 +1,11 @@ + + + -_模型预测 + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/ontology.svg b/web/src/assets/images/menuNew/ontology.svg new file mode 100644 index 00000000..68798ccd --- /dev/null +++ b/web/src/assets/images/menuNew/ontology.svg @@ -0,0 +1,17 @@ + + + 本体管理 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/ontology_active.svg b/web/src/assets/images/menuNew/ontology_active.svg new file mode 100644 index 00000000..f05a1069 --- /dev/null +++ b/web/src/assets/images/menuNew/ontology_active.svg @@ -0,0 +1,17 @@ + + + 本体管理 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/pricing.svg b/web/src/assets/images/menuNew/pricing.svg new file mode 100644 index 00000000..8c412ac0 --- /dev/null +++ b/web/src/assets/images/menuNew/pricing.svg @@ -0,0 +1,13 @@ + + + 收费管理 + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/pricing_active.svg b/web/src/assets/images/menuNew/pricing_active.svg new file mode 100644 index 00000000..54a0afb4 --- /dev/null +++ b/web/src/assets/images/menuNew/pricing_active.svg @@ -0,0 +1,11 @@ + + + 收费管理 + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/prompt.svg b/web/src/assets/images/menuNew/prompt.svg new file mode 100644 index 00000000..8007982b --- /dev/null +++ b/web/src/assets/images/menuNew/prompt.svg @@ -0,0 +1,21 @@ + + + 提示词 + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/prompt_active.svg b/web/src/assets/images/menuNew/prompt_active.svg new file mode 100644 index 00000000..4c94cac2 --- /dev/null +++ b/web/src/assets/images/menuNew/prompt_active.svg @@ -0,0 +1,21 @@ + + + 提示词 + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/skills.svg b/web/src/assets/images/menuNew/skills.svg new file mode 100644 index 00000000..3c8dd525 --- /dev/null +++ b/web/src/assets/images/menuNew/skills.svg @@ -0,0 +1,18 @@ + + + skills-icon + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/skills_active.svg b/web/src/assets/images/menuNew/skills_active.svg new file mode 100644 index 00000000..86191a8a --- /dev/null +++ b/web/src/assets/images/menuNew/skills_active.svg @@ -0,0 +1,16 @@ + + + skills-icon + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/space.svg b/web/src/assets/images/menuNew/space.svg new file mode 100644 index 00000000..d0e7a5e4 --- /dev/null +++ b/web/src/assets/images/menuNew/space.svg @@ -0,0 +1,15 @@ + + + 模型管理 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/spaceConfig.svg b/web/src/assets/images/menuNew/spaceConfig.svg new file mode 100644 index 00000000..f03b2f05 --- /dev/null +++ b/web/src/assets/images/menuNew/spaceConfig.svg @@ -0,0 +1,19 @@ + + + 空间配置 + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/spaceConfig_active.svg b/web/src/assets/images/menuNew/spaceConfig_active.svg new file mode 100644 index 00000000..578963a0 --- /dev/null +++ b/web/src/assets/images/menuNew/spaceConfig_active.svg @@ -0,0 +1,19 @@ + + + 空间配置 + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/space_active.svg b/web/src/assets/images/menuNew/space_active.svg new file mode 100644 index 00000000..e55efb3e --- /dev/null +++ b/web/src/assets/images/menuNew/space_active.svg @@ -0,0 +1,13 @@ + + + 模型管理 + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/tool.svg b/web/src/assets/images/menuNew/tool.svg new file mode 100644 index 00000000..0a14a626 --- /dev/null +++ b/web/src/assets/images/menuNew/tool.svg @@ -0,0 +1,18 @@ + + + 工具管理 (2) + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/tool_active.svg b/web/src/assets/images/menuNew/tool_active.svg new file mode 100644 index 00000000..00544dac --- /dev/null +++ b/web/src/assets/images/menuNew/tool_active.svg @@ -0,0 +1,16 @@ + + + 工具管理 (2) + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menuNew/user.svg b/web/src/assets/images/menuNew/user.svg new file mode 100644 index 00000000..d04fb501 --- /dev/null +++ b/web/src/assets/images/menuNew/user.svg @@ -0,0 +1,13 @@ + + + 用户管理 + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/menu/userMemory.svg b/web/src/assets/images/menuNew/userMemory.svg similarity index 76% rename from web/src/assets/images/menu/userMemory.svg rename to web/src/assets/images/menuNew/userMemory.svg index c4b9cd51..9eb5b1fc 100644 --- a/web/src/assets/images/menu/userMemory.svg +++ b/web/src/assets/images/menuNew/userMemory.svg @@ -1,11 +1,11 @@ 编组 29 - - - - - + + + + + diff --git a/web/src/assets/images/menu/userMemory_active.svg b/web/src/assets/images/menuNew/userMemory_active.svg similarity index 76% rename from web/src/assets/images/menu/userMemory_active.svg rename to web/src/assets/images/menuNew/userMemory_active.svg index 554dc0bc..d31e4859 100644 --- a/web/src/assets/images/menu/userMemory_active.svg +++ b/web/src/assets/images/menuNew/userMemory_active.svg @@ -1,11 +1,11 @@ 编组 29 - - - - - + + + + + diff --git a/web/src/assets/images/menuNew/user_active.svg b/web/src/assets/images/menuNew/user_active.svg new file mode 100644 index 00000000..33778047 --- /dev/null +++ b/web/src/assets/images/menuNew/user_active.svg @@ -0,0 +1,11 @@ + + + 用户管理 + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/model/bedrock.png b/web/src/assets/images/model/bedrock.png new file mode 100644 index 00000000..a16ee6f7 Binary files /dev/null and b/web/src/assets/images/model/bedrock.png differ diff --git a/web/src/assets/images/model/bedrock.svg b/web/src/assets/images/model/bedrock.svg deleted file mode 100644 index 6a0235af..00000000 --- a/web/src/assets/images/model/bedrock.svg +++ /dev/null @@ -1,15 +0,0 @@ - - - - - - - - - - - - - - - diff --git a/web/src/assets/images/model/dashscope.png b/web/src/assets/images/model/dashscope.png index c1aff40e..e57821f0 100644 Binary files a/web/src/assets/images/model/dashscope.png and b/web/src/assets/images/model/dashscope.png differ diff --git a/web/src/assets/images/model/gpustack.png b/web/src/assets/images/model/gpustack.png index b154821d..39d303ae 100644 Binary files a/web/src/assets/images/model/gpustack.png and b/web/src/assets/images/model/gpustack.png differ diff --git a/web/src/assets/images/model/ollama.png b/web/src/assets/images/model/ollama.png new file mode 100644 index 00000000..068d066d Binary files /dev/null and b/web/src/assets/images/model/ollama.png differ diff --git a/web/src/assets/images/model/ollama.svg b/web/src/assets/images/model/ollama.svg deleted file mode 100644 index f8482a96..00000000 --- a/web/src/assets/images/model/ollama.svg +++ /dev/null @@ -1,15 +0,0 @@ - - - - - - - - - - - - - - - diff --git a/web/src/assets/images/model/openai.png b/web/src/assets/images/model/openai.png new file mode 100644 index 00000000..db9fabaa Binary files /dev/null and b/web/src/assets/images/model/openai.png differ diff --git a/web/src/assets/images/model/openai.svg b/web/src/assets/images/model/openai.svg deleted file mode 100644 index 70686f9b..00000000 --- a/web/src/assets/images/model/openai.svg +++ /dev/null @@ -1,4 +0,0 @@ - - - - diff --git a/web/src/assets/images/model/volcano.png b/web/src/assets/images/model/volcano.png new file mode 100644 index 00000000..ba0dce10 Binary files /dev/null and b/web/src/assets/images/model/volcano.png differ diff --git a/web/src/assets/images/model/xinference.png b/web/src/assets/images/model/xinference.png new file mode 100644 index 00000000..71a4821b Binary files /dev/null and b/web/src/assets/images/model/xinference.png differ diff --git a/web/src/assets/images/model/xinference.svg b/web/src/assets/images/model/xinference.svg deleted file mode 100644 index f5c5f75e..00000000 --- a/web/src/assets/images/model/xinference.svg +++ /dev/null @@ -1,24 +0,0 @@ - - - - - - - - - - - - - - - - - - - - - - - - diff --git a/web/src/assets/images/prompt/delete.svg b/web/src/assets/images/prompt/delete.svg new file mode 100644 index 00000000..f413ffa0 --- /dev/null +++ b/web/src/assets/images/prompt/delete.svg @@ -0,0 +1,19 @@ + + + 编组 33 + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/prompt/delete_hover.svg b/web/src/assets/images/prompt/delete_hover.svg new file mode 100644 index 00000000..aebdc48c --- /dev/null +++ b/web/src/assets/images/prompt/delete_hover.svg @@ -0,0 +1,20 @@ + + + 编组 33 + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/prompt/edit.svg b/web/src/assets/images/prompt/edit.svg new file mode 100644 index 00000000..89668678 --- /dev/null +++ b/web/src/assets/images/prompt/edit.svg @@ -0,0 +1,16 @@ + + + 编辑 + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/prompt/edit_bg.svg b/web/src/assets/images/prompt/edit_bg.svg new file mode 100644 index 00000000..4711afa4 --- /dev/null +++ b/web/src/assets/images/prompt/edit_bg.svg @@ -0,0 +1,17 @@ + + + 编辑 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/prompt/eye.svg b/web/src/assets/images/prompt/eye.svg new file mode 100644 index 00000000..df2af1cf --- /dev/null +++ b/web/src/assets/images/prompt/eye.svg @@ -0,0 +1,16 @@ + + + 编辑 + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/prompt/eye_bg.svg b/web/src/assets/images/prompt/eye_bg.svg new file mode 100644 index 00000000..275c13c2 --- /dev/null +++ b/web/src/assets/images/prompt/eye_bg.svg @@ -0,0 +1,17 @@ + + + 编辑 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/question.svg b/web/src/assets/images/question.svg new file mode 100644 index 00000000..539ab03a --- /dev/null +++ b/web/src/assets/images/question.svg @@ -0,0 +1,17 @@ + + + 问号小 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/refresh.svg b/web/src/assets/images/refresh.svg index c592feff..79d2f836 100644 --- a/web/src/assets/images/refresh.svg +++ b/web/src/assets/images/refresh.svg @@ -1,12 +1,16 @@ - - 刷新 - - - - - - + + 编组 28 + + + + + + + + + + diff --git a/web/src/assets/images/refresh_hover.svg b/web/src/assets/images/refresh_dark.svg similarity index 97% rename from web/src/assets/images/refresh_hover.svg rename to web/src/assets/images/refresh_dark.svg index 1d4dcf7c..07864e99 100644 --- a/web/src/assets/images/refresh_hover.svg +++ b/web/src/assets/images/refresh_dark.svg @@ -2,7 +2,7 @@ 刷新 - + diff --git a/web/src/assets/images/tool/market.png b/web/src/assets/images/tool/market.png new file mode 100644 index 00000000..9639e253 Binary files /dev/null and b/web/src/assets/images/tool/market.png differ diff --git a/web/src/assets/images/userMemory/aboutMe.svg b/web/src/assets/images/userMemory/aboutMe.svg new file mode 100644 index 00000000..16630fdb --- /dev/null +++ b/web/src/assets/images/userMemory/aboutMe.svg @@ -0,0 +1,18 @@ + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/aboutMe_active.svg b/web/src/assets/images/userMemory/aboutMe_active.svg new file mode 100644 index 00000000..53e6362e --- /dev/null +++ b/web/src/assets/images/userMemory/aboutMe_active.svg @@ -0,0 +1,18 @@ + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/aboutUs.svg b/web/src/assets/images/userMemory/aboutUs.svg index 1d75eeae..b8fa9e45 100644 --- a/web/src/assets/images/userMemory/aboutUs.svg +++ b/web/src/assets/images/userMemory/aboutUs.svg @@ -1,13 +1,15 @@ - + - - - - - - - + + + + + + + + + diff --git a/web/src/assets/images/userMemory/ai.png b/web/src/assets/images/userMemory/ai.png new file mode 100644 index 00000000..3783a543 Binary files /dev/null and b/web/src/assets/images/userMemory/ai.png differ diff --git a/web/src/assets/images/userMemory/arrow_right.svg b/web/src/assets/images/userMemory/arrow_right.svg index aca820f8..3fa0eb49 100644 --- a/web/src/assets/images/userMemory/arrow_right.svg +++ b/web/src/assets/images/userMemory/arrow_right.svg @@ -1,12 +1,14 @@ 编组 5 - - - - - - + + + + + + + + diff --git a/web/src/assets/images/userMemory/arrow_right_dark.svg b/web/src/assets/images/userMemory/arrow_right_dark.svg new file mode 100644 index 00000000..38cfd953 --- /dev/null +++ b/web/src/assets/images/userMemory/arrow_right_dark.svg @@ -0,0 +1,16 @@ + + + 编组 5 + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/arrow_right_hover.svg b/web/src/assets/images/userMemory/arrow_right_hover.svg index 0fed7c6b..444a7a03 100644 --- a/web/src/assets/images/userMemory/arrow_right_hover.svg +++ b/web/src/assets/images/userMemory/arrow_right_hover.svg @@ -1,12 +1,14 @@ 编组 5 - - - - - - + + + + + + + + diff --git a/web/src/assets/images/userMemory/chat.svg b/web/src/assets/images/userMemory/chat.svg new file mode 100644 index 00000000..11b34345 --- /dev/null +++ b/web/src/assets/images/userMemory/chat.svg @@ -0,0 +1,17 @@ + + + 编组 61 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/close.svg b/web/src/assets/images/userMemory/close.svg new file mode 100644 index 00000000..1b511252 --- /dev/null +++ b/web/src/assets/images/userMemory/close.svg @@ -0,0 +1,13 @@ + + + 关闭 + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/down.svg b/web/src/assets/images/userMemory/down.svg index ae263f65..07a70e0d 100644 --- a/web/src/assets/images/userMemory/down.svg +++ b/web/src/assets/images/userMemory/down.svg @@ -1,13 +1,15 @@ - 下拉备份 - - - - - - - + 下拉 + + + + + + + + + diff --git a/web/src/assets/images/userMemory/download.svg b/web/src/assets/images/userMemory/download.svg new file mode 100644 index 00000000..1aa4f1ac --- /dev/null +++ b/web/src/assets/images/userMemory/download.svg @@ -0,0 +1,21 @@ + + + 更多 + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/download_hover.svg b/web/src/assets/images/userMemory/download_hover.svg new file mode 100644 index 00000000..5079a1ff --- /dev/null +++ b/web/src/assets/images/userMemory/download_hover.svg @@ -0,0 +1,22 @@ + + + 更多 + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/entity.svg b/web/src/assets/images/userMemory/entity.svg new file mode 100644 index 00000000..ad6a2692 --- /dev/null +++ b/web/src/assets/images/userMemory/entity.svg @@ -0,0 +1,20 @@ + + + 编组 5 + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/file.svg b/web/src/assets/images/userMemory/file.svg new file mode 100644 index 00000000..6bfd562e --- /dev/null +++ b/web/src/assets/images/userMemory/file.svg @@ -0,0 +1,85 @@ + + + 编组 9 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + TEXT + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/forget.png b/web/src/assets/images/userMemory/forget.png new file mode 100644 index 00000000..f38ff9bd Binary files /dev/null and b/web/src/assets/images/userMemory/forget.png differ diff --git a/web/src/assets/images/userMemory/interestDistribution.svg b/web/src/assets/images/userMemory/interestDistribution.svg index f39d3bb3..d26a9952 100644 --- a/web/src/assets/images/userMemory/interestDistribution.svg +++ b/web/src/assets/images/userMemory/interestDistribution.svg @@ -1,16 +1,18 @@ - + 兴趣爱好 - - - - - - - - - - + + + + + + + + + + + + diff --git a/web/src/assets/images/userMemory/interestDistribution_active.svg b/web/src/assets/images/userMemory/interestDistribution_active.svg new file mode 100644 index 00000000..87b8d548 --- /dev/null +++ b/web/src/assets/images/userMemory/interestDistribution_active.svg @@ -0,0 +1,21 @@ + + + 兴趣爱好 + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/logo.png b/web/src/assets/images/userMemory/logo.png new file mode 100644 index 00000000..ab37dda6 Binary files /dev/null and b/web/src/assets/images/userMemory/logo.png differ diff --git a/web/src/assets/images/userMemory/logout.svg b/web/src/assets/images/userMemory/logout.svg new file mode 100644 index 00000000..8c21f4e2 --- /dev/null +++ b/web/src/assets/images/userMemory/logout.svg @@ -0,0 +1,13 @@ + + + 退出 (1) + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/long_term_number.svg b/web/src/assets/images/userMemory/long_term_number.svg new file mode 100644 index 00000000..134af714 --- /dev/null +++ b/web/src/assets/images/userMemory/long_term_number.svg @@ -0,0 +1,19 @@ + + + 编组 5 + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/me.svg b/web/src/assets/images/userMemory/me.svg new file mode 100644 index 00000000..b8fa9e45 --- /dev/null +++ b/web/src/assets/images/userMemory/me.svg @@ -0,0 +1,18 @@ + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/memoryInsight.svg b/web/src/assets/images/userMemory/memoryInsight.svg new file mode 100644 index 00000000..7dfa3dcf --- /dev/null +++ b/web/src/assets/images/userMemory/memoryInsight.svg @@ -0,0 +1,32 @@ + + + 编组 26 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/memoryInsight_active.svg b/web/src/assets/images/userMemory/memoryInsight_active.svg new file mode 100644 index 00000000..43c73a4b --- /dev/null +++ b/web/src/assets/images/userMemory/memoryInsight_active.svg @@ -0,0 +1,15 @@ + + + 热点洞察 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/mp3.svg b/web/src/assets/images/userMemory/mp3.svg new file mode 100644 index 00000000..6bc6f2c6 --- /dev/null +++ b/web/src/assets/images/userMemory/mp3.svg @@ -0,0 +1,60 @@ + + + 编组 9 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + MP3 + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/pause.svg b/web/src/assets/images/userMemory/pause.svg new file mode 100644 index 00000000..95e5d0ca --- /dev/null +++ b/web/src/assets/images/userMemory/pause.svg @@ -0,0 +1,20 @@ + + + 播放 + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/play.svg b/web/src/assets/images/userMemory/play.svg new file mode 100644 index 00000000..a3caf5be --- /dev/null +++ b/web/src/assets/images/userMemory/play.svg @@ -0,0 +1,15 @@ + + + 播放 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/play_opacity.svg b/web/src/assets/images/userMemory/play_opacity.svg new file mode 100644 index 00000000..78de47cf --- /dev/null +++ b/web/src/assets/images/userMemory/play_opacity.svg @@ -0,0 +1,15 @@ + + + 播放 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/play_speed.svg b/web/src/assets/images/userMemory/play_speed.svg new file mode 100644 index 00000000..0245a19e --- /dev/null +++ b/web/src/assets/images/userMemory/play_speed.svg @@ -0,0 +1,13 @@ + + + iconfont-PREV + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/question.svg b/web/src/assets/images/userMemory/question.svg new file mode 100644 index 00000000..f8b0fee4 --- /dev/null +++ b/web/src/assets/images/userMemory/question.svg @@ -0,0 +1,15 @@ + + + 问号小 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/refresh.svg b/web/src/assets/images/userMemory/refresh.svg new file mode 100644 index 00000000..46627009 --- /dev/null +++ b/web/src/assets/images/userMemory/refresh.svg @@ -0,0 +1,20 @@ + + + 重新生成 + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/retrieval_number.svg b/web/src/assets/images/userMemory/retrieval_number.svg new file mode 100644 index 00000000..0257ad37 --- /dev/null +++ b/web/src/assets/images/userMemory/retrieval_number.svg @@ -0,0 +1,22 @@ + + + 编组 5 + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/up_border.svg b/web/src/assets/images/userMemory/up_border.svg index a7fe9978..9435cb19 100644 --- a/web/src/assets/images/userMemory/up_border.svg +++ b/web/src/assets/images/userMemory/up_border.svg @@ -1,12 +1,12 @@ 下拉备份 - - + + - - - + + + diff --git a/web/src/assets/images/userMemory/user.png b/web/src/assets/images/userMemory/user.png new file mode 100644 index 00000000..671ab044 Binary files /dev/null and b/web/src/assets/images/userMemory/user.png differ diff --git a/web/src/assets/images/userMemory/userProfile.svg b/web/src/assets/images/userMemory/userProfile.svg new file mode 100644 index 00000000..fd996bf0 --- /dev/null +++ b/web/src/assets/images/userMemory/userProfile.svg @@ -0,0 +1,22 @@ + + + 知识库 + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/userMemory/userProfile_active.svg b/web/src/assets/images/userMemory/userProfile_active.svg new file mode 100644 index 00000000..76a1e9e8 --- /dev/null +++ b/web/src/assets/images/userMemory/userProfile_active.svg @@ -0,0 +1,22 @@ + + + 知识库 + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/agent_arbitration.png b/web/src/assets/images/workflow/agent_arbitration.png deleted file mode 100644 index d555e3e2..00000000 Binary files a/web/src/assets/images/workflow/agent_arbitration.png and /dev/null differ diff --git a/web/src/assets/images/workflow/agent_collaboration.png b/web/src/assets/images/workflow/agent_collaboration.png deleted file mode 100644 index 7a92aecf..00000000 Binary files a/web/src/assets/images/workflow/agent_collaboration.png and /dev/null differ diff --git a/web/src/assets/images/workflow/agent_scheduling.png b/web/src/assets/images/workflow/agent_scheduling.png deleted file mode 100644 index 97028422..00000000 Binary files a/web/src/assets/images/workflow/agent_scheduling.png and /dev/null differ diff --git a/web/src/assets/images/workflow/aggregator.png b/web/src/assets/images/workflow/aggregator.png deleted file mode 100644 index 6253733a..00000000 Binary files a/web/src/assets/images/workflow/aggregator.png and /dev/null differ diff --git a/web/src/assets/images/workflow/aggregator.svg b/web/src/assets/images/workflow/aggregator.svg new file mode 100644 index 00000000..c757e1a1 --- /dev/null +++ b/web/src/assets/images/workflow/aggregator.svg @@ -0,0 +1,31 @@ + + + 编组 6 + + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/assigner.png b/web/src/assets/images/workflow/assigner.png deleted file mode 100644 index 4370bfdd..00000000 Binary files a/web/src/assets/images/workflow/assigner.png and /dev/null differ diff --git a/web/src/assets/images/workflow/assigner.svg b/web/src/assets/images/workflow/assigner.svg new file mode 100644 index 00000000..c653694f --- /dev/null +++ b/web/src/assets/images/workflow/assigner.svg @@ -0,0 +1,30 @@ + + + 编组 6 + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/break.png b/web/src/assets/images/workflow/break.png deleted file mode 100644 index 473ab068..00000000 Binary files a/web/src/assets/images/workflow/break.png and /dev/null differ diff --git a/web/src/assets/images/workflow/break.svg b/web/src/assets/images/workflow/break.svg new file mode 100644 index 00000000..aefc203a --- /dev/null +++ b/web/src/assets/images/workflow/break.svg @@ -0,0 +1,30 @@ + + + 编组 14 + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/classification.png b/web/src/assets/images/workflow/classification.png deleted file mode 100644 index 87d34bb8..00000000 Binary files a/web/src/assets/images/workflow/classification.png and /dev/null differ diff --git a/web/src/assets/images/workflow/clear.svg b/web/src/assets/images/workflow/clear.svg new file mode 100644 index 00000000..10502289 --- /dev/null +++ b/web/src/assets/images/workflow/clear.svg @@ -0,0 +1,13 @@ + + + clear-outlined + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/code_execution.png b/web/src/assets/images/workflow/code_execution.png deleted file mode 100644 index 7f802b3c..00000000 Binary files a/web/src/assets/images/workflow/code_execution.png and /dev/null differ diff --git a/web/src/assets/images/workflow/code_execution.svg b/web/src/assets/images/workflow/code_execution.svg new file mode 100644 index 00000000..4d749ddd --- /dev/null +++ b/web/src/assets/images/workflow/code_execution.svg @@ -0,0 +1,27 @@ + + + 编组 13 + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/condition.png b/web/src/assets/images/workflow/condition.png deleted file mode 100644 index a0bf9160..00000000 Binary files a/web/src/assets/images/workflow/condition.png and /dev/null differ diff --git a/web/src/assets/images/workflow/condition.svg b/web/src/assets/images/workflow/condition.svg new file mode 100644 index 00000000..addb1122 --- /dev/null +++ b/web/src/assets/images/workflow/condition.svg @@ -0,0 +1,27 @@ + + + 编组 14 + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/delete.svg b/web/src/assets/images/workflow/delete.svg new file mode 100644 index 00000000..238a729c --- /dev/null +++ b/web/src/assets/images/workflow/delete.svg @@ -0,0 +1,23 @@ + + + 编组 33 + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/delete_hover.svg b/web/src/assets/images/workflow/delete_hover.svg new file mode 100644 index 00000000..2f145453 --- /dev/null +++ b/web/src/assets/images/workflow/delete_hover.svg @@ -0,0 +1,23 @@ + + + 编组 33 + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/document-extractor.svg b/web/src/assets/images/workflow/document-extractor.svg new file mode 100644 index 00000000..eea39cc6 --- /dev/null +++ b/web/src/assets/images/workflow/document-extractor.svg @@ -0,0 +1,32 @@ + + + 3备份 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/end.png b/web/src/assets/images/workflow/end.png deleted file mode 100644 index 7f4628c6..00000000 Binary files a/web/src/assets/images/workflow/end.png and /dev/null differ diff --git a/web/src/assets/images/workflow/end.svg b/web/src/assets/images/workflow/end.svg new file mode 100644 index 00000000..7c8eb34e --- /dev/null +++ b/web/src/assets/images/workflow/end.svg @@ -0,0 +1,23 @@ + + + 编组 13 + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/features.svg b/web/src/assets/images/workflow/features.svg new file mode 100644 index 00000000..2ff48584 --- /dev/null +++ b/web/src/assets/images/workflow/features.svg @@ -0,0 +1,15 @@ + + + 参与 + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/file_fold.svg b/web/src/assets/images/workflow/file_fold.svg new file mode 100644 index 00000000..b50f10de --- /dev/null +++ b/web/src/assets/images/workflow/file_fold.svg @@ -0,0 +1,13 @@ + + + 文件夹 + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/http_request.png b/web/src/assets/images/workflow/http_request.png deleted file mode 100644 index 64e55d36..00000000 Binary files a/web/src/assets/images/workflow/http_request.png and /dev/null differ diff --git a/web/src/assets/images/workflow/http_request.svg b/web/src/assets/images/workflow/http_request.svg new file mode 100644 index 00000000..36c8995f --- /dev/null +++ b/web/src/assets/images/workflow/http_request.svg @@ -0,0 +1,27 @@ + + + 编组 12 + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/iteration.png b/web/src/assets/images/workflow/iteration.png deleted file mode 100644 index dd73767b..00000000 Binary files a/web/src/assets/images/workflow/iteration.png and /dev/null differ diff --git a/web/src/assets/images/workflow/iteration.svg b/web/src/assets/images/workflow/iteration.svg new file mode 100644 index 00000000..5bc4d840 --- /dev/null +++ b/web/src/assets/images/workflow/iteration.svg @@ -0,0 +1,32 @@ + + + 编组 6 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/llm.png b/web/src/assets/images/workflow/llm.png deleted file mode 100644 index 5d9e7465..00000000 Binary files a/web/src/assets/images/workflow/llm.png and /dev/null differ diff --git a/web/src/assets/images/workflow/llm.svg b/web/src/assets/images/workflow/llm.svg new file mode 100644 index 00000000..54cee4e0 --- /dev/null +++ b/web/src/assets/images/workflow/llm.svg @@ -0,0 +1,20 @@ + + + 编组 14 + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/loop.png b/web/src/assets/images/workflow/loop.png deleted file mode 100644 index a4313229..00000000 Binary files a/web/src/assets/images/workflow/loop.png and /dev/null differ diff --git a/web/src/assets/images/workflow/loop.svg b/web/src/assets/images/workflow/loop.svg new file mode 100644 index 00000000..78fbe8a2 --- /dev/null +++ b/web/src/assets/images/workflow/loop.svg @@ -0,0 +1,27 @@ + + + 编组 15 + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/memory-read.png b/web/src/assets/images/workflow/memory-read.png deleted file mode 100644 index 4b0cdc1d..00000000 Binary files a/web/src/assets/images/workflow/memory-read.png and /dev/null differ diff --git a/web/src/assets/images/workflow/memory-read.svg b/web/src/assets/images/workflow/memory-read.svg new file mode 100644 index 00000000..d385748e --- /dev/null +++ b/web/src/assets/images/workflow/memory-read.svg @@ -0,0 +1,26 @@ + + + 编组 13 + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/memory-write.png b/web/src/assets/images/workflow/memory-write.png deleted file mode 100644 index 83a50fd4..00000000 Binary files a/web/src/assets/images/workflow/memory-write.png and /dev/null differ diff --git a/web/src/assets/images/workflow/memory-write.svg b/web/src/assets/images/workflow/memory-write.svg new file mode 100644 index 00000000..404275b4 --- /dev/null +++ b/web/src/assets/images/workflow/memory-write.svg @@ -0,0 +1,29 @@ + + + 编组 13 + + + + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/memory_enhancement.png b/web/src/assets/images/workflow/memory_enhancement.png deleted file mode 100644 index 998c02fe..00000000 Binary files a/web/src/assets/images/workflow/memory_enhancement.png and /dev/null differ diff --git a/web/src/assets/images/workflow/menuFold.svg b/web/src/assets/images/workflow/menuFold.svg new file mode 100644 index 00000000..77dc38ac --- /dev/null +++ b/web/src/assets/images/workflow/menuFold.svg @@ -0,0 +1,17 @@ + + + 收起 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/minus.png b/web/src/assets/images/workflow/minus.png new file mode 100644 index 00000000..8dabd4dc Binary files /dev/null and b/web/src/assets/images/workflow/minus.png differ diff --git a/web/src/assets/images/workflow/model_selection.png b/web/src/assets/images/workflow/model_selection.png deleted file mode 100644 index e3e93962..00000000 Binary files a/web/src/assets/images/workflow/model_selection.png and /dev/null differ diff --git a/web/src/assets/images/workflow/model_voting.png b/web/src/assets/images/workflow/model_voting.png deleted file mode 100644 index 8324541e..00000000 Binary files a/web/src/assets/images/workflow/model_voting.png and /dev/null differ diff --git a/web/src/assets/images/workflow/node_plus.png b/web/src/assets/images/workflow/node_plus.png new file mode 100644 index 00000000..61e83c65 Binary files /dev/null and b/web/src/assets/images/workflow/node_plus.png differ diff --git a/web/src/assets/images/workflow/output_audit.png b/web/src/assets/images/workflow/output_audit.png deleted file mode 100644 index 50128f82..00000000 Binary files a/web/src/assets/images/workflow/output_audit.png and /dev/null differ diff --git a/web/src/assets/images/workflow/parallel.png b/web/src/assets/images/workflow/parallel.png deleted file mode 100644 index e77d79d8..00000000 Binary files a/web/src/assets/images/workflow/parallel.png and /dev/null differ diff --git a/web/src/assets/images/workflow/parameter_extraction.png b/web/src/assets/images/workflow/parameter_extraction.png deleted file mode 100644 index d4b50ee0..00000000 Binary files a/web/src/assets/images/workflow/parameter_extraction.png and /dev/null differ diff --git a/web/src/assets/images/workflow/parameter_extraction.svg b/web/src/assets/images/workflow/parameter_extraction.svg new file mode 100644 index 00000000..f3472516 --- /dev/null +++ b/web/src/assets/images/workflow/parameter_extraction.svg @@ -0,0 +1,22 @@ + + + 编组 15 + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/plus.png b/web/src/assets/images/workflow/plus.png new file mode 100644 index 00000000..05f5066a Binary files /dev/null and b/web/src/assets/images/workflow/plus.png differ diff --git a/web/src/assets/images/workflow/process_evolution.png b/web/src/assets/images/workflow/process_evolution.png deleted file mode 100644 index 8262c00d..00000000 Binary files a/web/src/assets/images/workflow/process_evolution.png and /dev/null differ diff --git a/web/src/assets/images/workflow/question-classifier.png b/web/src/assets/images/workflow/question-classifier.png index 754a0a62..9a95e4ab 100644 Binary files a/web/src/assets/images/workflow/question-classifier.png and b/web/src/assets/images/workflow/question-classifier.png differ diff --git a/web/src/assets/images/workflow/question-classifier.svg b/web/src/assets/images/workflow/question-classifier.svg new file mode 100644 index 00000000..3a85ff8b --- /dev/null +++ b/web/src/assets/images/workflow/question-classifier.svg @@ -0,0 +1,23 @@ + + + 编组 14 + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/rag.png b/web/src/assets/images/workflow/rag.png deleted file mode 100644 index 3749dbfa..00000000 Binary files a/web/src/assets/images/workflow/rag.png and /dev/null differ diff --git a/web/src/assets/images/workflow/rag.svg b/web/src/assets/images/workflow/rag.svg new file mode 100644 index 00000000..e9648fc8 --- /dev/null +++ b/web/src/assets/images/workflow/rag.svg @@ -0,0 +1,26 @@ + + + 编组 14 + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/reasoning_control.png b/web/src/assets/images/workflow/reasoning_control.png deleted file mode 100644 index 649e165c..00000000 Binary files a/web/src/assets/images/workflow/reasoning_control.png and /dev/null differ diff --git a/web/src/assets/images/workflow/refresh_active.svg b/web/src/assets/images/workflow/refresh_active.svg new file mode 100644 index 00000000..f9b0b3d8 --- /dev/null +++ b/web/src/assets/images/workflow/refresh_active.svg @@ -0,0 +1,18 @@ + + + 刷新 + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/return.svg b/web/src/assets/images/workflow/return.svg new file mode 100644 index 00000000..b7cfe153 --- /dev/null +++ b/web/src/assets/images/workflow/return.svg @@ -0,0 +1,17 @@ + + + 退出 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/robot-2-line@2x.png b/web/src/assets/images/workflow/robot-2-line@2x.png deleted file mode 100644 index f1dc247e..00000000 Binary files a/web/src/assets/images/workflow/robot-2-line@2x.png and /dev/null differ diff --git a/web/src/assets/images/workflow/run.svg b/web/src/assets/images/workflow/run.svg new file mode 100644 index 00000000..5d320106 --- /dev/null +++ b/web/src/assets/images/workflow/run.svg @@ -0,0 +1,13 @@ + + + 编组 31 + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/save.svg b/web/src/assets/images/workflow/save.svg new file mode 100644 index 00000000..681c7633 --- /dev/null +++ b/web/src/assets/images/workflow/save.svg @@ -0,0 +1,17 @@ + + + 保存 + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/self_optimization.png b/web/src/assets/images/workflow/self_optimization.png deleted file mode 100644 index 08ed8598..00000000 Binary files a/web/src/assets/images/workflow/self_optimization.png and /dev/null differ diff --git a/web/src/assets/images/workflow/self_reflection.png b/web/src/assets/images/workflow/self_reflection.png deleted file mode 100644 index 099aac60..00000000 Binary files a/web/src/assets/images/workflow/self_reflection.png and /dev/null differ diff --git a/web/src/assets/images/workflow/sensitive_detection.png b/web/src/assets/images/workflow/sensitive_detection.png deleted file mode 100644 index 637a4f13..00000000 Binary files a/web/src/assets/images/workflow/sensitive_detection.png and /dev/null differ diff --git a/web/src/assets/images/workflow/start.png b/web/src/assets/images/workflow/start.png deleted file mode 100644 index f6828988..00000000 Binary files a/web/src/assets/images/workflow/start.png and /dev/null differ diff --git a/web/src/assets/images/workflow/start.svg b/web/src/assets/images/workflow/start.svg new file mode 100644 index 00000000..7b89c1f7 --- /dev/null +++ b/web/src/assets/images/workflow/start.svg @@ -0,0 +1,21 @@ + + + 编组 12 + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/task_planning.png b/web/src/assets/images/workflow/task_planning.png deleted file mode 100644 index 33f322fd..00000000 Binary files a/web/src/assets/images/workflow/task_planning.png and /dev/null differ diff --git a/web/src/assets/images/workflow/template_rendering.png b/web/src/assets/images/workflow/template_rendering.png deleted file mode 100644 index 064caeb6..00000000 Binary files a/web/src/assets/images/workflow/template_rendering.png and /dev/null differ diff --git a/web/src/assets/images/workflow/template_rendering.svg b/web/src/assets/images/workflow/template_rendering.svg new file mode 100644 index 00000000..e52bf30d --- /dev/null +++ b/web/src/assets/images/workflow/template_rendering.svg @@ -0,0 +1,26 @@ + + + 编组 13 + + + + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/tools.png b/web/src/assets/images/workflow/tools.png deleted file mode 100644 index 49ff2fa4..00000000 Binary files a/web/src/assets/images/workflow/tools.png and /dev/null differ diff --git a/web/src/assets/images/workflow/tools.svg b/web/src/assets/images/workflow/tools.svg new file mode 100644 index 00000000..7c772245 --- /dev/null +++ b/web/src/assets/images/workflow/tools.svg @@ -0,0 +1,23 @@ + + + 编组 6 + + + + + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/assets/images/workflow/variable.svg b/web/src/assets/images/workflow/variable.svg new file mode 100644 index 00000000..cdb8338e --- /dev/null +++ b/web/src/assets/images/workflow/variable.svg @@ -0,0 +1,16 @@ + + + 聊天 + + + + + + + + + + + + + \ No newline at end of file diff --git a/web/src/components/AudioRecorder/index.tsx b/web/src/components/AudioRecorder/index.tsx index 639a9109..8df31398 100644 --- a/web/src/components/AudioRecorder/index.tsx +++ b/web/src/components/AudioRecorder/index.tsx @@ -2,12 +2,13 @@ * @Author: ZhaoYing * @Date: 2026-02-06 21:11:51 * @Last Modified by: ZhaoYing - * @Last Modified time: 2026-03-17 18:39:09 + * @Last Modified time: 2026-03-20 14:25:26 */ import { type FC, useRef, useState } from 'react' import RecordRTC from 'recordrtc' -import { App } from 'antd' +import { App, Tooltip } from 'antd' import { useTranslation } from 'react-i18next'; +import clsx from 'clsx'; import { fileUploadUrlWithoutApiPrefix } from '@/api/fileStorage' import { request } from '@/utils/request' @@ -91,14 +92,17 @@ const AudioRecorder: FC = ({ // Toggle between recording/idle states on click; // swap background image to reflect current state return ( -
+ +
+ ) } diff --git a/web/src/components/BtnTabs/index.tsx b/web/src/components/BtnTabs/index.tsx new file mode 100644 index 00000000..772a4c8d --- /dev/null +++ b/web/src/components/BtnTabs/index.tsx @@ -0,0 +1,49 @@ +/* + * @Author: ZhaoYing + * @Date: 2026-03-19 14:05:09 + * @Last Modified by: ZhaoYing + * @Last Modified time: 2026-03-19 14:05:09 + */ +import { type FC } from 'react' +import { Flex } from 'antd'; +import clsx from 'clsx' + +/** A single tab item with a display label and unique key */ +interface Tab { + label: string + key: string +} + +/** Props for the BtnTabs component */ +interface BtnTabsProps { + /** List of tab items to render */ + items: Tab[] + /** Key of the currently active tab */ + activeKey: string + /** Callback fired when a tab is clicked */ + onChange: (key: string) => void; + /** Optional extra class name for the container */ + className?: string; +} + +/** Button-style tab switcher — renders tabs as pill-shaped buttons with active highlight */ +const BtnTabs: FC = ({ items, activeKey, onChange, className }) => { + return ( + + {items.map((tab) => ( +
onChange(tab.key)} + className={clsx('rb:px-2 rb:py-1 rb:rounded-[13px] rb:text-[12px] rb:leading-4.5 rb:cursor-pointer', { + 'rb:bg-[#F6F6F6]': activeKey !== tab.key, + 'rb:bg-[#171719] rb:text-white': activeKey === tab.key, + })} + > + {tab.label} +
+ ))} +
+ ) +} + +export default BtnTabs diff --git a/web/src/components/ButtonCheckbox/index.tsx b/web/src/components/ButtonCheckbox/index.tsx index 18bca7c6..8c52701b 100644 --- a/web/src/components/ButtonCheckbox/index.tsx +++ b/web/src/components/ButtonCheckbox/index.tsx @@ -2,7 +2,7 @@ * @Author: ZhaoYing * @Date: 2026-02-02 15:01:59 * @Last Modified by: ZhaoYing - * @Last Modified time: 2026-03-19 13:41:26 + * @Last Modified time: 2026-03-19 20:45:13 */ /** @@ -64,13 +64,11 @@ const ButtonCheckbox: FC = ({ align="center" justify={cicle ? 'center' : 'start'} gap={4} - className={clsx("rb:flex rb:items-center rb:cursor-pointer rb:px-2! rb:border rb:hover:bg-[#F6F6F6]", { - 'rb:size-7 rb:rounded-[14px] rb:border-[0.5px] rb:border-[#EBEBEB]': cicle, - 'rb:rounded-lg rb:text-[12px] rb:h-6': !cicle, + className={clsx("rb:border rb:rounded-lg rb:px-2! rb:text-[12px] rb:h-6 rb:cursor-pointer", { // Checked state: blue background and border - "rb:bg-[rgba(21,94,239,0.06)] rb:border-[rgba(21,94,239,0.25)] rb:hover:bg-[rgba(21,94,239,0.06)] rb:text-[#155EEF]": checked, + "rb:bg-[#FAFAFA] rb:border-[#171719]": checked, // Unchecked state: gray border and dark text - "rb:border-[#DFE4ED] rb:text-[#212332]": !checked, + "rb:border-[#EBEBEB] rb:text-[#212332] rb:hover:bg-[#F0F3F8]": !checked, "rb:opacity-65 rb:cursor-not-allowed!": disabled })} onClick={handleChange} diff --git a/web/src/components/Charts/AreaLineChart.tsx b/web/src/components/Charts/AreaLineChart.tsx new file mode 100644 index 00000000..40bfabb1 --- /dev/null +++ b/web/src/components/Charts/AreaLineChart.tsx @@ -0,0 +1,306 @@ +/* + * @Author: ZhaoYing + * @Date: 2026-02-10 13:36:03 + * @Last Modified by: ZhaoYing + * @Last Modified time: 2026-02-25 13:51:52 + */ +/* + * AreaLineChart Component + * + * A reusable area line chart component built with ECharts that displays time-series data + * with gradient-filled areas under the lines. Supports multiple data series with + * customizable colors and responsive behavior. + * + * Features: + * - Multiple line series with gradient area fills + * - Gradient line colors (white to color to white) + * - Customizable x-axis key for flexible data structures + * - Date-based x-axis with formatted labels (DD/MM) + * - Responsive resizing using ResizeObserver + * - Interactive tooltips on hover + * - Customizable grid layout and colors + * - Legend at the bottom for series identification + * - Empty state when no data is available + * - Smooth rendering with requestAnimationFrame + */ +import { type FC, useEffect, useRef, useMemo } from 'react' +import ReactEcharts from 'echarts-for-react'; +import * as echarts from 'echarts'; + +import { formatDateTime } from '@/utils/format'; +import Empty from '@/components/Empty' + +/** Base configuration for all line series */ +const SeriesConfig = { + type: 'line', + stack: 'Total', + symbol: 'circle', + symbolSize: 5, + showSymbol: true, + label: { + show: false, + position: 'top' + }, + emphasis: { + focus: 'series' + }, +} + +/** Default color palette for area line series */ +const Colors = ['#155EEF', '#FFB048', '#4DA8FF'] + +/** + * Data structure for chart data points + * Flexible structure allowing any string key with string or number values + * + * @interface ChartData + * @property {string | number} [key: string] - Dynamic properties for x-axis and data series + */ +export interface ChartData { + [key: string]: string | number; +} + +/** + * Props for the AreaLineChart component + * + * @interface AreaLineChartProps + * @property {string} xAxisKey - Key name in chartData to use for x-axis values + * @property {ChartData[]} chartData - Array of data points with dynamic properties + * @property {Record} seriesList - Map of data keys to display names + * @property {string} [className] - Additional CSS classes for the container + * @property {number} [height] - Height of the chart in pixels + * @property {string[]} [colors] - Custom color array for line series and gradients + * @property {any} [grid] - ECharts grid configuration for chart positioning + */ +interface AreaLineChartProps { + xAxisKey: string; + chartData: ChartData[]; + seriesList: Record; + className?: string; + height?: number; + colors?: string[]; + grid?: any; + lineStyle?: any; + showLegend?: boolean; + smooth?: boolean; +} + +/** + * AreaLineChart Component + * + * Renders a multi-series area line chart with gradient fills. + * The area gradient goes from the series color at the top to white at the bottom. + * The line gradient goes from white to the series color and back to white. + * Automatically resizes when container dimensions change. + * + * @param {AreaLineChartProps} props - Component props + * @returns {JSX.Element} Rendered area line chart or empty state + * + * @example + * ```tsx + * + * ``` + */ +const AreaLineChart: FC = ({ + xAxisKey, + chartData, + seriesList, + height, + colors = Colors, + grid = { + top: 7, + left: 4, + right: 16, + bottom: 32, + containLabel: true + }, + lineStyle, + showLegend = true, + smooth = true +}) => { + /** Reference to the ECharts instance for programmatic control */ + const chartRef = useRef(null); + /** Flag to prevent multiple simultaneous resize operations */ + const resizeScheduledRef = useRef(false) + + /** + * Generate series configuration for each data series with gradient effects + * Creates area fills with vertical gradients (color to white) + * and line colors with horizontal gradients (white to color to white) + * + * @returns {Array} Array of ECharts series configurations with gradient styles + */ + const getSeries = () => { + return Object.entries(seriesList).map(([key, name], index) => ({ + ...SeriesConfig, + name: name, + data: chartData.map(vo => vo[key as keyof ChartData]), + areaStyle: { + opacity: 0.8, + color: new echarts.graphic.LinearGradient(0, 0, 0, 1, [ + { + offset: 0, + color: colors[index] + }, + { + offset: 1, + color: '#FFFFFF' + } + ]) + }, + lineStyle: lineStyle || { + width: 3, + color: new echarts.graphic.LinearGradient(0, 0, 1, 0, [ + { + offset: 0, + color: '#FFFFFF' + }, + { + offset: 0.8, + color: colors[index] + }, + { + offset: 1, + color: '#FFFFFF' + } + ]) + }, + smooth + })) + } + /** + * Memoized legend data to prevent unnecessary recalculations + * Formats series list for display in chart legend + */ + const formatSeriesList = useMemo(() => { + return Object.entries(seriesList).map(([_key, name]) => ({ + ...SeriesConfig, + name: name, + })) + }, [seriesList]) + + /** + * Set up responsive behavior using ResizeObserver + * Resizes chart when parent container dimensions change + */ + useEffect(() => { + const handleResize = () => { + if (chartRef.current && !resizeScheduledRef.current) { + resizeScheduledRef.current = true + requestAnimationFrame(() => { + chartRef.current?.getEchartsInstance().resize(); + resizeScheduledRef.current = false + }); + } + } + + const resizeObserver = new ResizeObserver(handleResize) + const chartElement = chartRef.current?.getEchartsInstance().getDom().parentElement + if (chartElement) { + resizeObserver.observe(chartElement) + } + + return () => { + resizeObserver.disconnect() + } + }, [chartData]) + + return ( +
+ {chartData && chartData.length > 0 + ? formatDateTime(item[xAxisKey], 'DD/MM')), + boundaryGap: false, + axisLabel: { + color: '#5B6167', + fontFamily: 'PingFangSC, PingFang SC', + lineHeight: 17, + }, + axisLine: { + show: false, + lineStyle: { + color: '#EBEBEB', + } + }, + splitLine: { + show: false, + }, + axisTick: { + show: false + } + }, + yAxis: { + type: 'value', + axisLabel: { + color: '#A8A9AA', + fontFamily: 'PingFangSC, PingFang SC', + align: 'right', + lineHeight: 17, + }, + axisLine: { + lineStyle: { + color: '#EBEBEB', + } + }, + }, + series: getSeries() + }} + style={{ height: `${height}px`, width: '100%', minWidth: '100%', boxSizing: 'border-box' }} + opts={{ renderer: 'canvas' }} + notMerge={true} + lazyUpdate={true} + /> + : + } +
+ ) +} + +export default AreaLineChart diff --git a/web/src/components/Charts/BarChart.tsx b/web/src/components/Charts/BarChart.tsx new file mode 100644 index 00000000..e476a2cb --- /dev/null +++ b/web/src/components/Charts/BarChart.tsx @@ -0,0 +1,295 @@ +/* + * @Author: ZhaoYing + * @Date: 2026-02-10 13:36:03 + * @Last Modified by: ZhaoYing + * @Last Modified time: 2026-02-25 13:49:04 + */ +/* + * BarChart Component + * + * A reusable area line chart component built with ECharts that displays time-series data + * with gradient-filled areas under the lines. Supports multiple data series with + * customizable colors and responsive behavior. + * + * Features: + * - Multiple line series with gradient area fills + * - Gradient line colors (white to color to white) + * - Customizable x-axis key for flexible data structures + * - Date-based x-axis with formatted labels (DD/MM) + * - Responsive resizing using ResizeObserver + * - Interactive tooltips on hover + * - Customizable grid layout and colors + * - Legend at the bottom for series identification + * - Empty state when no data is available + * - Smooth rendering with requestAnimationFrame + */ +import { type FC, useEffect, useRef, useMemo } from 'react' +import ReactEcharts from 'echarts-for-react'; +import * as echarts from 'echarts'; + +import { formatDateTime } from '@/utils/format'; +import Empty from '@/components/Empty' + +/** Base configuration for all line series */ +const SeriesConfig = { + type: 'bar', + stack: 'Total', + symbol: 'circle', + symbolSize: 5, + showSymbol: true, + label: { + show: false, + position: 'top' + }, + emphasis: { + focus: 'series' + }, + showBackground: true, +} + +/** Default color palette for area line series */ +const Colors = ['#155EEF', '#FFB048', '#4DA8FF'] + +/** + * Data structure for chart data points + * Flexible structure allowing any string key with string or number values + * + * @interface ChartData + * @property {string | number} [key: string] - Dynamic properties for x-axis and data series + */ +export interface ChartData { + [key: string]: string | number; +} + +/** + * Props for the BarChart component + * + * @interface BarChartProps + * @property {string} xAxisKey - Key name in chartData to use for x-axis values + * @property {ChartData[]} chartData - Array of data points with dynamic properties + * @property {Record} seriesList - Map of data keys to display names + * @property {string} [className] - Additional CSS classes for the container + * @property {number} [height] - Height of the chart in pixels + * @property {string[]} [colors] - Custom color array for line series and gradients + * @property {any} [grid] - ECharts grid configuration for chart positioning + */ +interface BarChartProps { + xAxisKey: string; + chartData: ChartData[]; + seriesList: Record; + className?: string; + height?: number; + colors?: string[]; + grid?: any; + itemStyle?: any; + showLegend?: boolean; + showBackground?: boolean; +} + +/** + * BarChart Component + * + * Renders a multi-series area line chart with gradient fills. + * The area gradient goes from the series color at the top to white at the bottom. + * The line gradient goes from white to the series color and back to white. + * Automatically resizes when container dimensions change. + * + * @param {BarChartProps} props - Component props + * @returns {JSX.Element} Rendered area line chart or empty state + * + * @example + * ```tsx + * + * ``` + */ +const BarChart: FC = ({ + xAxisKey, + chartData, + seriesList, + height, + colors = Colors, + grid = { + top: 7, + left: 4, + right: 16, + bottom: 32, + containLabel: true + }, + itemStyle, + showLegend = true, + showBackground = true, +}) => { + /** Reference to the ECharts instance for programmatic control */ + const chartRef = useRef(null); + /** Flag to prevent multiple simultaneous resize operations */ + const resizeScheduledRef = useRef(false) + + /** + * Generate series configuration for each data series with gradient effects + * Creates area fills with vertical gradients (color to white) + * and line colors with horizontal gradients (white to color to white) + * + * @returns {Array} Array of ECharts series configurations with gradient styles + */ + const getSeries = () => { + return Object.entries(seriesList).map(([key, name], index) => ({ + ...SeriesConfig, + name: name, + data: chartData.map(vo => vo[key as keyof ChartData]), + barWidth: 16, + itemStyle: itemStyle || { + color: new echarts.graphic.LinearGradient(0, 0, 0, 1, [ + { + offset: 0, + color: colors[index] + }, + { + offset: 1, + color: '#FFFFFF' + } + ]), + }, + emphasis: { + itemStyle: { + } + }, + barGap: '-100%', + showBackground: showBackground, + })) + } + /** + * Memoized legend data to prevent unnecessary recalculations + * Formats series list for display in chart legend + */ + const formatSeriesList = useMemo(() => { + return Object.entries(seriesList).map(([_key, name]) => ({ + ...SeriesConfig, + name: name, + })) + }, [seriesList]) + + /** + * Set up responsive behavior using ResizeObserver + * Resizes chart when parent container dimensions change + */ + useEffect(() => { + const handleResize = () => { + if (chartRef.current && !resizeScheduledRef.current) { + resizeScheduledRef.current = true + requestAnimationFrame(() => { + chartRef.current?.getEchartsInstance().resize(); + resizeScheduledRef.current = false + }); + } + } + + const resizeObserver = new ResizeObserver(handleResize) + const chartElement = chartRef.current?.getEchartsInstance().getDom().parentElement + if (chartElement) { + resizeObserver.observe(chartElement) + } + + return () => { + resizeObserver.disconnect() + } + }, [chartData]) + + return ( +
+ {chartData && chartData.length > 0 + ? formatDateTime(item[xAxisKey], 'DD/MM')), + boundaryGap: false, + axisLabel: { + color: '#5B6167', + fontFamily: 'PingFangSC, PingFang SC', + lineHeight: 17, + }, + axisLine: { + show: false, + itemStyle: { + color: '#EBEBEB', + } + }, + splitLine: { + show: false, + }, + axisTick: { + show: false + } + }, + yAxis: { + type: 'value', + axisLabel: { + color: '#A8A9AA', + fontFamily: 'PingFangSC, PingFang SC', + align: 'right', + lineHeight: 17, + }, + axisLine: { + itemStyle: { + color: '#EBEBEB', + } + }, + }, + series: getSeries() + }} + style={{ height: `${height}px`, width: '100%', minWidth: '100%', boxSizing: 'border-box' }} + opts={{ renderer: 'canvas' }} + notMerge={true} + lazyUpdate={true} + /> + : + } +
+ ) +} + +export default BarChart diff --git a/web/src/components/Charts/GraphNetworkChart.tsx b/web/src/components/Charts/GraphNetworkChart.tsx new file mode 100644 index 00000000..8f4ec796 --- /dev/null +++ b/web/src/components/Charts/GraphNetworkChart.tsx @@ -0,0 +1,200 @@ +/* + * @Author: ZhaoYing + * @Date: 2026-02-10 14:06:09 + * @Last Modified by: ZhaoYing + * @Last Modified time: 2026-02-10 14:06:09 + */ +/** + * GraphNetworkChart Component + * + * A force-directed graph visualization component built with ECharts. + * Displays nodes and edges in an interactive network diagram with physics-based layout. + * Supports zooming, panning, dragging nodes, and click interactions. + */ +import { type FC, useEffect, useRef, type SetStateAction, type Dispatch } from 'react' +import ReactEcharts from 'echarts-for-react'; + +import PageEmpty from '@/components/Empty/PageEmpty' + +// Default color palette for node categories +const Colors = ['#171719', '#155EEF', '#9C6FFF', '#FF8A4C'] + +/** + * Node interface representing a graph node/vertex + */ +export interface Node { + id: string; // Unique identifier for the node + label: string; // Display label for the node + category: number; // Category index for grouping and coloring + symbolSize: number; // Size of the node symbol in pixels + name: string; // Node name (used in ECharts) + itemStyle: { + color: string; // Custom color for this node + } + caption: string; // Additional description or caption + [key: string]: any; // Allow additional custom properties +} + +/** + * Edge interface representing a connection between two nodes + */ +export interface Edge { + id: string; // Unique identifier for the edge + source: string; // Source node ID + target: string; // Target node ID + type: string; // Type/category of the relationship + caption: string; // Description of the relationship + value: number; // Numeric value associated with the edge + weight: number; // Weight/strength of the connection +} + +/** + * Props for the GraphNetworkChart component + */ +interface GraphNetworkChartProps { + nodes: Node[]; // Array of nodes to display in the graph + links: Edge[]; // Array of edges connecting the nodes + categories: { name: string }[]; // Category definitions for node grouping + colors?: string[]; // Optional custom color palette (defaults to Colors) + onNodeClick: Dispatch>; // Callback when a node is clicked +} + +const GraphNetworkChart: FC = ({ + nodes, + links, + categories, + colors = Colors, + onNodeClick, +}) => { + // Reference to the ECharts instance for programmatic control + const chartRef = useRef(null); + + // Flag to prevent multiple simultaneous resize operations (debouncing) + const resizeScheduledRef = useRef(false) + + /** + * Effect: Handle responsive chart resizing + * + * Uses ResizeObserver to detect container size changes and resize the chart accordingly. + * Implements requestAnimationFrame for smooth, debounced resize operations. + * Re-runs when nodes change to ensure proper sizing with new data. + */ + useEffect(() => { + const handleResize = () => { + if (chartRef.current && !resizeScheduledRef.current) { + resizeScheduledRef.current = true + // Use requestAnimationFrame for smooth, optimized resize + requestAnimationFrame(() => { + chartRef.current?.getEchartsInstance().resize(); + resizeScheduledRef.current = false + }); + } + } + + // Observe the chart container for size changes + const resizeObserver = new ResizeObserver(handleResize) + const chartElement = chartRef.current?.getEchartsInstance().getDom().parentElement + if (chartElement) { + resizeObserver.observe(chartElement) + } + + // Cleanup: disconnect observer when component unmounts + return () => { + resizeObserver.disconnect() + } + }, [nodes]) + + return ( +
+ {/* Render chart only if nodes exist, otherwise show empty state */} + {nodes && nodes.length > 0 + ? { + // Only trigger callback for node clicks (not edges or background) + if (params.dataType === 'node') { + onNodeClick(params.data) + } + } + }} + /> + : + } +
+ ) +} + +export default GraphNetworkChart diff --git a/web/src/components/Charts/LineChart.tsx b/web/src/components/Charts/LineChart.tsx new file mode 100644 index 00000000..e5217336 --- /dev/null +++ b/web/src/components/Charts/LineChart.tsx @@ -0,0 +1,260 @@ +/* + * @Author: ZhaoYing + * @Date: 2026-02-10 13:35:55 + * @Last Modified by: ZhaoYing + * @Last Modified time: 2026-02-10 13:35:55 + */ +/* + * LineChart Component + * + * A reusable line chart component built with ECharts for displaying time-series data + * with multiple data series. Supports customizable colors, responsive behavior, + * and interactive tooltips. + * + * Features: + * - Multiple line series with different colors + * - Date-based x-axis with formatted labels (DD/MM) + * - Responsive resizing using ResizeObserver + * - Interactive tooltips on hover + * - Customizable grid layout and colors + * - Legend at the bottom for series identification + * - Empty state when no data is available + * - Smooth rendering with requestAnimationFrame + */ +import { type FC, useEffect, useRef, useMemo } from 'react' +import ReactEcharts from 'echarts-for-react'; + +import { formatDateTime } from '@/utils/format'; +import Empty from '@/components/Empty' + +/** Base configuration for all line series */ +const SeriesConfig = { + type: 'line', + stack: 'Total', + symbol: 'circle', + symbolSize: 5, + showSymbol: true, + label: { + show: false, + position: 'top' + }, + emphasis: { + focus: 'series' + }, +} + +/** Default color palette for line series */ +const Colors = ['#171719', '#155EEF', '#FF5D34'] + +/** + * Data structure for chart data points + * + * @interface ChartData + * @property {string | number} date - Date value for x-axis (timestamp or date string) + * @property {string | number} [key: string] - Dynamic properties for different data series + */ +export interface ChartData { + date: string | number; + [key: string]: string | number; +} + +/** + * Props for the LineChart component + * + * @interface LineChartProps + * @property {ChartData[]} chartData - Array of data points with date and series values + * @property {Record} seriesList - Map of data keys to display names + * @property {string} [className] - Additional CSS classes for the container + * @property {number} [height] - Height of the chart in pixels + * @property {string[]} [colors] - Custom color array for line series + * @property {any} [grid] - ECharts grid configuration for chart positioning + */ +interface LineChartProps { + chartData: ChartData[]; + seriesList: Record; + className?: string; + height?: number; + colors?: string[]; + grid?: any; +} + +/** + * LineChart Component + * + * Renders a multi-series line chart with date-based x-axis. + * Automatically resizes when container dimensions change. + * + * @param {LineChartProps} props - Component props + * @returns {JSX.Element} Rendered line chart or empty state + * + * @example + * ```tsx + * + * ``` + */ +const LineChart: FC = ({ + chartData, + seriesList, + height, + colors = Colors, + grid = { + top: 7, + right: 16, + } +}) => { + /** Reference to the ECharts instance for programmatic control */ + const chartRef = useRef(null); + /** Flag to prevent multiple simultaneous resize operations */ + const resizeScheduledRef = useRef(false) + + /** + * Generate series configuration for each data series + * Maps seriesList keys to chart series with corresponding data and colors + * + * @returns {Array} Array of ECharts series configurations + */ + const getSeries = () => { + return Object.entries(seriesList).map(([key, name], index) => ({ + ...SeriesConfig, + name: name, + data: chartData.map(vo => vo[key as keyof ChartData]), + lineStyle: { + width: 2, + color: colors[index] + }, + })) + } + /** + * Memoized legend data to prevent unnecessary recalculations + * Formats series list for display in chart legend + */ + const formatSeriesList = useMemo(() => { + return Object.entries(seriesList).map(([_key, name]) => ({ + ...SeriesConfig, + name: name, + })) + }, [seriesList]) + + /** + * Set up responsive behavior using ResizeObserver + * Resizes chart when parent container dimensions change + */ + useEffect(() => { + const handleResize = () => { + if (chartRef.current && !resizeScheduledRef.current) { + resizeScheduledRef.current = true + requestAnimationFrame(() => { + chartRef.current?.getEchartsInstance().resize(); + resizeScheduledRef.current = false + }); + } + } + + const resizeObserver = new ResizeObserver(handleResize) + const chartElement = chartRef.current?.getEchartsInstance().getDom().parentElement + if (chartElement) { + resizeObserver.observe(chartElement) + } + + return () => { + resizeObserver.disconnect() + } + }, [chartData]) + + return ( +
+ {chartData && chartData.length > 0 + ? formatDateTime(item.date, 'DD/MM')), + boundaryGap: false, + axisLabel: { + color: '#5B6167', + fontFamily: 'PingFangSC, PingFang SC', + lineHeight: 17, + }, + axisLine: { + show: false, + lineStyle: { + color: '#EBEBEB', + } + }, + splitLine: { + show: false, + }, + axisTick: { + show: false + } + }, + yAxis: { + type: 'value', + axisLabel: { + color: '#A8A9AA', + fontFamily: 'PingFangSC, PingFang SC', + align: 'right', + lineHeight: 17, + }, + axisLine: { + lineStyle: { + color: '#EBEBEB', + } + }, + }, + series: getSeries() + }} + style={{ height: '100%', width: '100%', minWidth: '100%', boxSizing: 'border-box' }} + opts={{ renderer: 'canvas' }} + notMerge={true} + lazyUpdate={true} + /> + : + } +
+ ) +} + +export default LineChart diff --git a/web/src/components/Charts/PieChart.tsx b/web/src/components/Charts/PieChart.tsx new file mode 100644 index 00000000..b0c67549 --- /dev/null +++ b/web/src/components/Charts/PieChart.tsx @@ -0,0 +1,204 @@ +/* + * @Author: ZhaoYing + * @Date: 2026-02-10 13:35:45 + * @Last Modified by: ZhaoYing + * @Last Modified time: 2026-03-16 11:34:30 + */ +/* + * PieChart Component + * + * A reusable pie chart component built with ECharts that displays data distribution + * in a donut chart format with customizable colors and responsive behavior. + * + * Features: + * - Donut-style pie chart with percentage labels + * - Customizable color palette + * - Responsive resizing using ResizeObserver + * - Hover tooltips showing percentage values + * - Legend at the bottom with horizontal layout + * - Empty state when no data is available + * - Shadow effects for better visual depth + */ +import { type FC, useEffect, useRef } from 'react' +import ReactEcharts from 'echarts-for-react'; + +import Empty from '@/components/Empty' + +/** Default color palette for pie chart segments */ +const Colors = ['#171719', '#155EEF', '#4DA8FF', '#9C6FFF', '#ABEBFF', '#DFE4ED'] + +/** + * Data structure for each pie chart segment + * + * @interface ChartData + * @property {string} name - Label for the segment (displayed in legend) + * @property {number} value - Numeric value for the segment (determines size) + */ +export interface ChartData { + name: string; + value: number; +} + +/** + * Props for the PieChart component + * + * @interface PieChartProps + * @property {ChartData[]} chartData - Array of data points to display in the chart + * @property {number} [height=260] - Height of the chart in pixels + * @property {string[]} [colors] - Custom color array for chart segments (defaults to Colors) + */ +interface PieChartProps { + chartData: ChartData[]; + height?: number; + colors?: string[]; + itemGap?: number; + seriesWidth?: number; + seriesHeight?: number; + seriesLabel?: boolean; + seriesTop?: number; +} + +/** + * PieChart Component + * + * Renders a donut-style pie chart with percentage labels and legend. + * Automatically resizes when container dimensions change. + * + * @param {PieChartProps} props - Component props + * @returns {JSX.Element} Rendered pie chart or empty state + * + * @example + * ```tsx + * + * ``` + */ +const PieChart: FC = ({ + chartData, + height = 260, + seriesWidth = 182, + seriesHeight = 182, + colors = Colors, + itemGap = 48, + seriesLabel = true, + seriesTop = 24, +}) => { + /** Reference to the ECharts instance for programmatic control */ + const chartRef = useRef(null); + /** Flag to prevent multiple simultaneous resize operations */ + const resizeScheduledRef = useRef(false) + + /** + * Set up responsive behavior using ResizeObserver + * Resizes chart when parent container dimensions change + */ + useEffect(() => { + const handleResize = () => { + if (chartRef.current && !resizeScheduledRef.current) { + resizeScheduledRef.current = true + // Use requestAnimationFrame for smooth resize performance + requestAnimationFrame(() => { + chartRef.current?.getEchartsInstance().resize(); + resizeScheduledRef.current = false + }); + } + } + + const resizeObserver = new ResizeObserver(handleResize) + const chartElement = chartRef.current?.getEchartsInstance().getDom().parentElement + if (chartElement) { + resizeObserver.observe(chartElement) + } + + // Cleanup: disconnect observer when component unmounts + return () => { + resizeObserver.disconnect() + } + }, [chartData]) + + return ( +
+ {chartData && chartData.length > 0 + ? + : + } +
+ ) +} + +export default PieChart diff --git a/web/src/components/Chat/ChatContent.tsx b/web/src/components/Chat/ChatContent.tsx index c7d3cffb..0276916f 100644 --- a/web/src/components/Chat/ChatContent.tsx +++ b/web/src/components/Chat/ChatContent.tsx @@ -2,15 +2,15 @@ * @Author: ZhaoYing * @Date: 2025-12-10 16:46:17 * @Last Modified by: ZhaoYing - * @Last Modified time: 2026-03-19 19:45:40 + * @Last Modified time: 2026-03-27 14:17:38 */ import { type FC, useRef, useEffect, useState } from 'react' import clsx from 'clsx' import Markdown from '@/components/Markdown' import type { ChatContentProps } from './types' -import { Spin, Divider, Space, Image, Flex } from 'antd' +import { Spin, Divider, Space, Image, Flex, Button } from 'antd' import { SoundOutlined } from '@ant-design/icons' - +import { t } from 'i18next' const getFileUrl = (file: any) => { return file.thumbUrl || file.url || (file.originFileObj ? URL.createObjectURL(file.originFileObj) : undefined) @@ -29,17 +29,19 @@ const ChatContent: FC = ({ labelPosition = 'bottom', labelFormat, errorDesc, - renderRuntime + renderRuntime, + onSend }) => { // Scroll container reference for controlling auto-scroll to bottom const scrollContainerRef = useRef<(HTMLDivElement | null)>(null) const prevDataLengthRef = useRef(data.length); const isScrolledToBottomRef = useRef(true); const audioRef = useRef(null) - const [playingIndex, setPlayingIndex] = useState(null) + const [playingIndex, setPlayingIndex] = useState(null) - const handlePlay = (index: number, audio_url: string) => { - if (playingIndex === index) { + const handlePlay = (audio_url: string, audio_status?: string) => { + if (audio_status !== 'completed' && typeof audio_status === 'string') return + if (playingIndex === audio_url) { audioRef.current?.pause() setPlayingIndex(null) return @@ -50,7 +52,7 @@ const ChatContent: FC = ({ const audio = new Audio(audio_url) audioRef.current = audio audio.play() - setPlayingIndex(index) + setPlayingIndex(audio_url) audio.onended = () => setPlayingIndex(null) } @@ -77,12 +79,16 @@ const ChatContent: FC = ({ } }; }, []); - + // Auto-scroll to bottom when data changes to show latest messages // When data array length remains unchanged, if data is updated and user manually scrolled up, don't auto-scroll to bottom // When data array length changes, auto-scroll to bottom // If already scrolled to bottom, will auto-scroll to bottom useEffect(() => { + if (playingIndex && !data.some(item => item.meta_data?.audio_url === playingIndex)) { + audioRef.current?.pause() + setPlayingIndex(null) + } setTimeout(() => { if (scrollContainerRef.current) { // Auto-scroll if data length changed OR user is currently at bottom @@ -114,7 +120,7 @@ const ChatContent: FC = ({ : <> {/* Top label (such as timestamp, username, etc.) */} {labelPosition === 'top' && -
+
{labelFormat(item)}
} @@ -162,26 +168,56 @@ const ChatContent: FC = ({ })} } {/* Message bubble */} -
+ {item.status &&
} {item.subContent && renderRuntime && renderRuntime(item, index)} {/* Render message content using Markdown component */} + {item.meta_data?.suggested_questions && item.meta_data?.suggested_questions?.length > 0 && + {item.meta_data?.suggested_questions?.map((question, idx) => ( + + ))} + } + {item.meta_data?.citations && item.meta_data?.citations.length > 0 &&
+
{t('memoryConversation.citations')}
+ {item.meta_data?.citations?.map((citation, idx) => ( + + ))} +
} {item.meta_data?.audio_url && <> - {playingIndex !== index - ? handlePlay(index, item.meta_data?.audio_url!)} /> + {playingIndex !== item.meta_data?.audio_url && item.meta_data?.audio_status === 'pending' + ? + : playingIndex !== item.meta_data?.audio_url + ? handlePlay(item.meta_data?.audio_url!, item.meta_data?.audio_status)} /> :
handlePlay(index, item.meta_data?.audio_url!)} + onClick={() => handlePlay(item.meta_data?.audio_url!, item.meta_data?.audio_status)} /> } @@ -189,7 +225,7 @@ const ChatContent: FC = ({
{/* Bottom label (such as timestamp, username, etc.) */} {labelPosition === 'bottom' && -
+
{labelFormat(item)}
} diff --git a/web/src/components/Chat/ChatInput.tsx b/web/src/components/Chat/ChatInput.tsx index aa0dd2f6..6495ff06 100644 --- a/web/src/components/Chat/ChatInput.tsx +++ b/web/src/components/Chat/ChatInput.tsx @@ -2,15 +2,12 @@ * @Author: ZhaoYing * @Date: 2025-12-10 16:46:14 * @Last Modified by: ZhaoYing - * @Last Modified time: 2026-03-19 18:44:51 + * @Last Modified time: 2026-03-23 17:46:25 */ -import { type FC, useEffect, useMemo } from 'react' -import { Flex, Input, Form, Spin } from 'antd' +import { type FC, useEffect, useMemo, useState } from 'react' +import { Flex, Input, Spin } from 'antd' import clsx from 'clsx' -import SendIcon from '@/assets/images/conversation/send.svg' -import SendDisabledIcon from '@/assets/images/conversation/sendDisabled.svg' -import LoadingIcon from '@/assets/images/conversation/loading.svg' import type { ChatInputProps } from './types' /** @@ -27,37 +24,27 @@ const ChatInput: FC = ({ className = '', onChange }) => { - const [form] = Form.useForm() - const values = Form.useWatch([], form) - // Monitor form value changes to control send button state + const [inputValue, setInputValue] = useState('') + const [isFocus, setIsFocus] = useState(false) - // Clear form when external message is empty + // Clear input when external message is cleared useEffect(() => { - if (!message) { - form.setFieldsValue({ - message: undefined, - }) - } - }, [form, message]) - + if (!message) setInputValue('') + }, [message]) + // Clear input when loading useEffect(() => { - if (loading) { - form.setFieldsValue({ - message: undefined, - }) - } + if (loading) setInputValue('') }, [loading]) - const handleDelete = (file: any) => { fileChange?.(fileList?.filter(item => { return item.thumbUrl && file.thumbUrl ? item.thumbUrl !== file.thumbUrl : item.url && file.url ? item.url !== file.url - : item.uid !== file.uid + : item.uid !== file.uid }) || []) } - // Convert file object to preview URL + const previewFileList = useMemo(() => { return fileList?.map(file => ({ ...file, @@ -66,24 +53,27 @@ const ChatInput: FC = ({ }, [fileList]) const handleSend = () => { - if (loading || !values || !values?.message || values?.message?.trim() === '') return - onSend(values.message) + if (loading || !inputValue || inputValue.trim() === '') return + onSend(inputValue) } - console.log('previewFileList', previewFileList) + const canSend = !loading && inputValue.trim() !== '' return (
- - {previewFileList.length > 0 &&
+ + {previewFileList.length > 0 &&
+ {previewFileList.map((file) => { - if (file.type.includes('image')) { + if (file.type?.includes('image')) { return ( -
- {file.name} + {file.name}
handleDelete(file)} @@ -92,30 +82,30 @@ const ChatInput: FC = ({ ) } - if (file.type.includes('video')) { + if (file.type?.includes('video')) { return ( -
-
) } - if (file.type.includes('audio')) { + if (file.type?.includes('audio')) { return ( -
-
@@ -124,68 +114,86 @@ const ChatInput: FC = ({ } return ( -
- {file.type.includes('pdf') - ?
- : (file.type.includes('excel') || file.type.includes('spreadsheetml.sheet') || file.type.includes('csv')) - ?
- : (file.type.includes('doc') || file.type.includes('docx') || file.type.includes('word') || file.type.includes('wordprocessingml.document')) - ?
- : null - } +
{file.name}
-
{file.type} · {file.size}
+
{file.type?.split('/')[file.type?.split('/').length - 1]} · {file.size}
handleDelete(file)} >
-
+
) })} -
} - {/* Message input form */} -
- - onChange?.(e.target.value)} - onKeyDown={(e) => { - // Enter to send, Shift+Enter for new line - if (e.key === 'Enter' && !e.shiftKey && (e.target as HTMLTextAreaElement).value?.trim() !== '' && !loading) { - e.preventDefault(); - handleSend(); - } - }} - /> - -
+ +
} + {/* Message input area */} + { + setInputValue(e.target.value) + onChange?.(e.target.value) + }} + onKeyDown={(e) => { + // Enter to send, Shift+Enter for new line + if (e.key === 'Enter' && !e.shiftKey && (e.target as HTMLTextAreaElement).value?.trim() !== '' && !loading) { + e.preventDefault(); + handleSend(); + } + }} + onFocus={() => setIsFocus(true)} + onBlur={() => setIsFocus(false)} + /> {/* Bottom action area */} - - {/* Child component content (such as buttons) */} +
{children}
-
- {/* Send button - display different icons based on state */} - {loading - ? - : !values || !values?.message || values?.message?.trim() === '' - ? - : - } -
+ +
+
diff --git a/web/src/components/Chat/ChatToolbar.tsx b/web/src/components/Chat/ChatToolbar.tsx index 936e7e63..c5db0c4c 100644 --- a/web/src/components/Chat/ChatToolbar.tsx +++ b/web/src/components/Chat/ChatToolbar.tsx @@ -2,12 +2,11 @@ * @Author: ZhaoYing * @Date: 2026-03-17 14:22:25 * @Last Modified by: ZhaoYing - * @Last Modified time: 2026-03-19 18:59:37 + * @Last Modified time: 2026-03-27 17:54:47 */ // Toolbar component for chat input area, supporting file upload, audio recording, and variable configuration import { useRef, forwardRef, useImperativeHandle, type ReactNode, useEffect } from 'react' -import { Flex, Dropdown, Divider, App, Form, type MenuProps } from 'antd' -import { SettingOutlined } from '@ant-design/icons' +import { Flex, Dropdown, Divider, App, Form, type MenuProps, Tooltip } from 'antd' import { useTranslation } from 'react-i18next' import clsx from 'clsx' @@ -19,6 +18,8 @@ import type { FeaturesConfigForm } from '@/views/ApplicationConfig/types' import type { UploadFileListModalRef } from '@/views/Conversation/types' import type { VariableConfigModalRef } from '@/views/Workflow/types' import type { Variable } from '@/views/Workflow/components/Properties/VariableList/types' +import { getFileInfoByUrl } from '@/api/fileStorage'; +import { transform_file_type } from '@/views/Conversation/components/FileUpload' // Exposed methods via ref for parent components to access/set form state export interface ChatToolbarRef { @@ -31,7 +32,8 @@ export interface ChatToolbarRef { // Props for configuring toolbar features, upload settings, and event callbacks export interface ChatToolbarProps { features: FeaturesConfigForm - extra?: ReactNode + leftExtra?: ReactNode; + rightExtra?: ReactNode uploadAction?: string uploadRequestConfig?: { data?: Record @@ -52,7 +54,8 @@ interface FormValues { const max_file_count = 1; const ChatToolbar = forwardRef(({ features, - extra, + leftExtra, + rightExtra, uploadAction, uploadRequestConfig, onFilesChange, @@ -96,8 +99,6 @@ const ChatToolbar = forwardRef(({ } form.setFieldValue('files', [...lastFiles]) onFilesChange?.([...lastFiles]) - - console.log('lastFiles', lastFiles) } // Append recorded audio file to the file list and notify parent @@ -111,9 +112,33 @@ const ChatToolbar = forwardRef(({ // Merge a batch of files (e.g. from remote URL modal) into the file list const addFileList = (list?: any[]) => { if (!list?.length) return - const files = [...(queryValues?.files || []), ...list] + const uploadingList = list.map(f => ({ ...f, status: 'uploading' })) + const files = [...(queryValues?.files || []), ...uploadingList] form.setFieldValue('files', files) onFilesChange?.(files) + + uploadingList.forEach(file => { + getFileInfoByUrl(file.url) + .then((res) => { + const { file_name, file_size, content_type } = res as { file_name: string; file_size: number; content_type: string; } + const current: any[] = form.getFieldValue('files') || [] + const updated = current.map(f => f.uid === file.uid ? { + ...f, + status: 'done', + name: file_name, + size: file_size, + type: transform_file_type[content_type] || content_type, + } : f) + form.setFieldValue('files', updated) + onFilesChange?.(updated) + }) + .catch(() => { + const current: any[] = form.getFieldValue('files') || [] + const updated = current.map(f => f.uid === file.uid ? { ...f, status: 'error' } : f) + form.setFieldValue('files', updated) + onFilesChange?.(updated) + }) + }) } // Persist variable values from the config modal and notify parent @@ -163,28 +188,34 @@ const ChatToolbar = forwardRef(({ return (
- +