fix(memory): add coalesce defaults for activation_value and related node properties

- Add coalesce fallbacks for importance_score, activation_value, and
access_count in statement, entity, weak entity, strong entity, and
memory summary MERGE queries to prevent null values on new nodes
- Set activation_value default to coalesce(importance_score, 0.5) for
consistency with the forgetting/activation scoring logic
- Suppress Neo4j UNRECOGNIZED property key notifications in driver init
since missing keys like last_access_time and activation_value are
expected for newly created nodes

.github/workflows/release-notify-wechat.yml

@@ -0,0 +1,157 @@
+name: Release Notify Workflow
+
+on:
+  pull_request:
+    types: [closed]
+
+jobs:
+  notify:
+    if: >
+      github.event.pull_request.merged == true &&
+      startsWith(github.event.pull_request.base.ref, 'release')
+    runs-on: ubuntu-latest
+
+    steps:
+      # 防止 GitHub HEAD 未同步
+      - run: sleep 3
+
+      # 1️⃣ 获取分支 HEAD
+      - name: Get HEAD
+        id: head
+        run: |
+          HEAD_SHA=$(curl -s \
+            -H "Authorization: Bearer ${{ secrets.GITHUB_TOKEN }}" \
+            https://api.github.com/repos/${{ github.repository }}/git/ref/heads/${{ github.event.pull_request.base.ref }} \
+            | jq -r '.object.sha')
+          echo "head_sha=$HEAD_SHA" >> $GITHUB_OUTPUT
+
+      # 2️⃣ 判断是否最终PR
+      - name: Check Latest
+        id: check
+        run: |
+          if [ "${{ github.event.pull_request.merge_commit_sha }}" = "${{ steps.head.outputs.head_sha }}" ]; then
+            echo "ok=true" >> $GITHUB_OUTPUT
+          else
+            echo "ok=false" >> $GITHUB_OUTPUT
+          fi
+
+      # 3️⃣ 尝试从 PR body 提取 Sourcery 摘要
+      - name: Extract Sourcery Summary
+        if: steps.check.outputs.ok == 'true'
+        id: sourcery
+        env:
+          PR_BODY: ${{ github.event.pull_request.body }}
+        run: |
+          python3 << 'PYEOF'
+          import os, re
+
+          body = os.environ.get("PR_BODY", "") or ""
+          match = re.search(
+              r"## Summary by Sourcery\s*\n(.*?)(?=\n## |\Z)",
+              body,
+              re.DOTALL
+          )
+
+          if match:
+              summary = match.group(1).strip()
+              found = "true"
+          else:
+              summary = ""
+              found = "false"
+
+          with open("sourcery_summary.txt", "w", encoding="utf-8") as f:
+              f.write(summary)
+
+          with open(os.environ["GITHUB_OUTPUT"], "a") as gh:
+              gh.write(f"found={found}\n")
+              gh.write("summary<<EOF\n")
+              gh.write(summary + "\n")
+              gh.write("EOF\n")
+          PYEOF
+
+      # 4️⃣ Fallback: 获取 commits + 通义千问总结
+      - name: Get Commits
+        if: steps.check.outputs.ok == 'true' && steps.sourcery.outputs.found == 'false'
+        run: |
+          curl -s \
+            -H "Authorization: Bearer ${{ secrets.GITHUB_TOKEN }}" \
+            ${{ github.event.pull_request.commits_url }} \
+            | jq -r '.[].commit.message' | head -n 20 > commits.txt
+
+      - name: AI Summary (Qwen Fallback)
+        if: steps.check.outputs.ok == 'true' && steps.sourcery.outputs.found == 'false'
+        id: qwen
+        env:
+          DASHSCOPE_API_KEY: ${{ secrets.DASHSCOPE_API_KEY }}
+        run: |
+          python3 << 'PYEOF'
+          import json, os, urllib.request
+
+          with open("commits.txt", "r") as f:
+              commits = f.read().strip()
+
+          prompt = "请用中文总结以下代码提交，输出3-5条要点，面向测试人员。直接输出编号列表，不要输出标题或前言：\n" + commits
+          payload = {"model": "qwen-plus", "input": {"prompt": prompt}}
+          data = json.dumps(payload, ensure_ascii=False).encode("utf-8")
+
+          req = urllib.request.Request(
+              "https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation",
+              data=data,
+              headers={
+                  "Authorization": "Bearer " + os.environ["DASHSCOPE_API_KEY"],
+                  "Content-Type": "application/json"
+              }
+          )
+          resp = urllib.request.urlopen(req)
+          result = json.loads(resp.read().decode())
+          summary = result.get("output", {}).get("text", "AI 摘要生成失败")
+
+          with open(os.environ["GITHUB_OUTPUT"], "a") as gh:
+              gh.write("summary<<EOF\n")
+              gh.write(summary + "\n")
+              gh.write("EOF\n")
+          PYEOF
+
+      # 5️⃣ 企业微信通知（Markdown）
+      - name: Notify WeChat
+        if: steps.check.outputs.ok == 'true'
+        env:
+          WECHAT_WEBHOOK: ${{ secrets.WECHAT_WEBHOOK }}
+          BRANCH: ${{ github.event.pull_request.base.ref }}
+          AUTHOR: ${{ github.event.pull_request.user.login }}
+          PR_TITLE: ${{ github.event.pull_request.title }}
+          PR_URL: ${{ github.event.pull_request.html_url }}
+          SOURCERY_FOUND: ${{ steps.sourcery.outputs.found }}
+          SOURCERY_SUMMARY: ${{ steps.sourcery.outputs.summary }}
+          QWEN_SUMMARY: ${{ steps.qwen.outputs.summary }}
+        run: |
+          python3 << 'PYEOF'
+          import json, os, urllib.request
+
+          if os.environ.get("SOURCERY_FOUND") == "true":
+              label = "Summary by Sourcery"
+              summary = os.environ.get("SOURCERY_SUMMARY", "")
+          else:
+              label = "AI变更摘要"
+              summary = os.environ.get("QWEN_SUMMARY", "AI 摘要生成失败")
+
+          content = (
+            "## 🚀 Release 发布通知\n"
+            "> 📦 **分支**: " + os.environ["BRANCH"] + "\n"
+            "> 👤 **提交人**: " + os.environ["AUTHOR"] + "\n"
+            "> 📝 **标题**: " + os.environ["PR_TITLE"] + "\n\n"
+            "### 🧠 " + label + "\n" +
+            summary + "\n\n"
+            "---\n"
+            "🔗 [查看PR详情](" + os.environ["PR_URL"] + ")"
+          )
+          payload = {"msgtype": "markdown", "markdown": {"content": content}}
+          data = json.dumps(payload, ensure_ascii=False).encode("utf-8")
+          req = urllib.request.Request(
+            os.environ["WECHAT_WEBHOOK"],
+            data=data,
+            headers={"Content-Type": "application/json"}
+          )
+          resp = urllib.request.urlopen(req)
+          print(resp.read().decode())
+          PYEOF

.gitignore

@@ -27,6 +27,7 @@ time.log
 celerybeat-schedule.db
 search_results.json
 redbear-mem-metrics/
+redbear-mem-benchmark/
 pitch-deck/
 
 api/migrations/versions

api/app/celery_app.py

@@ -111,11 +111,17 @@ celery_app.conf.update(
         # 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'},
 
+        # Metadata extraction → memory_tasks queue
+        'app.tasks.extract_user_metadata': {'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'},
         'app.core.rag.tasks.sync_knowledge_for_kb': {'queue': 'document_tasks'},
 
+        # GraphRAG tasks → graphrag_tasks queue (独立队列，避免阻塞文档解析)
+        'app.core.rag.tasks.build_graphrag_for_kb': {'queue': 'graphrag_tasks'},
+        'app.core.rag.tasks.build_graphrag_for_document': {'queue': 'graphrag_tasks'},
+
         # Beat/periodic tasks → periodic_tasks queue (dedicated periodic worker)
         'app.tasks.workspace_reflection_task': {'queue': 'periodic_tasks'},
         'app.tasks.regenerate_memory_cache': {'queue': 'periodic_tasks'},

api/app/controllers/app_controller.py

@@ -1250,9 +1250,11 @@ async def export_app(
 async def import_app(
         file: UploadFile = File(...),
         db: Session = Depends(get_db),
-        current_user: User = Depends(get_current_user)
+        current_user: User = Depends(get_current_user),
+        app_id: Optional[str] = Form(None),
 ):
     """从 YAML 文件导入 agent / multi_agent / workflow 应用。
+    传入 app_id 时覆盖该应用的配置（类型必须一致），否则创建新应用。
     跨空间/跨租户导入时，模型/工具/知识库会按名称匹配，匹配不到则置空并返回 warnings。
     """
     if not file.filename.lower().endswith((".yaml", ".yml")):
@@ -1263,13 +1265,15 @@ async def import_app(
     if not dsl or "app" not in dsl:
         return fail(msg="YAML 格式无效，缺少 app 字段", code=BizCode.BAD_REQUEST)
 
-    new_app, warnings = AppDslService(db).import_dsl(
+    target_app_id = uuid.UUID(app_id) if app_id else None
+    result_app, warnings = AppDslService(db).import_dsl(
         dsl=dsl,
         workspace_id=current_user.current_workspace_id,
         tenant_id=current_user.tenant_id,
         user_id=current_user.id,
+        app_id=target_app_id,
     )
     return success(
-        data={"app": app_schema.App.model_validate(new_app), "warnings": warnings},
+        data={"app": app_schema.App.model_validate(result_app), "warnings": warnings},
         msg="应用导入成功" + ("，但部分资源需手动配置" if warnings else "")
     )

api/app/controllers/auth_controller.py

@@ -136,7 +136,7 @@ async def refresh_token(
     # 检查用户是否存在
     user = auth_service.get_user_by_id(db, userId)
     if not user:
-        raise BusinessException(t("auth.user.not_found"), code=BizCode.USER_NOT_FOUND)
+        raise BusinessException(t("auth.user.not_found"), code=BizCode.USER_NO_ACCESS)
     
     # 检查 refresh token 黑名单
     if settings.ENABLE_SINGLE_SESSION:

api/app/controllers/chunk_controller.py

@@ -23,6 +23,7 @@ from app.models.user_model import User
 from app.schemas import chunk_schema
 from app.schemas.response_schema import ApiResponse
 from app.services import knowledge_service, document_service, file_service, knowledgeshare_service
+from app.services.model_service import ModelApiKeyService
 
 # Obtain a dedicated API logger
 api_logger = get_api_logger()
@@ -460,18 +461,20 @@ async def retrieve_chunks(
             if retrieve_data.retrieve_type == chunk_schema.RetrieveType.Graph:
                 kb_ids = [str(kb_id) for kb_id in private_kb_ids]
                 workspace_ids = [str(workspace_id) for workspace_id in private_workspace_ids]
+                llm_key = ModelApiKeyService.get_available_api_key(db, db_knowledge.llm_id)
+                emb_key = ModelApiKeyService.get_available_api_key(db, db_knowledge.embedding_id)
                 # Prepare to configure chat_mdl、embedding_model、vision_model information
                 chat_model = Base(
-                    key=db_knowledge.llm.api_keys[0].api_key,
-                    model_name=db_knowledge.llm.api_keys[0].model_name,
-                    base_url=db_knowledge.llm.api_keys[0].api_base
+                    key=llm_key.api_key,
+                    model_name=llm_key.model_name,
+                    base_url=llm_key.api_base
                 )
                 embedding_model = OpenAIEmbed(
-                    key=db_knowledge.embedding.api_keys[0].api_key,
-                    model_name=db_knowledge.embedding.api_keys[0].model_name,
-                    base_url=db_knowledge.embedding.api_keys[0].api_base
+                    key=emb_key.api_key,
+                    model_name=emb_key.model_name,
+                    base_url=emb_key.api_base
                 )
-                doc = kg_retriever.retrieval(question=retrieve_data.query, workspace_ids=workspace_ids, kb_ids= kb_ids, emb_mdl=embedding_model, llm=chat_model)
+                doc = kg_retriever.retrieval(question=retrieve_data.query, workspace_ids=workspace_ids, kb_ids=kb_ids, emb_mdl=embedding_model, llm=chat_model)
                 if doc:
                     rs.insert(0, doc)
             return success(data=jsonable_encoder(rs), msg="retrieval successful")

api/app/controllers/prompt_optimizer_controller.py

@@ -124,10 +124,11 @@ async def get_prompt_opt(
                     skill=data.skill
             ):
                 # chunk 是 prompt 的增量内容
-                yield f"event:message\ndata: {json.dumps(chunk)}\n\n"
+                yield f"event:message\ndata: {json.dumps(chunk, ensure_ascii=False)}\n\n"
         except Exception as e:
             yield f"event:error\ndata: {json.dumps(
-                {"error": str(e)}
+                {"error": str(e)},
+                ensure_ascii=False
             )}\n\n"
         yield "event:end\ndata: {}\n\n"
 

api/app/controllers/service/app_api_controller.py

@@ -14,6 +14,7 @@ from app.core.response_utils import success
 from app.db import get_db
 from app.models.app_model import App
 from app.models.app_model import AppType
+from app.models.app_release_model import AppRelease
 from app.repositories import knowledge_repository
 from app.repositories.end_user_repository import EndUserRepository
 from app.schemas import AppChatRequest, conversation_schema
@@ -61,18 +62,18 @@ async def list_apps():
 #     return success(data={"received": True}, msg="消息已接收")
 
 
-def _checkAppConfig(app: App):
-    if app.type == AppType.AGENT:
-        if not app.current_release.config:
+def _checkAppConfig(release: AppRelease):
+    if release.type == AppType.AGENT:
+        if not release.config:
             raise BusinessException("Agent 应用未配置模型", BizCode.AGENT_CONFIG_MISSING)
-    elif app.type == AppType.MULTI_AGENT:
-        if not app.current_release.config:
+    elif release.type == AppType.MULTI_AGENT:
+        if not release.config:
             raise BusinessException("Multi-Agent 应用未配置模型", BizCode.AGENT_CONFIG_MISSING)
-    elif app.type == AppType.WORKFLOW:
-        if not app.current_release.config:
+    elif release.type == AppType.WORKFLOW:
+        if not release.config:
             raise BusinessException("工作流应用未配置模型", BizCode.AGENT_CONFIG_MISSING)
     else:
-        raise BusinessException("不支持的应用类型", BizCode.AGENT_CONFIG_MISSING)
+        raise BusinessException("不支持的应用类型", BizCode.APP_TYPE_NOT_SUPPORTED)
 
 
 @router.post("/chat")
@@ -86,10 +87,22 @@ async def chat(
         app_service: Annotated[AppService, Depends(get_app_service)] = None,
         message: str = Body(..., description="聊天消息内容"),
 ):
+    """
+    Agent/Workflow 聊天接口
+
+    - 不传 version：使用当前生效版本（current_release，回滚后为回滚目标版本）
+    - 传 version=release_id：使用指定版本uuid的历史快照，例如 {"version": "{{release_id}}"}
+    """
     body = await request.json()
     payload = AppChatRequest(**body)
 
     app = app_service.get_app(api_key_auth.resource_id, api_key_auth.workspace_id)
+
+    # 版本切换：指定 release_id 时查找对应历史快照，否则使用当前激活版本
+    if payload.version is not None:
+        active_release = app_service.get_release_by_id(app.id, payload.version)
+    else:
+        active_release = app.current_release
     other_id = payload.user_id
     workspace_id = api_key_auth.workspace_id
     end_user_repo = EndUserRepository(db)
@@ -127,7 +140,7 @@ async def chat(
             storage_type = 'neo4j'
     app_type = app.type
     # check app config
-    _checkAppConfig(app)
+    _checkAppConfig(active_release)
 
     # 获取或创建会话（提前验证）
     conversation = conversation_service.create_or_get_conversation(
@@ -142,7 +155,7 @@ async def chat(
 
         # print("="*50)
         # print(app.current_release.default_model_config_id)
-        agent_config = agent_config_4_app_release(app.current_release)
+        agent_config = agent_config_4_app_release(active_release)
         # print(agent_config.default_model_config_id)
 
         # thinking 开关：仅当 agent 配置了 deep_thinking 且请求 thinking=True 时才启用
@@ -194,7 +207,7 @@ async def chat(
         return success(data=conversation_schema.ChatResponse(**result).model_dump(mode="json"))
     elif app_type == AppType.MULTI_AGENT:
         # 多 Agent 流式返回
-        config = multi_agent_config_4_app_release(app.current_release)
+        config = multi_agent_config_4_app_release(active_release)
         if payload.stream:
             async def event_generator():
                 async for event in app_chat_service.multi_agent_chat_stream(
@@ -237,7 +250,7 @@ async def chat(
         return success(data=conversation_schema.ChatResponse(**result).model_dump(mode="json"))
     elif app_type == AppType.WORKFLOW:
         # 多 Agent 流式返回
-        config = workflow_config_4_app_release(app.current_release)
+        config = workflow_config_4_app_release(active_release)
         if payload.stream:
             async def event_generator():
                 async for event in app_chat_service.workflow_chat_stream(
@@ -253,7 +266,7 @@ async def chat(
                         user_rag_memory_id=user_rag_memory_id,
                         app_id=app.id,
                         workspace_id=workspace_id,
-                        release_id=app.current_release.id,
+                        release_id=active_release.id,
                         public=True
                 ):
                     event_type = event.get("event", "message")
@@ -288,7 +301,7 @@ async def chat(
             files=payload.files,
             app_id=app.id,
             workspace_id=workspace_id,
-            release_id=app.current_release.id
+            release_id=active_release.id
         )
         logger.debug(
             "工作流试运行返回结果",
@@ -302,6 +315,4 @@ async def chat(
             msg="工作流任务执行成功"
         )
     else:
-        from app.core.exceptions import BusinessException
-        from app.core.error_codes import BizCode
         raise BusinessException(f"不支持的应用类型: {app_type}", BizCode.APP_TYPE_NOT_SUPPORTED)

api/app/core/error_codes.py

@@ -41,6 +41,7 @@ class BizCode(IntEnum):
     FILE_NOT_FOUND = 4006
     APP_NOT_FOUND = 4007
     RELEASE_NOT_FOUND = 4008
+    USER_NO_ACCESS = 4009
 
     # 冲突/状态（5xxx）
     DUPLICATE_NAME = 5001
@@ -118,6 +119,7 @@ HTTP_MAPPING = {
     BizCode.WORKSPACE_ACCESS_DENIED: 403,
     BizCode.NOT_FOUND: 400,
     BizCode.USER_NOT_FOUND: 200,
+    BizCode.USER_NO_ACCESS: 401,
     BizCode.WORKSPACE_NOT_FOUND: 400,
     BizCode.MODEL_NOT_FOUND: 400,
     BizCode.KNOWLEDGE_NOT_FOUND: 400,

api/app/core/memory/agent/langgraph_graph/nodes/perceptual_retrieve_node.py

@@ -153,7 +153,7 @@ class PerceptualSearchService:
             return []
         try:
             r = await search_perceptual(
-                connector=connector, q=escaped,
+                connector=connector, query=escaped,
                 end_user_id=self.end_user_id,
                 limit=limit * 5,  # 多查一些以提高命中率
             )
@@ -178,7 +178,7 @@ class PerceptualSearchService:
             if not escaped.strip():
                 return []
             r = await search_perceptual(
-                connector=connector, q=escaped,
+                connector=connector, query=escaped,
                 end_user_id=self.end_user_id, limit=limit,
             )
             return r.get("perceptuals", [])

api/app/core/memory/agent/utils/write_tools.py

@@ -14,6 +14,7 @@ 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.deduplication.deduped_and_disamb import _USER_PLACEHOLDER_NAMES
 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
@@ -191,15 +192,37 @@ async def write(
             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:
+                    end_user_id = all_entity_nodes[0].end_user_id
+
+                    # Neo4j 写入完成后，用 PgSQL 权威 aliases 覆盖 Neo4j 用户实体
+                    try:
+                        from app.repositories.end_user_info_repository import EndUserInfoRepository
+                        if end_user_id:
+                            with get_db_context() as db_session:
+                                info = EndUserInfoRepository(db_session).get_by_end_user_id(uuid.UUID(end_user_id))
+                                pg_aliases = info.aliases if info and info.aliases else []
+                            if info is not None:
+                                # 将 Python 侧占位名集合作为参数传入，避免 Cypher 硬编码
+                                placeholder_names = list(_USER_PLACEHOLDER_NAMES)
+                                await neo4j_connector.execute_query(
+                                    """
+                                    MATCH (e:ExtractedEntity)
+                                    WHERE e.end_user_id = $end_user_id AND toLower(e.name) IN $placeholder_names
+                                    SET e.aliases = $aliases
+                                    """,
+                                    end_user_id=end_user_id, aliases=pg_aliases,
+                                    placeholder_names=placeholder_names,
+                                )
+                                logger.info(f"[AliasSync] Neo4j 用户实体 aliases 已用 PgSQL 权威源覆盖: {pg_aliases}")
+                    except Exception as sync_err:
+                        logger.warning(f"[AliasSync] PgSQL→Neo4j aliases 同步失败（不影响主流程）: {sync_err}")
+
+                    # 使用 Celery 异步任务触发聚类（不阻塞主流程）
                     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,
@@ -207,7 +230,6 @@ async def write(
                                 "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(
@@ -215,7 +237,6 @@ async def write(
                             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

api/app/core/memory/models/__init__.py

@@ -58,6 +58,14 @@ from app.core.memory.models.triplet_models import (
     TripletExtractionResponse,
 )
 
+# User metadata models
+from app.core.memory.models.metadata_models import (
+    UserMetadata,
+    UserMetadataProfile,
+    MetadataExtractionResponse,
+    MetadataFieldChange,
+)
+
 # Ontology scenario models (LLM extracted from scenarios)
 from app.core.memory.models.ontology_scenario_models import (
     OntologyClass,
@@ -124,6 +132,10 @@ __all__ = [
     "Entity",
     "Triplet",
     "TripletExtractionResponse",
+    "UserMetadata",
+    "UserMetadataProfile",
+    "MetadataExtractionResponse",
+    "MetadataFieldChange",
     # Ontology models
     "OntologyClass",
     "OntologyExtractionResponse",

api/app/core/memory/models/graph_models.py

@@ -364,12 +364,14 @@ class ChunkNode(Node):
     Attributes:
         dialog_id: ID of the parent dialog
         content: The text content of the chunk
+        speaker: Speaker identifier ('user' or 'assistant')
         chunk_embedding: Optional embedding vector for the chunk
         sequence_number: Order of this chunk within the dialog
         metadata: Additional chunk metadata as key-value pairs
     """
     dialog_id: str = Field(..., description="ID of the parent dialog")
     content: str = Field(..., description="The text content of the chunk")
+    speaker: Optional[str] = Field(None, description="Speaker identifier: 'user' for user messages, 'assistant' for AI responses")
     chunk_embedding: Optional[List[float]] = Field(None, description="Chunk embedding vector")
     sequence_number: int = Field(..., description="Order of this chunk within the dialog")
     metadata: dict = Field(default_factory=dict, description="Additional chunk metadata")

api/app/core/memory/models/metadata_models.py

@@ -0,0 +1,63 @@
+"""Models for user metadata extraction.
+
+Independent from triplet_models.py - these models are used by the
+standalone metadata extraction pipeline (post-dedup async Celery task).
+"""
+
+from typing import List, Literal, Optional
+
+from pydantic import BaseModel, ConfigDict, Field
+
+
+class UserMetadataProfile(BaseModel):
+    """用户画像信息"""
+
+    model_config = ConfigDict(extra="ignore")
+    role: List[str] = Field(default_factory=list, description="用户职业或角色")
+    domain: List[str] = Field(default_factory=list, description="用户所在领域")
+    expertise: List[str] = Field(
+        default_factory=list, description="用户擅长的技能或工具"
+    )
+    interests: List[str] = Field(
+        default_factory=list, description="用户关注的话题或领域标签"
+    )
+
+
+class UserMetadata(BaseModel):
+    """用户元数据顶层结构"""
+
+    model_config = ConfigDict(extra="ignore")
+    profile: UserMetadataProfile = Field(default_factory=UserMetadataProfile)
+
+
+class MetadataFieldChange(BaseModel):
+    """单个元数据字段的变更操作"""
+
+    model_config = ConfigDict(extra="ignore")
+    field_path: str = Field(
+        description="字段路径，用点号分隔，如 'profile.role'、'profile.expertise'"
+    )
+    action: Literal["set", "remove"] = Field(
+        description="操作类型：'set' 表示新增或修改，'remove' 表示移除"
+    )
+    value: Optional[str] = Field(
+        default=None,
+        description="字段的新值（action='set' 时必填）。标量字段直接填值，列表字段填单个要新增的元素"
+    )
+
+
+class MetadataExtractionResponse(BaseModel):
+    """元数据提取 LLM 响应结构（增量模式）"""
+
+    model_config = ConfigDict(extra="ignore")
+    metadata_changes: List[MetadataFieldChange] = Field(
+        default_factory=list,
+        description="元数据的增量变更列表，每项描述一个字段的新增、修改或移除操作",
+    )
+    aliases_to_add: List[str] = Field(
+        default_factory=list,
+        description="本次新发现的用户别名（用户自我介绍或他人对用户的称呼）",
+    )
+    aliases_to_remove: List[str] = Field(
+        default_factory=list, description="用户明确否认的别名（如'我不叫XX了'）"
+    )

api/app/core/memory/src/search.py

@@ -1,4 +1,3 @@
-import argparse
 import asyncio
 import json
 import math
@@ -6,7 +5,6 @@ import os
 import time
 from datetime import datetime
 from typing import TYPE_CHECKING, Any, Dict, List, Optional
-from uuid import UUID
 
 if TYPE_CHECKING:
     from app.schemas.memory_config_schema import MemoryConfig
@@ -23,7 +21,7 @@ from app.core.memory.utils.config.config_utils import (
 )
 from app.core.memory.utils.data.text_utils import extract_plain_query
 from app.core.memory.utils.data.time_utils import normalize_date_safe
-from app.core.memory.utils.llm.llm_utils import get_reranker_client
+# from app.core.memory.utils.llm.llm_utils import get_reranker_client
 from app.core.models.base import RedBearModelConfig
 from app.db import get_db_context
 from app.repositories.neo4j.graph_search import (
@@ -748,11 +746,10 @@ async def run_hybrid_search(
         if search_type in ["keyword", "hybrid"]:
             # Keyword-based search
             logger.info("[PERF] Starting keyword search...")
-            keyword_start = time.time()
             keyword_task = asyncio.create_task(
                 search_graph(
                     connector=connector,
-                    q=query_text,
+                    query=query_text,
                     end_user_id=end_user_id,
                     limit=limit,
                     include=include
@@ -762,7 +759,6 @@ async def run_hybrid_search(
         if search_type in ["embedding", "hybrid"]:
             # Embedding-based search
             logger.info("[PERF] Starting embedding search...")
-            embedding_start = time.time()
 
             # 从数据库读取嵌入器配置（按 ID）并构建 RedBearModelConfig
             config_load_start = time.time()
@@ -904,10 +900,10 @@ async def run_hybrid_search(
         else:
             results["latency_metrics"] = latency_metrics
 
-        logger.info(f"[PERF] ===== SEARCH PERFORMANCE SUMMARY =====")
+        logger.info("[PERF] ===== SEARCH PERFORMANCE SUMMARY =====")
         logger.info(f"[PERF] Total search completed in {total_latency:.4f}s")
         logger.info(f"[PERF] Latency breakdown: {json.dumps(latency_metrics, indent=2)}")
-        logger.info(f"[PERF] =========================================")
+        logger.info("[PERF] =========================================")
 
         # Sanitize results: drop large/unused fields
         _remove_keys_recursive(results, ["name_embedding"])  # drop entity name embeddings from outputs

api/app/core/memory/storage_services/extraction_engine/deduplication/deduped_and_disamb.py

@@ -82,51 +82,38 @@ def _merge_attribute(canonical: ExtractedEntityNode, ent: ExtractedEntityNode):
             canonical.connect_strength = next(iter(pair))
 
     # 别名合并（去重保序，使用标准化工具）
+    # 用户实体的 aliases 由 PgSQL end_user_info 作为唯一权威源，去重合并时不修改
     try:
         canonical_name = (getattr(canonical, "name", "") or "").strip()
-        incoming_name = (getattr(ent, "name", "") or "").strip()
-        
-        # 收集所有需要合并的别名
-        all_aliases = []
-        
-        # 1. 添加canonical现有的别名
-        existing = getattr(canonical, "aliases", []) or []
-        all_aliases.extend(existing)
-        
-        # 2. 添加incoming实体的名称（如果不同于canonical的名称）
-        if incoming_name and incoming_name != canonical_name:
-            all_aliases.append(incoming_name)
-        
-        # 3. 添加incoming实体的所有别名
-        incoming = getattr(ent, "aliases", []) or []
-        all_aliases.extend(incoming)
-        
-        # 4. 标准化并去重（优先使用alias_utils工具函数）
-        try:
-            from app.core.memory.utils.alias_utils import normalize_aliases
-            canonical.aliases = normalize_aliases(canonical_name, all_aliases)
-        except Exception:
-            # 如果导入失败，使用增强的去重逻辑
-            seen_normalized = set()
-            unique_aliases = []
+        if canonical_name.lower() not in _USER_PLACEHOLDER_NAMES:
+            incoming_name = (getattr(ent, "name", "") or "").strip()
             
-            for alias in all_aliases:
-                if not alias:
-                    continue
-                
-                alias_stripped = str(alias).strip()
-                if not alias_stripped or alias_stripped == canonical_name:
-                    continue
-                
-                # 标准化：转小写用于去重判断
-                alias_normalized = alias_stripped.lower()
-                
-                if alias_normalized not in seen_normalized:
-                    seen_normalized.add(alias_normalized)
-                    unique_aliases.append(alias_stripped)
+            # 收集所有需要合并的别名，过滤掉用户占位名避免污染非用户实体
+            all_aliases = list(getattr(canonical, "aliases", []) or [])
+            if incoming_name and incoming_name != canonical_name and incoming_name.lower() not in _USER_PLACEHOLDER_NAMES:
+                all_aliases.append(incoming_name)
+            all_aliases.extend(
+                a for a in (getattr(ent, "aliases", []) or [])
+                if a and a.strip().lower() not in _USER_PLACEHOLDER_NAMES
+            )
             
-            # 排序并赋值
-            canonical.aliases = sorted(unique_aliases)
+            try:
+                from app.core.memory.utils.alias_utils import normalize_aliases
+                canonical.aliases = normalize_aliases(canonical_name, all_aliases)
+            except Exception:
+                seen_normalized = set()
+                unique_aliases = []
+                for alias in all_aliases:
+                    if not alias:
+                        continue
+                    alias_stripped = str(alias).strip()
+                    if not alias_stripped or alias_stripped == canonical_name:
+                        continue
+                    alias_normalized = alias_stripped.lower()
+                    if alias_normalized not in seen_normalized:
+                        seen_normalized.add(alias_normalized)
+                        unique_aliases.append(alias_stripped)
+                canonical.aliases = sorted(unique_aliases)
     except Exception:
         pass
 
@@ -733,66 +720,37 @@ def fuzzy_match(
 
 
     def _merge_entities_with_aliases(canonical: ExtractedEntityNode, losing: ExtractedEntityNode):
-        """ 模糊匹配中的实体合并。
+        """模糊匹配中的实体合并（别名部分）。
         
-        合并策略：
-        1. 保留canonical的主名称不变
-        2. 将losing的主名称添加为alias（如果不同）
-        3. 合并两个实体的所有aliases
-        4. 自动去重（case-insensitive）并排序
-        
-        Args:
-            canonical: 规范实体（保留）
-            losing: 被合并实体（删除）
-            
-        Note:
-            使用alias_utils.normalize_aliases进行标准化去重
+        用户实体的 aliases 由 PgSQL end_user_info 作为唯一权威源，跳过合并。
         """
-        # 获取规范实体的名称
         canonical_name = (getattr(canonical, "name", "") or "").strip()
+        if canonical_name.lower() in _USER_PLACEHOLDER_NAMES:
+            return
+
         losing_name = (getattr(losing, "name", "") or "").strip()
         
-        # 收集所有需要合并的别名
-        all_aliases = []
-        
-        # 1. 添加canonical现有的别名
-        current_aliases = getattr(canonical, "aliases", []) or []
-        all_aliases.extend(current_aliases)
-        
-        # 2. 添加losing实体的名称（如果不同于canonical的名称）
+        all_aliases = list(getattr(canonical, "aliases", []) or [])
         if losing_name and losing_name != canonical_name:
             all_aliases.append(losing_name)
+        all_aliases.extend(getattr(losing, "aliases", []) or [])
         
-        # 3. 添加losing实体的所有别名
-        losing_aliases = getattr(losing, "aliases", []) or []
-        all_aliases.extend(losing_aliases)
-        
-        # 4. 标准化并去重（使用标准化后的字符串进行去重）
         try:
             from app.core.memory.utils.alias_utils import normalize_aliases
             canonical.aliases = normalize_aliases(canonical_name, all_aliases)
         except Exception:
-            # 如果导入失败，使用增强的去重逻辑
-            # 使用标准化后的字符串作为key进行去重
             seen_normalized = set()
             unique_aliases = []
-            
             for alias in all_aliases:
                 if not alias:
                     continue
-                
                 alias_stripped = str(alias).strip()
                 if not alias_stripped or alias_stripped == canonical_name:
                     continue
-                
-                # 标准化：转小写用于去重判断
                 alias_normalized = alias_stripped.lower()
-                
                 if alias_normalized not in seen_normalized:
                     seen_normalized.add(alias_normalized)
                     unique_aliases.append(alias_stripped)
-            
-            # 排序并赋值
             canonical.aliases = sorted(unique_aliases)
     
     # ========== 主循环：遍历所有实体对进行模糊匹配 ==========

api/app/core/memory/storage_services/extraction_engine/extraction_orchestrator.py

@@ -311,10 +311,53 @@ class ExtractionOrchestrator:
                 dialog_data_list,
             )
 
-            # 步骤 7: 同步用户别名到数据库表（仅正式模式）
+            # 步骤 7: 触发异步元数据和别名提取（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)
+                try:
+                    from app.core.memory.storage_services.extraction_engine.knowledge_extraction.metadata_extractor import (
+                        MetadataExtractor,
+                    )
+
+                    metadata_extractor = MetadataExtractor(
+                        llm_client=self.llm_client, language=self.language
+                    )
+                    user_statements = (
+                        metadata_extractor.collect_user_related_statements(
+                            entity_nodes, statement_nodes, statement_entity_edges
+                        )
+                    )
+                    if user_statements:
+                        end_user_id = (
+                            dialog_data_list[0].end_user_id
+                            if dialog_data_list
+                            else None
+                        )
+                        config_id = (
+                            dialog_data_list[0].config_id
+                            if dialog_data_list
+                            and hasattr(dialog_data_list[0], "config_id")
+                            else None
+                        )
+                        if end_user_id:
+                            from app.tasks import extract_user_metadata_task
+
+                            extract_user_metadata_task.delay(
+                                end_user_id=str(end_user_id),
+                                statements=user_statements,
+                                config_id=str(config_id) if config_id else None,
+                                language=self.language,
+                            )
+                            logger.info(
+                                f"已触发异步元数据提取任务，共 {len(user_statements)} 条用户相关 statement"
+                            )
+                    else:
+                        logger.info("未找到用户相关 statement，跳过元数据提取")
+                except Exception as e:
+                    logger.error(
+                        f"触发元数据提取任务失败（不影响主流程）: {e}", exc_info=True
+                    )
+
+                # 别名同步已迁移到 Celery 元数据提取任务中，不再在此处执行
 
             logger.info(f"知识提取流水线运行完成（{mode_str}）")
             return (
@@ -1107,6 +1150,7 @@ class ExtractionOrchestrator:
                     end_user_id=dialog_data.end_user_id,
                     run_id=dialog_data.run_id,  # 使用 dialog_data 的 run_id
                     content=chunk.content,
+                    speaker=getattr(chunk, 'speaker', None),
                     chunk_embedding=chunk.chunk_embedding,
                     sequence_number=chunk_idx,  # 添加必需的 sequence_number 字段
                     created_at=dialog_data.created_at,
@@ -1342,23 +1386,23 @@ class ExtractionOrchestrator:
     async def _update_end_user_other_name(
             self,
             entity_nodes: List[ExtractedEntityNode],
-            dialog_data_list: List[DialogData]
+            dialog_data_list: List[DialogData],
     ) -> None:
         """
         将本轮提取的用户别名同步到 end_user 和 end_user_info 表。
 
-        注意：此方法在 Neo4j 写入之前调用，因此不能依赖 Neo4j 作为别名的权威数据源。
-        改为直接使用内存中去重后的 entity_nodes 的 aliases，与 PgSQL 已有的 aliases 合并。
+        PgSQL end_user_info.aliases 是用户别名的唯一权威源。
+        此方法仅将本轮 LLM 从对话中新提取的别名增量追加到 PgSQL，
+        不再从 Neo4j 二层去重合并历史别名，避免脏数据反向污染 PgSQL。
 
         策略：
-        1. 从内存中的 entity_nodes 提取本轮用户别名（current_aliases）
-        2. 从去重后的 entity_nodes 中提取完整别名（含 Neo4j 二层去重合并的历史别名）
-        3. 从 PgSQL end_user_info 读取已有的 aliases（db_aliases）
-        4. 合并 db_aliases + deduped_aliases + current_aliases，去重保序
-        5. 写回 PgSQL
+        1. 从本轮对话原始发言中提取用户别名（current_aliases）
+        2. 从 PgSQL end_user_info 读取已有的 aliases（db_aliases）
+        3. 合并 db_aliases + current_aliases，去重保序
+        4. 写回 PgSQL
 
         Args:
-            entity_nodes: 去重后的实体节点列表（内存中，含二层去重合并结果）
+            entity_nodes: 去重后的实体节点列表（内存中）
             dialog_data_list: 对话数据列表
         """
         try:
@@ -1374,11 +1418,6 @@ class ExtractionOrchestrator:
             # 1. 提取本轮对话的用户别名（保持 LLM 提取的原始顺序，不排序）
             current_aliases = self._extract_current_aliases(entity_nodes, dialog_data_list)
 
-            # 1.5 从去重后的 entity_nodes 中提取完整别名
-            # 二层去重会将 Neo4j 中已有的历史别名合并到 entity_nodes 中，
-            # 这里提取出来确保 PgSQL 与 Neo4j 的别名保持同步
-            deduped_aliases = self._extract_deduped_entity_aliases(entity_nodes)
-
             # 1.6 从 Neo4j 查询已有的 AI 助手别名，作为额外的排除源
             # （防止 LLM 未提取出 AI 助手实体时，AI 别名泄漏到用户别名中）
             neo4j_assistant_aliases = await self._fetch_neo4j_assistant_aliases(end_user_id)
@@ -1390,19 +1429,12 @@ class ExtractionOrchestrator:
                 ]
                 if len(current_aliases) < before_count:
                     logger.info(f"通过 Neo4j AI 助手别名排除了 {before_count - len(current_aliases)} 个误归属别名")
-                # 同样过滤 deduped_aliases
-                deduped_aliases = [
-                    a for a in deduped_aliases
-                    if a.strip().lower() not in neo4j_assistant_aliases
-                ]
 
-            if not current_aliases and not deduped_aliases:
+            if not current_aliases:
                 logger.debug(f"本轮未提取到用户别名，跳过同步: end_user_id={end_user_id}")
                 return
 
             logger.info(f"本轮对话提取的 aliases: {current_aliases}")
-            if deduped_aliases:
-                logger.info(f"去重后实体的完整 aliases（含历史）: {deduped_aliases}")
 
             # 2. 同步到数据库
             end_user_uuid = uuid.UUID(end_user_id)
@@ -1413,21 +1445,15 @@ class ExtractionOrchestrator:
                     logger.warning(f"未找到 end_user_id={end_user_id} 的用户记录")
                     return
 
-                # 3. 从 PgSQL 读取已有 aliases 并与本轮合并
+                # 3. 从 PgSQL 读取已有 aliases 并与本轮新增合并
                 info = EndUserInfoRepository(db).get_by_end_user_id(end_user_uuid)
                 db_aliases = (info.aliases if info and info.aliases else [])
                 # 过滤掉占位名称
                 db_aliases = [a for a in db_aliases if a.strip().lower() not in self.USER_PLACEHOLDER_NAMES]
 
-                # 合并：已有 + 去重后完整别名 + 本轮新增，去重保序
+                # 合并：PgSQL 已有 + 本轮新增，去重保序（不再合并 Neo4j 历史别名）
                 merged_aliases = list(db_aliases)
                 seen_lower = {a.strip().lower() for a in merged_aliases}
-                # 先合并去重后实体的完整别名（含 Neo4j 历史别名）
-                for alias in deduped_aliases:
-                    if alias.strip().lower() not in seen_lower:
-                        merged_aliases.append(alias)
-                        seen_lower.add(alias.strip().lower())
-                # 再合并本轮新提取的别名
                 for alias in current_aliases:
                     if alias.strip().lower() not in seen_lower:
                         merged_aliases.append(alias)
@@ -1461,16 +1487,13 @@ class ExtractionOrchestrator:
                         info.aliases = merged_aliases
                         logger.info(f"同步合并后 aliases 到 end_user_info: {merged_aliases}")
                 else:
-                    first_alias = current_aliases[0].strip() if current_aliases else (
-                        deduped_aliases[0].strip() if deduped_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.lower() not in self.USER_PLACEHOLDER_NAMES:
                         db.add(EndUserInfo(
                             end_user_id=end_user_uuid,
                             other_name=first_alias,
                             aliases=merged_aliases,
-                            meta_data={}
                         ))
                         logger.info(f"创建 end_user_info 记录，other_name={first_alias}, aliases={merged_aliases}")
 
@@ -1478,9 +1501,6 @@ class ExtractionOrchestrator:
 
         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
     # 复用 deduped_and_disamb 模块级常量，避免重复维护
     USER_PLACEHOLDER_NAMES = _USER_PLACEHOLDER_NAMES
@@ -1587,7 +1607,6 @@ class ExtractionOrchestrator:
             if candidate and candidate.lower() in self.USER_PLACEHOLDER_NAMES:
                 return None
             return candidate
-        
         return None
 
     async def _run_dedup_and_write_summary(

api/app/core/memory/storage_services/extraction_engine/knowledge_extraction/metadata_extractor.py

@@ -0,0 +1,176 @@
+"""
+Metadata extractor module.
+
+Collects user-related statements from post-dedup graph data and
+extracts user metadata via an independent LLM call.
+"""
+
+import logging
+from typing import List, Optional
+
+from app.core.memory.models.graph_models import (
+    ExtractedEntityNode,
+    StatementEntityEdge,
+    StatementNode,
+)
+
+logger = logging.getLogger(__name__)
+
+# Reuse the same user-entity detection logic from dedup module
+_USER_NAMES = {"用户", "我", "user", "i"}
+_CANONICAL_USER_TYPE = "用户"
+
+
+def _is_user_entity(ent: ExtractedEntityNode) -> bool:
+    """判断实体是否为用户实体"""
+    name = (getattr(ent, "name", "") or "").strip().lower()
+    etype = (getattr(ent, "entity_type", "") or "").strip()
+    return name in _USER_NAMES or etype == _CANONICAL_USER_TYPE
+
+
+class MetadataExtractor:
+    """Extracts user metadata from post-dedup graph data via independent LLM call."""
+
+    def __init__(self, llm_client, language: Optional[str] = None):
+        self.llm_client = llm_client
+        self.language = language
+
+    @staticmethod
+    def detect_language(statements: List[str]) -> str:
+        """根据 statement 文本内容检测语言。
+        如果文本中包含中文字符则返回 "zh"，否则返回 "en"。
+        """
+        import re
+
+        combined = " ".join(statements)
+        if re.search(r"[\u4e00-\u9fff]", combined):
+            return "zh"
+        return "en"
+
+    def collect_user_related_statements(
+        self,
+        entity_nodes: List[ExtractedEntityNode],
+        statement_nodes: List[StatementNode],
+        statement_entity_edges: List[StatementEntityEdge],
+    ) -> List[str]:
+        """
+        从去重后的数据中筛选与用户直接相关且由用户发言的 statement 文本。
+
+        筛选逻辑：
+        1. 用户实体 → StatementEntityEdge → statement（直接关联）
+        2. 只保留 speaker="user" 的 statement（过滤 assistant 回复的噪声）
+
+        Returns:
+            用户发言的 statement 文本列表
+        """
+        # Find user entity IDs
+        user_entity_ids = set()
+        for ent in entity_nodes:
+            if _is_user_entity(ent):
+                user_entity_ids.add(ent.id)
+
+        if not user_entity_ids:
+            logger.debug("未找到用户实体节点，跳过 statement 收集")
+            return []
+
+        # 用户实体 → StatementEntityEdge → statement
+        target_stmt_ids = set()
+        for edge in statement_entity_edges:
+            if edge.target in user_entity_ids:
+                target_stmt_ids.add(edge.source)
+
+        # Collect: only speaker="user" statements, preserving order
+        result = []
+        seen = set()
+        total_associated = 0
+        skipped_non_user = 0
+        for stmt_node in statement_nodes:
+            if stmt_node.id in target_stmt_ids and stmt_node.id not in seen:
+                total_associated += 1
+                speaker = getattr(stmt_node, "speaker", None) or "unknown"
+                if speaker == "user":
+                    text = (stmt_node.statement or "").strip()
+                    if text:
+                        result.append(text)
+                else:
+                    skipped_non_user += 1
+                seen.add(stmt_node.id)
+
+        logger.info(
+            f"收集到 {len(result)} 条用户发言 statement "
+            f"(直接关联: {total_associated}, speaker=user: {len(result)}, "
+            f"跳过非user: {skipped_non_user})"
+        )
+        if result:
+            for i, text in enumerate(result):
+                logger.info(f"  [user statement {i + 1}] {text}")
+        if total_associated > 0 and len(result) == 0:
+            logger.warning(
+                f"有 {total_associated} 条直接关联 statement 但全部被 speaker 过滤，"
+                f"可能本次写入不包含 user 消息"
+            )
+        return result
+
+    async def extract_metadata(
+        self,
+        statements: List[str],
+        existing_metadata: Optional[dict] = None,
+        existing_aliases: Optional[List[str]] = None,
+    ) -> Optional[tuple]:
+        """
+        对筛选后的 statement 列表调用 LLM 提取元数据增量变更和用户别名。
+
+        Args:
+            statements: 用户发言的 statement 文本列表
+            existing_metadata: 数据库已有的元数据（可选）
+            existing_aliases: 数据库已有的用户别名列表（可选）
+
+        Returns:
+            (List[MetadataFieldChange], List[str], List[str]) tuple:
+            (metadata_changes, aliases_to_add, aliases_to_remove) on success, None on failure
+        """
+        if not statements:
+            return None
+
+        try:
+            from app.core.memory.utils.prompt.prompt_utils import prompt_env
+
+            if self.language:
+                detected_language = self.language
+                logger.info(f"元数据提取使用显式指定语言: {detected_language}")
+            else:
+                detected_language = self.detect_language(statements)
+                logger.info(f"元数据提取语言自动检测结果: {detected_language}")
+
+            template = prompt_env.get_template("extract_user_metadata.jinja2")
+            prompt = template.render(
+                statements=statements,
+                language=detected_language,
+                existing_metadata=existing_metadata,
+                existing_aliases=existing_aliases,
+                json_schema="",
+            )
+
+            from app.core.memory.models.metadata_models import (
+                MetadataExtractionResponse,
+            )
+
+            response = await self.llm_client.response_structured(
+                messages=[{"role": "user", "content": prompt}],
+                response_model=MetadataExtractionResponse,
+            )
+
+            if response:
+                changes = response.metadata_changes if response.metadata_changes else []
+                to_add = response.aliases_to_add if response.aliases_to_add else []
+                to_remove = (
+                    response.aliases_to_remove if response.aliases_to_remove else []
+                )
+                return changes, to_add, to_remove
+
+            logger.warning("LLM 返回的响应为空")
+            return None
+
+        except Exception as e:
+            logger.error(f"元数据提取 LLM 调用失败: {e}", exc_info=True)
+            return None

api/app/core/memory/storage_services/extraction_engine/knowledge_extraction/statement_extraction.py

@@ -1,6 +1,5 @@
 import asyncio
 import logging
-import os
 from datetime import datetime
 from typing import Any, Dict, List, Optional
 
@@ -82,6 +81,7 @@ class StatementExtractor:
         logger.warning(f"Chunk {getattr(chunk, 'id', 'unknown')} has no speaker field or is empty")
         return None
 
+
     async def _extract_statements(self, chunk, end_user_id: Optional[str] = None, dialogue_content: str = None) -> List[Statement]:
         """Process a single chunk and return extracted statements
 
@@ -94,7 +94,8 @@ class StatementExtractor:
             List of ExtractedStatement objects extracted from the chunk
         """
         chunk_content = chunk.content
-        
+        chunk_speaker = self._get_speaker_from_chunk(chunk)
+
         if not chunk_content or len(chunk_content.strip()) < 5:
             logger.warning(f"Chunk {chunk.id} content too short or empty, skipping")
             return []
@@ -149,8 +150,6 @@ class StatementExtractor:
                     relevence_info = RelevenceInfo[relevence_str] if relevence_str in RelevenceInfo.__members__ else RelevenceInfo.RELEVANT
                 except (KeyError, ValueError):
                     relevence_info = RelevenceInfo.RELEVANT
-               
-                chunk_speaker = self._get_speaker_from_chunk(chunk)
             
                 chunk_statement = Statement(
                     statement=extracted_stmt.statement,

api/app/core/memory/storage_services/extraction_engine/knowledge_extraction/triplet_extraction.py

@@ -1,4 +1,3 @@
-import os
 import asyncio
 from typing import List, Dict, Optional
 

api/app/core/memory/storage_services/forgetting_engine/access_history_manager.py

@@ -42,22 +42,21 @@ class AccessHistoryManager:
     - access_count: 访问次数
     
     特性：
-    - 原子性更新：使用Neo4j事务确保所有字段同时更新或回滚
-    - 并发安全：使用乐观锁机制防止并发冲突
+    - 原子性更新：使用 APOC 原子操作确保并发安全
+    - 批次内合并：同一批次中对同一节点的多次访问合并为一次更新
     - 一致性保证：提供一致性检查和自动修复功能
     - 智能修剪：自动修剪过长的访问历史
     
     Attributes:
         connector: Neo4j连接器实例
         actr_calculator: ACT-R激活值计算器实例
-        max_retries: 并发冲突时的最大重试次数
     """
     
     def __init__(
         self,
         connector: Neo4jConnector,
         actr_calculator: ACTRCalculator,
-        max_retries: int = 3
+        max_retries: int = 5
     ):
         """
         初始化访问历史管理器
@@ -65,47 +64,35 @@ class AccessHistoryManager:
         Args:
             connector: Neo4j连接器实例
             actr_calculator: ACT-R激活值计算器实例
-            max_retries: 并发冲突时的最大重试次数（默认3次）
+            max_retries: 已废弃，保留参数兼容性（APOC 原子操作无需重试）
         """
         self.connector = connector
         self.actr_calculator = actr_calculator
-        self.max_retries = max_retries
-    
+
     async def record_access(
         self,
         node_id: str,
         node_label: str,
         end_user_id: Optional[str] = None,
-        current_time: Optional[datetime] = None
+        current_time: Optional[datetime] = None,
+        access_times: int = 1
     ) -> Dict[str, Any]:
         """
         记录节点访问并原子性更新所有相关字段
         
-        这是核心方法，实现了：
-        1. 首次访问：初始化access_history，计算初始激活值
-        2. 后续访问：追加访问历史，重新计算激活值
-        3. 历史修剪：当历史过长时自动修剪
-        4. 原子性：所有字段在单个事务中更新
-        5. 并发安全：使用乐观锁重试机制
-        
         Args:
             node_id: 节点ID
             node_label: 节点标签（Statement, ExtractedEntity, MemorySummary）
             end_user_id: 组ID（可选，用于过滤）
             current_time: 当前时间（可选，默认使用系统时间）
+            access_times: 本次访问次数（默认1，批量合并时可能大于1）
         
         Returns:
-            Dict[str, Any]: 更新后的节点数据，包含：
-                - id: 节点ID
-                - activation_value: 更新后的激活值
-                - access_history: 更新后的访问历史
-                - last_access_time: 最后访问时间
-                - access_count: 访问次数
-                - importance_score: 重要性分数
+            Dict[str, Any]: 更新后的节点数据
         
         Raises:
             ValueError: 如果节点不存在或节点标签无效
-            RuntimeError: 如果重试次数耗尽仍然失败
+            RuntimeError: 如果更新失败
         """
         if current_time is None:
             current_time = datetime.now()
@@ -119,55 +106,48 @@ class AccessHistoryManager:
                 f"Invalid node_label: {node_label}. Must be one of {valid_labels}"
             )
         
-        # 使用乐观锁重试机制处理并发冲突
-        for attempt in range(self.max_retries):
-            try:
-                # 步骤1：读取当前节点状态
-                node_data = await self._fetch_node(node_id, node_label, end_user_id)
-                
-                if not node_data:
-                    raise ValueError(
-                        f"Node not found: {node_label} with id={node_id}"
-                    )
-                
-                # 步骤2：计算新的访问历史和激活值
-                update_data = await self._calculate_update(
-                    node_data=node_data,
-                    current_time=current_time,
-                    current_time_iso=current_time_iso
+        try:
+            # 步骤1：读取当前节点状态
+            node_data = await self._fetch_node(node_id, node_label, end_user_id)
+            
+            if not node_data:
+                raise ValueError(
+                    f"Node not found: {node_label} with id={node_id}"
                 )
-                
-                # 步骤3：原子性更新节点（使用事务）
-                updated_node = await self._atomic_update(
-                    node_id=node_id,
-                    node_label=node_label,
-                    update_data=update_data,
-                    end_user_id=end_user_id
-                )
-                
-                logger.info(
-                    f"成功记录访问: {node_label}[{node_id}], "
-                    f"activation={update_data['activation_value']:.4f}, "
-                    f"access_count={update_data['access_count']}"
-                )
-                
-                return updated_node
-                
-            except Exception as e:
-                if attempt < self.max_retries - 1:
-                    logger.warning(
-                        f"访问记录失败（尝试 {attempt + 1}/{self.max_retries}）: {str(e)}"
-                    )
-                    continue
-                else:
-                    logger.error(
-                        f"访问记录失败，重试次数耗尽: {node_label}[{node_id}], "
-                        f"错误: {str(e)}"
-                    )
-                    raise RuntimeError(
-                        f"Failed to record access after {self.max_retries} attempts: {str(e)}"
-                    )
-    
+            
+            # 步骤2：计算新的访问历史和激活值
+            update_data = await self._calculate_update(
+                node_data=node_data,
+                current_time=current_time,
+                current_time_iso=current_time_iso,
+                access_times=access_times
+            )
+            
+            # 步骤3：使用 APOC 原子操作更新节点（无需重试）
+            updated_node = await self._atomic_update(
+                node_id=node_id,
+                node_label=node_label,
+                update_data=update_data,
+                end_user_id=end_user_id
+            )
+            
+            logger.info(
+                f"成功记录访问: {node_label}[{node_id}], "
+                f"activation={update_data['activation_value']:.4f}, "
+                f"access_count={update_data['access_count']}"
+                f"{f', 合并访问次数={access_times}' if access_times > 1 else ''}"
+            )
+            
+            return updated_node
+            
+        except Exception as e:
+            logger.error(
+                f"访问记录失败: {node_label}[{node_id}], 错误: {str(e)}"
+            )
+            raise RuntimeError(
+                f"Failed to record access: {str(e)}"
+            ) from e
+
     async def record_batch_access(
         self,
         node_ids: List[str],
@@ -178,11 +158,10 @@ class AccessHistoryManager:
         """
         批量记录多个节点的访问
         
-        为提高性能，批量更新多个节点的访问历史。
-        每个节点独立更新，失败的节点不影响其他节点。
+        对同一个节点的多次访问会先在内存中合并，只发起一次更新。
         
         Args:
-            node_ids: 节点ID列表
+            node_ids: 节点ID列表（可包含重复ID）
             node_label: 节点标签（所有节点必须是同一类型）
             end_user_id: 组ID（可选）
             current_time: 当前时间（可选）
@@ -196,25 +175,38 @@ class AccessHistoryManager:
         if current_time is None:
             current_time = datetime.now()
         
-        # PERFORMANCE FIX: Process all nodes in parallel instead of sequentially
-        tasks = []
+        # 合并同一节点的访问次数，避免对同一节点并发写入
+        access_count_map: Dict[str, int] = {}
         for node_id in node_ids:
+            access_count_map[node_id] = access_count_map.get(node_id, 0) + 1
+        
+        merged_count = len(node_ids) - len(access_count_map)
+        if merged_count > 0:
+            logger.info(
+                f"批量访问合并: 原始={len(node_ids)}, "
+                f"去重后={len(access_count_map)}, 合并={merged_count}"
+            )
+        
+        # 对去重后的节点并行发起更新
+        tasks = []
+        for node_id, access_times in access_count_map.items():
             task = self.record_access(
                 node_id=node_id,
                 node_label=node_label,
                 end_user_id=end_user_id,
-                current_time=current_time
+                current_time=current_time,
+                access_times=access_times
             )
-            tasks.append(task)
+            tasks.append((node_id, task))
         
-        # Execute all tasks in parallel
-        task_results = await asyncio.gather(*tasks, return_exceptions=True)
+        task_results = await asyncio.gather(
+            *[t for _, t in tasks], return_exceptions=True
+        )
         
-        # Collect successful results and count failures
         results = []
         failed_count = 0
         
-        for node_id, result in zip(node_ids, task_results):
+        for (node_id, _), result in zip(tasks, task_results):
             if isinstance(result, Exception):
                 failed_count += 1
                 logger.warning(
@@ -225,12 +217,12 @@ class AccessHistoryManager:
         
         batch_duration = time.time() - batch_start
         logger.info(
-            f"[PERF] 批量访问记录完成: 成功 {len(results)}/{len(node_ids)}, "
+            f"[PERF] 批量访问记录完成: 成功 {len(results)}/{len(access_count_map)}, "
             f"失败 {failed_count}, 耗时 {batch_duration:.4f}s"
         )
         
         return results
-    
+
     async def check_consistency(
         self,
         node_id: str,
@@ -239,22 +231,6 @@ class AccessHistoryManager:
     ) -> Tuple[ConsistencyCheckResult, Optional[str]]:
         """
         检查节点数据的一致性
-        
-        验证以下一致性规则：
-        1. access_history[-1] == last_access_time
-        2. len(access_history) == access_count
-        3. 如果有访问历史，必须有激活值
-        4. 激活值必须在有效范围内 [offset, 1.0]
-        
-        Args:
-            node_id: 节点ID
-            node_label: 节点标签
-            end_user_id: 组ID（可选）
-        
-        Returns:
-            Tuple[ConsistencyCheckResult, Optional[str]]: 
-                - 一致性检查结果枚举
-                - 错误描述（如果不一致）
         """
         node_data = await self._fetch_node(node_id, node_label, end_user_id)
         
@@ -266,7 +242,6 @@ class AccessHistoryManager:
         access_count = node_data.get('access_count', 0)
         activation_value = node_data.get('activation_value')
         
-        # 检查1：access_history[-1] == last_access_time
         if access_history and last_access_time:
             if access_history[-1] != last_access_time:
                 return (
@@ -275,7 +250,6 @@ class AccessHistoryManager:
                     f"last_access_time={last_access_time}"
                 )
         
-        # 检查2：len(access_history) == access_count
         if len(access_history) != access_count:
             return (
                 ConsistencyCheckResult.INCONSISTENT_HISTORY_COUNT,
@@ -283,14 +257,12 @@ class AccessHistoryManager:
                 f"access_count={access_count}"
             )
         
-        # 检查3：有访问历史必须有激活值
         if access_history and activation_value is None:
             return (
                 ConsistencyCheckResult.MISSING_ACTIVATION,
                 "Node has access_history but activation_value is None"
             )
         
-        # 检查4：激活值范围
         if activation_value is not None:
             offset = self.actr_calculator.offset
             if not (offset <= activation_value <= 1.0):
@@ -301,30 +273,14 @@ class AccessHistoryManager:
                 )
         
         return ConsistencyCheckResult.CONSISTENT, None
-    
+
     async def check_batch_consistency(
         self,
         node_label: str,
         end_user_id: Optional[str] = None,
         limit: int = 1000
     ) -> Dict[str, Any]:
-        """
-        批量检查多个节点的一致性
-        
-        Args:
-            node_label: 节点标签
-            end_user_id: 组ID（可选）
-            limit: 检查的最大节点数
-        
-        Returns:
-            Dict[str, Any]: 一致性检查报告，包含：
-                - total_checked: 检查的节点总数
-                - consistent_count: 一致的节点数
-                - inconsistent_count: 不一致的节点数
-                - inconsistencies: 不一致节点的详细信息列表
-                - consistency_rate: 一致性率（0-1）
-        """
-        # 查询所有相关节点
+        """批量检查多个节点的一致性"""
         query = f"""
         MATCH (n:{node_label})
         WHERE n.access_history IS NOT NULL
@@ -343,7 +299,6 @@ class AccessHistoryManager:
         results = await self.connector.execute_query(query, **params)
         node_ids = [r['id'] for r in results]
         
-        # 检查每个节点
         inconsistencies = []
         consistent_count = 0
         
@@ -382,32 +337,15 @@ class AccessHistoryManager:
         )
         
         return report
-    
+
     async def repair_inconsistency(
         self,
         node_id: str,
         node_label: str,
         end_user_id: Optional[str] = None
     ) -> bool:
-        """
-        自动修复节点的数据不一致问题
-        
-        修复策略：
-        1. 如果access_history[-1] != last_access_time：使用access_history[-1]
-        2. 如果len(access_history) != access_count：使用len(access_history)
-        3. 如果有历史但无激活值：重新计算激活值
-        4. 如果激活值超出范围：重新计算激活值
-        
-        Args:
-            node_id: 节点ID
-            node_label: 节点标签
-            end_user_id: 组ID（可选）
-        
-        Returns:
-            bool: 修复成功返回True，否则返回False
-        """
+        """自动修复节点的数据不一致问题"""
         try:
-            # 检查一致性
             result, message = await self.check_consistency(
                 node_id=node_id,
                 node_label=node_label,
@@ -418,7 +356,6 @@ class AccessHistoryManager:
                 logger.info(f"节点数据一致，无需修复: {node_label}[{node_id}]")
                 return True
             
-            # 获取节点数据
             node_data = await self._fetch_node(node_id, node_label, end_user_id)
             if not node_data:
                 logger.error(f"节点不存在，无法修复: {node_label}[{node_id}]")
@@ -427,17 +364,13 @@ class AccessHistoryManager:
             access_history = node_data.get('access_history') or []
             importance_score = node_data.get('importance_score', 0.5)
             
-            # 准备修复数据
             repair_data = {}
             
-            # 修复last_access_time
             if access_history:
                 repair_data['last_access_time'] = access_history[-1]
             
-            # 修复access_count
             repair_data['access_count'] = len(access_history)
             
-            # 修复activation_value
             if access_history:
                 current_time = datetime.now()
                 last_access_dt = datetime.fromisoformat(access_history[-1])
@@ -453,7 +386,6 @@ class AccessHistoryManager:
                 )
                 repair_data['activation_value'] = activation_value
             
-            # 执行修复
             query = f"""
             MATCH (n:{node_label} {{id: $node_id}})
             """
@@ -484,26 +416,16 @@ class AccessHistoryManager:
                 f"修复节点失败: {node_label}[{node_id}], 错误: {str(e)}"
             )
             return False
-    
+
     # ==================== 私有辅助方法 ====================
-    
+
     async def _fetch_node(
         self,
         node_id: str,
         node_label: str,
         end_user_id: Optional[str] = None
     ) -> Optional[Dict[str, Any]]:
-        """
-        获取节点数据
-        
-        Args:
-            node_id: 节点ID
-            node_label: 节点标签
-            end_user_id: 组ID（可选）
-        
-        Returns:
-            Optional[Dict[str, Any]]: 节点数据，如果不存在返回None
-        """
+        """获取节点数据"""
         query = f"""
         MATCH (n:{node_label} {{id: $node_id}})
         """
@@ -527,12 +449,13 @@ class AccessHistoryManager:
         if results:
             return results[0]
         return None
-    
+
     async def _calculate_update(
         self,
         node_data: Dict[str, Any],
         current_time: datetime,
-        current_time_iso: str
+        current_time_iso: str,
+        access_times: int = 1
     ) -> Dict[str, Any]:
         """
         计算更新数据
@@ -541,45 +464,40 @@ class AccessHistoryManager:
             node_data: 当前节点数据
             current_time: 当前时间（datetime对象）
             current_time_iso: 当前时间（ISO格式字符串）
+            access_times: 本次访问次数（合并后可能大于1）
         
         Returns:
-            Dict[str, Any]: 更新数据，包含所有需要更新的字段
+            Dict[str, Any]: 更新数据
         """
-        access_history = node_data.get('access_history') or []
-        # Handle None importance_score - default to 0.5
         importance_score = node_data.get('importance_score')
         if importance_score is None:
             importance_score = 0.5
         
-        # 追加新的访问时间
-        new_access_history = access_history + [current_time_iso]
+        # 本次新增的时间戳
+        new_timestamps = [current_time_iso] * access_times
         
-        # 修剪访问历史（如果过长）
-        access_history_dt = [
-            datetime.fromisoformat(ts) for ts in new_access_history
-        ]
+        # 仅用本次新增的访问记录计算激活值
+        new_history_dt = [current_time] * access_times
         trimmed_history_dt = self.actr_calculator.trim_access_history(
-            access_history=access_history_dt,
+            access_history=new_history_dt,
             current_time=current_time
         )
-        trimmed_history = [ts.isoformat() for ts in trimmed_history_dt]
         
-        # 计算新的激活值
         activation_value = self.actr_calculator.calculate_memory_activation(
             access_history=trimmed_history_dt,
             current_time=current_time,
-            last_access_time=current_time,  # 最后访问时间就是当前时间
+            last_access_time=current_time,
             importance_score=importance_score
         )
         
-        # 返回所有需要更新的字段
         return {
             'activation_value': activation_value,
-            'access_history': trimmed_history,
+            'new_timestamps': new_timestamps,
+            'access_count_delta': access_times,
+            'access_count': len(trimmed_history_dt),
             'last_access_time': current_time_iso,
-            'access_count': len(trimmed_history)
         }
-    
+
     async def _atomic_update(
         self,
         node_id: str,
@@ -588,10 +506,10 @@ class AccessHistoryManager:
         end_user_id: Optional[str] = None
     ) -> Dict[str, Any]:
         """
-        原子性更新节点（使用乐观锁）
+        原子性更新节点（使用 APOC 原子操作）
         
-        使用Neo4j事务和版本号确保所有字段同时更新或回滚。
-        实现乐观锁机制防止并发冲突。
+        使用 apoc.atomic.add 和 apoc.atomic.insert 保证并发安全，
+        无需 version 字段和乐观锁，数据库层面保证原子性。
         
         Args:
             node_id: 节点ID
@@ -603,126 +521,68 @@ class AccessHistoryManager:
             Dict[str, Any]: 更新后的节点数据
         
         Raises:
-            RuntimeError: 如果更新失败或发生版本冲突
+            RuntimeError: 如果更新失败
         """
-        # 定义事务函数
-        async def update_transaction(tx, node_id, node_label, update_data, end_user_id):
-            # 步骤1：读取当前节点并获取版本号
-            read_query = f"""
-            MATCH (n:{node_label} {{id: $node_id}})
-            """
-            if end_user_id:
-                read_query += " WHERE n.end_user_id = $end_user_id"
-            read_query += """
-            RETURN n.id as id,
-                   n.version as version,
-                   n.activation_value as activation_value,
-                   n.access_history as access_history,
-                   n.last_access_time as last_access_time,
-                   n.access_count as access_count,
-                   n.importance_score as importance_score
-            """
+        content_field_map = {
+            'Statement': 'n.statement as statement',
+            'MemorySummary': 'n.content as content',
+            'ExtractedEntity': 'null as content_placeholder',
+            'Community': 'n.summary as summary'
+        }
+        
+        if node_label not in content_field_map:
+            raise ValueError(
+                f"Unsupported node_label: {node_label}. "
+                f"Supported labels are: {list(content_field_map.keys())}"
+            )
+        
+        content_field = content_field_map[node_label]
+        
+        where_clause = ""
+        if end_user_id:
+            where_clause = " AND n.end_user_id = $end_user_id"
+        
+        query = f"""
+        MATCH (n:{node_label} {{id: $node_id}})
+        WHERE true{where_clause}
+        CALL apoc.atomic.add(n, 'access_count', $access_count_delta, 5) YIELD oldValue AS old_count
+        WITH n
+        CALL (n) {{
+            UNWIND $new_timestamps AS ts
+            CALL apoc.atomic.insert(n, 'access_history', size(n.access_history), ts, 5) YIELD oldValue
+            RETURN count(*) AS inserted
+        }}
+        SET n.activation_value = $activation_value,
+            n.last_access_time = $last_access_time
+        RETURN n.id as id,
+               n.activation_value as activation_value,
+               n.access_history as access_history,
+               n.last_access_time as last_access_time,
+               n.access_count as access_count,
+               n.importance_score as importance_score,
+               {content_field}
+        """
+        
+        params = {
+            'node_id': node_id,
+            'access_count_delta': update_data['access_count_delta'],
+            'new_timestamps': update_data['new_timestamps'],
+            'activation_value': update_data['activation_value'],
+            'last_access_time': update_data['last_access_time'],
+        }
+        if end_user_id:
+            params['end_user_id'] = end_user_id
+        
+        try:
+            results = await self.connector.execute_query(query, **params)
             
-            read_params = {'node_id': node_id}
-            if end_user_id:
-                read_params['end_user_id'] = end_user_id
-            
-            read_result = await tx.run(read_query, **read_params)
-            current_node = await read_result.single()
-            
-            if not current_node:
+            if not results:
                 raise RuntimeError(f"Node not found: {node_label}[{node_id}]")
             
-            # 获取当前版本号（如果不存在则为0）
-            current_version = current_node.get('version', 0) or 0
-            new_version = current_version + 1
-            
-            # 步骤2：使用乐观锁更新节点
-            # 根据节点类型构建完整的查询语句
-            content_field_map = {
-                'Statement': 'n.statement as statement',
-                'MemorySummary': 'n.content as content',
-                'ExtractedEntity': 'null as content_placeholder'  # 占位符，后续会被过滤
-            }
-            
-            # 显式检查节点类型，不支持的类型抛出错误
-            if node_label not in content_field_map:
-                raise ValueError(
-                    f"Unsupported node_label: {node_label}. "
-                    f"Supported labels are: {list(content_field_map.keys())}"
-                )
-            
-            content_field = content_field_map[node_label]
-            
-            # 构建 WHERE 子句
-            where_conditions = []
-            if end_user_id:
-                where_conditions.append("n.end_user_id = $end_user_id")
-            
-            # 添加版本检查
-            if current_version > 0:
-                where_conditions.append("n.version = $current_version")
-            else:
-                where_conditions.append("(n.version IS NULL OR n.version = 0)")
-            
-            where_clause = " AND ".join(where_conditions) if where_conditions else "true"
-            
-            # 构建完整的更新查询
-            update_query = f"""
-            MATCH (n:{node_label} {{id: $node_id}})
-            WHERE {where_clause}
-            SET n.activation_value = $activation_value,
-                n.access_history = $access_history,
-                n.last_access_time = $last_access_time,
-                n.access_count = $access_count,
-                n.version = $new_version
-            RETURN n.id as id,
-                   n.activation_value as activation_value,
-                   n.access_history as access_history,
-                   n.last_access_time as last_access_time,
-                   n.access_count as access_count,
-                   n.importance_score as importance_score,
-                   n.version as version,
-                   {content_field}
-            """
-            
-            update_params = {
-                'node_id': node_id,
-                'current_version': current_version,
-                'new_version': new_version,
-                'activation_value': update_data['activation_value'],
-                'access_history': update_data['access_history'],
-                'last_access_time': update_data['last_access_time'],
-                'access_count': update_data['access_count']
-            }
-            if end_user_id:
-                update_params['end_user_id'] = end_user_id
-            
-            update_result = await tx.run(update_query, **update_params)
-            updated_node = await update_result.single()
-            
-            if not updated_node:
-                raise RuntimeError(
-                    f"Version conflict detected for {node_label}[{node_id}]. "
-                    f"Expected version {current_version}, but node was modified by another transaction."
-                )
-            
-            # 转换为字典并移除占位符字段
-            result_dict = dict(updated_node)
+            result_dict = dict(results[0])
             result_dict.pop('content_placeholder', None)
             
             return result_dict
-        
-        # 执行事务
-        try:
-            result = await self.connector.execute_write_transaction(
-                update_transaction,
-                node_id=node_id,
-                node_label=node_label,
-                update_data=update_data,
-                end_user_id=end_user_id
-            )
-            return result
         except Exception as e:
             logger.error(
                 f"原子性更新失败: {node_label}[{node_id}], 错误: {str(e)}"

api/app/core/memory/storage_services/search/keyword_search.py

@@ -5,7 +5,7 @@
 使用Neo4j的全文索引进行高效的文本匹配。
 """
 
-from typing import List, Dict, Any, Optional
+from typing import List, Optional
 from app.core.logging_config import get_memory_logger
 from app.repositories.neo4j.neo4j_connector import Neo4jConnector
 from app.core.memory.storage_services.search.search_strategy import SearchStrategy, SearchResult
@@ -74,7 +74,7 @@ class KeywordSearchStrategy(SearchStrategy):
             # 调用底层的关键词搜索函数
             results_dict = await search_graph(
                 connector=self.connector,
-                q=query_text,
+                query=query_text,
                 end_user_id=end_user_id,
                 limit=limit,
                 include=include_list

api/app/core/memory/utils/data/text_utils.py

@@ -22,7 +22,9 @@ def escape_lucene_query(query: str) -> str:
     s = s.replace("\r", " ").replace("\n", " ").strip()
 
     # Lucene reserved tokens/special characters
-    specials = ['&&', '||', '\\', '+', '-', '!', '(', ')', '{', '}', '[', ']', '^', '"', '~', '*', '?', ':']
+    # NOTE: '/' is the regex delimiter in Lucene — must be escaped to prevent
+    #       TokenMgrError when the query contains unmatched slashes.
+    specials = ['&&', '||', '\\', '+', '-', '!', '(', ')', '{', '}', '[', ']', '^', '"', '~', '*', '?', ':', '/']
     # Replace longer tokens first to avoid partial double-escaping
     for token in sorted(specials, key=len, reverse=True):
         s = s.replace(token, f"\\{token}")

api/app/core/memory/utils/prompt/prompts/extract_statement.jinja2

@@ -43,8 +43,9 @@ Each statement must be labeled as per the criteria mentioned below.
 
 对话上下文和共指消解：
 - 将每个陈述句归属于说出它的参与者。
-- 如果参与者列表为说话者提供了名称（例如，"李雪（用户）"），请在提取的陈述句中使用具体名称（"李雪"），而不是通用角色（"用户"）。
-- 将所有代词解析为对话上下文中的具体人物或实体。
+- **对于用户的发言：必须使用"用户"作为主语**，禁止将"用户"或"我"替换为用户的真实姓名或别名。例如，用户说"我叫张三"应提取为"用户叫张三"，而不是"张三叫张三"。
+- 对于 AI 助手的发言：使用"助手"或"AI助手"作为主语。
+- 将所有代词解析为对话上下文中的具体人物或实体，但"我"必须解析为"用户"。
 - 识别并将抽象引用解析为其具体名称（如果提到）。
 - 将缩写和首字母缩略词扩展为其完整形式。
 {% else %}
@@ -68,8 +69,9 @@ Context Resolution Requirements:
 
 Conversational Context & Co-reference Resolution:
 - Attribute every statement to the participant who uttered it.
-- If the participant list provides a name for a speaker (e.g., "李雪 (用户)"), use the specific name ("李雪") in the extracted statement, not the generic role ("用户").
-- Resolve all pronouns to the specific person or entity from the conversation's context.
+- **For user's statements: always use "用户" (User) as the subject**. Do NOT replace "用户" or "I" with the user's real name or alias. For example, if the user says "I'm John", extract as "用户 is John", not "John is John".
+- For AI assistant's statements: use "助手" or "AI助手" as the subject.
+- Resolve all pronouns to the specific person or entity from the conversation's context, but "I"/"我" must always resolve to "用户".
 - Identify and resolve abstract references to their specific names if mentioned.
 - Expand abbreviations and acronyms to their full form.
 {% endif %}
@@ -139,13 +141,13 @@ AI: "水彩画很有趣！水彩颜料通常由颜料与阿拉伯树胶等粘合
 示例输出: {
   "statements": [
     {
-      "statement": "Sarah Chen 最近一直在尝试水彩画。",
+      "statement": "用户最近一直在尝试水彩画。",
       "statement_type": "FACT",
       "temporal_type": "DYNAMIC",
       "relevance": "RELEVANT"
     },
     {
-      "statement": "Sarah Chen 画了一些花朵。",
+      "statement": "用户画了一些花朵。",
       "statement_type": "FACT",
       "temporal_type": "DYNAMIC",
       "relevance": "RELEVANT"
@@ -157,13 +159,13 @@ AI: "水彩画很有趣！水彩颜料通常由颜料与阿拉伯树胶等粘合
       "relevance": "IRRELEVANT"
     },
     {
-      "statement": "Sarah Chen 认为她的水彩画中的色彩组合可以改进。",
+      "statement": "用户认为她的水彩画中的色彩组合可以改进。",
       "statement_type": "OPINION",
       "temporal_type": "STATIC",
       "relevance": "RELEVANT"
     },
     {
-      "statement": "Sarah Chen 真的很喜欢玫瑰和百合。",
+      "statement": "用户真的很喜欢玫瑰和百合。",
       "statement_type": "FACT",
       "temporal_type": "STATIC",
       "relevance": "RELEVANT"
@@ -186,13 +188,13 @@ AI: "水彩画很有趣！水彩颜料通常由颜料和阿拉伯树胶等粘合
 示例输出: {
   "statements": [
     {
-      "statement": "张曼婷最近在尝试水彩画。",
+      "statement": "用户最近在尝试水彩画。",
       "statement_type": "FACT",
       "temporal_type": "DYNAMIC",
       "relevance": "RELEVANT"
     },
     {
-      "statement": "张曼婷画了一些花朵。",
+      "statement": "用户画了一些花朵。",
       "statement_type": "FACT",
       "temporal_type": "DYNAMIC",
       "relevance": "RELEVANT"
@@ -204,13 +206,13 @@ AI: "水彩画很有趣！水彩颜料通常由颜料和阿拉伯树胶等粘合
       "relevance": "IRRELEVANT"
     },
     {
-      "statement": "张曼婷觉得水彩画的色彩搭配还有提升的空间。",
+      "statement": "用户觉得水彩画的色彩搭配还有提升的空间。",
       "statement_type": "OPINION",
       "temporal_type": "STATIC",
       "relevance": "RELEVANT"
     },
     {
-      "statement": "张曼婷很喜欢玫瑰和百合。",
+      "statement": "用户很喜欢玫瑰和百合。",
       "statement_type": "FACT",
       "temporal_type": "STATIC",
       "relevance": "RELEVANT"
@@ -233,13 +235,13 @@ User: "I think the color combinations could use some improvement, but I really l
 Example Output: {
   "statements": [
     {
-      "statement": "Sarah Chen has been trying watercolor painting recently.",
+      "statement": "用户 has been trying watercolor painting recently.",
       "statement_type": "FACT",
       "temporal_type": "DYNAMIC",
       "relevance": "RELEVANT"
     },
     {
-      "statement": "Sarah Chen painted some flowers.",
+      "statement": "用户 painted some flowers.",
       "statement_type": "FACT",
       "temporal_type": "DYNAMIC",
       "relevance": "RELEVANT"
@@ -251,13 +253,13 @@ Example Output: {
       "relevance": "IRRELEVANT"
     },
     {
-      "statement": "Sarah Chen thinks the color combinations in her watercolor paintings could use some improvement.",
+      "statement": "用户 thinks the color combinations in her watercolor paintings could use some improvement.",
       "statement_type": "OPINION",
       "temporal_type": "STATIC",
       "relevance": "RELEVANT"
     },
     {
-      "statement": "Sarah Chen really likes roses and lilies.",
+      "statement": "用户 really likes roses and lilies.",
       "statement_type": "FACT",
       "temporal_type": "STATIC",
       "relevance": "RELEVANT"
@@ -280,13 +282,13 @@ AI: "水彩画很有趣！水彩颜料通常由颜料和阿拉伯树胶等粘合
 Example Output: {
   "statements": [
     {
-      "statement": "张曼婷最近在尝试水彩画。",
+      "statement": "用户最近在尝试水彩画。",
       "statement_type": "FACT",
       "temporal_type": "DYNAMIC",
       "relevance": "RELEVANT"
     },
     {
-      "statement": "张曼婷画了一些花朵。",
+      "statement": "用户画了一些花朵。",
       "statement_type": "FACT",
       "temporal_type": "DYNAMIC",
       "relevance": "RELEVANT"
@@ -298,13 +300,13 @@ Example Output: {
       "relevance": "IRRELEVANT"
     },
     {
-      "statement": "张曼婷觉得水彩画的色彩搭配还有提升的空间。",
+      "statement": "用户觉得水彩画的色彩搭配还有提升的空间。",
       "statement_type": "OPINION",
       "temporal_type": "STATIC",
       "relevance": "RELEVANT"
     },
     {
-      "statement": "张曼婷很喜欢玫瑰和百合。",
+      "statement": "用户很喜欢玫瑰和百合。",
       "statement_type": "FACT",
       "temporal_type": "STATIC",
       "relevance": "RELEVANT"

api/app/core/memory/utils/prompt/prompts/extract_triplet.jinja2

@@ -406,4 +406,12 @@ Output:
 - **⚠️ ALIASES ORDER: preserve temporal order of appearance**
 - **🚨 MANDATORY FIELD: EVERY entity MUST include "aliases" field, even if empty array []**
 
+**Output JSON structure:**
+```json
+{
+  "triplets": [...],
+  "entities": [...]
+}
+```
+
 {{ json_schema }}

api/app/core/memory/utils/prompt/prompts/extract_user_metadata.jinja2

@@ -0,0 +1,140 @@
+===Task===
+Extract user metadata changes from the following conversation statements spoken by the user.
+
+{% if language == "zh" %}
+**"三度原则"判断标准：**
+- 复用度：该信息是否会被多个功能模块使用？
+- 约束度：该信息是否会影响系统行为？
+- 时效性：该信息是长期稳定的还是临时的？仅提取长期稳定信息。
+
+**提取规则：**
+- **只提取关于"用户本人"的画像信息**，忽略用户提到的第三方人物（如朋友、同事、家人）的信息
+- 仅提取文本中明确提到的信息，不要推测
+- **输出语言必须与输入文本的语言一致**（输入中文则输出中文值，输入英文则输出英文值）
+
+**增量模式（重要）：**
+你只需要输出**本次对话引起的变更操作**，不要输出完整的元数据。每个变更是一个对象，包含：
+- `field_path`：字段路径，用点号分隔（如 `profile.role`、`profile.expertise`）
+- `action`：操作类型
+  * `set`：新增或修改一个字段的值
+  * `remove`：移除一个字段的值
+- `value`：字段的新值（`action="set"` 时必填，`action="remove"` 时填要移除的元素值）
+  * 所有字段均为列表类型，每个元素一条变更记录
+
+**判断规则：**
+- 用户提到新信息 → `action="set"`，填入新值
+- 用户明确否定已有信息（如"我不再做老师了"、"我已经不学Python了"）→ `action="remove"`，`value` 填要移除的元素值
+- 如果本次对话没有任何可提取的变更，返回空的 `metadata_changes` 数组 `[]`
+- **不要为未被提及的字段生成任何变更操作**
+
+{% if existing_metadata %}
+**已有元数据（仅供参考，用于判断是否需要变更）：**
+请对比已有数据和用户最新发言，只输出差异部分的变更操作。
+- 如果用户说的信息和已有数据一致，不需要输出变更
+- 如果用户否定了已有数据中的某个值，输出 `remove` 操作
+- 如果用户提到了新信息，输出 `set` 操作
+{% endif %}
+
+**字段说明：**
+- profile.role：用户的职业或角色（列表），如 教师、医生、后端工程师，一个人可以有多个角色
+- profile.domain：用户所在领域（列表），如 教育、医疗、软件开发，一个人可以涉及多个领域
+- profile.expertise：用户擅长的技能或工具（列表），如 Python、心理咨询、高中物理
+- profile.interests：用户主动表达兴趣的话题或领域标签（列表）
+
+**用户别名变更（增量模式）：**
+- **aliases_to_add**：本次新发现的用户别名，包括：
+  * 用户主动自我介绍：如"我叫张三"、"我的名字是XX"、"我的网名是XX"
+  * 他人对用户的称呼：如"同事叫我陈哥"、"大家叫我小张"、"领导叫我老陈"
+  * 只提取原文中逐字出现的名字，严禁推测或创造
+  * 禁止提取：用户给 AI 取的名字、第三方人物自身的名字、"用户"/"我" 等占位词
+  * 如果没有新别名，返回空数组 `[]`
+- **aliases_to_remove**：用户明确否认的别名，包括：
+  * 用户说"我不叫XX了"、"别叫我XX"、"我改名了，不叫XX" → 将 XX 放入此数组
+  * **严格限制**：只将用户原文中**逐字提到**的被否认名字放入，不要推断关联的其他别名
+  * 如果没有要移除的别名，返回空数组 `[]`
+{% if existing_aliases %}
+- 已有别名：{{ existing_aliases | tojson }}（仅供参考，不需要在输出中重复）
+{% endif %}
+{% else %}
+**"Three-Degree Principle" criteria:**
+- Reusability: Will this information be used by multiple functional modules?
+- Constraint: Will this information affect system behavior?
+- Timeliness: Is this information long-term stable or temporary? Only extract long-term stable information.
+
+**Extraction rules:**
+- **Only extract profile information about the user themselves**, ignore information about third parties (friends, colleagues, family) mentioned by the user
+- Only extract information explicitly mentioned in the text, do not speculate
+- **Output language must match the input text language**
+
+**Incremental mode (important):**
+You should only output **the change operations caused by this conversation**, not the complete metadata. Each change is an object containing:
+- `field_path`: Field path separated by dots (e.g. `profile.role`, `profile.expertise`)
+- `action`: Operation type
+  * `set`: Add or update a field value
+  * `remove`: Remove a field value
+- `value`: The new value for the field (required when `action="set"`, for `action="remove"` fill in the element value to remove)
+  * All fields are list types, one change record per element
+
+**Decision rules:**
+- User mentions new information → `action="set"`, fill in the new value
+- User explicitly negates existing info (e.g. "I'm no longer a teacher", "I stopped learning Python") → `action="remove"`, `value` is the element to remove
+- If this conversation has no extractable changes, return an empty `metadata_changes` array `[]`
+- **Do NOT generate any change operations for fields not mentioned in the conversation**
+
+{% if existing_metadata %}
+**Existing metadata (for reference only, to determine if changes are needed):**
+Compare existing data with the user's latest statements, and only output change operations for the differences.
+- If the user's statement matches existing data, no change is needed
+- If the user negates a value in existing data, output a `remove` operation
+- If the user mentions new information, output a `set` operation
+{% endif %}
+
+**Field descriptions:**
+- profile.role: User's occupation or role (list), e.g. teacher, doctor, software engineer. A person can have multiple roles
+- profile.domain: User's domain (list), e.g. education, healthcare, software development. A person can span multiple domains
+- profile.expertise: User's skills or tools (list), e.g. Python, counseling, physics
+- profile.interests: Topics or domain tags the user actively expressed interest in (list)
+
+**User alias changes (incremental mode):**
+- **aliases_to_add**: Newly discovered user aliases from this conversation, including:
+  * User self-introductions: e.g. "I'm John", "My name is XX", "My username is XX"
+  * How others address the user: e.g. "My colleagues call me Johnny", "People call me Mike"
+  * Only extract names that appear VERBATIM in the text — never infer or fabricate
+  * Do NOT extract: names the user gives to the AI, third-party people's own names, placeholder words like "User"/"I"
+  * If no new aliases, return empty array `[]`
+- **aliases_to_remove**: Aliases the user explicitly denies, including:
+  * User says "Don't call me XX anymore", "I'm not called XX", "I changed my name from XX" → put XX in this array
+  * **Strict rule**: Only include the exact name the user **verbatim mentions** as denied. Do NOT infer or remove related aliases
+  * If no aliases to remove, return empty array `[]`
+{% if existing_aliases %}
+- Existing aliases: {{ existing_aliases | tojson }} (for reference only, do not repeat in output)
+{% endif %}
+{% endif %}
+
+===User Statements===
+{% for stmt in statements %}
+- {{ stmt }}
+{% endfor %}
+
+{% if existing_metadata %}
+===Existing User Metadata===
+```json
+{{ existing_metadata | tojson }}
+```
+{% endif %}
+
+===Output Format===
+Return a JSON object with the following structure:
+```json
+{
+  "metadata_changes": [
+    {"field_path": "profile.role", "action": "set", "value": "后端工程师"},
+    {"field_path": "profile.expertise", "action": "set", "value": "Python"},
+    {"field_path": "profile.expertise", "action": "remove", "value": "Java"}
+  ],
+  "aliases_to_add": [],
+  "aliases_to_remove": []
+}
+```
+
+{{ json_schema }}

api/app/core/models/base.py

@@ -112,22 +112,23 @@ class RedBearModelFactory:
                 params["stream_usage"] = True
             # 深度思考模式
             is_streaming = bool(config.extra_params.get("streaming"))
-            if is_streaming and not config.is_omni:
-                if provider == ModelProvider.VOLCANO:
-                    # 火山引擎深度思考仅流式调用支持，非流式时不传 thinking 参数
-                    thinking_config: Dict[str, Any] = {
-                        "type": "enabled" if config.deep_thinking else "disabled"
-                    }
-                    if config.deep_thinking and config.thinking_budget_tokens:
-                        thinking_config["budget_tokens"] = config.thinking_budget_tokens
-                    params["extra_body"] = {"thinking": thinking_config}
-                else:
-                    # 始终显式传递 enable_thinking，不支持该参数的模型（如 DeepSeek-R1）会直接忽略
-                    model_kwargs: Dict[str, Any] = config.extra_params.get("model_kwargs", {})
-                    model_kwargs["enable_thinking"] = config.deep_thinking
-                    if config.deep_thinking and config.thinking_budget_tokens:
-                        model_kwargs["thinking_budget"] = config.thinking_budget_tokens
-                    params["model_kwargs"] = model_kwargs
+            if config.support_thinking:
+                if is_streaming and not config.is_omni:
+                    if provider == ModelProvider.VOLCANO:
+                        # 火山引擎深度思考仅流式调用支持，非流式时不传 thinking 参数
+                        thinking_config: Dict[str, Any] = {
+                            "type": "enabled" if config.deep_thinking else "disabled"
+                        }
+                        if config.deep_thinking and config.thinking_budget_tokens:
+                            thinking_config["budget_tokens"] = config.thinking_budget_tokens
+                        params["extra_body"] = {"thinking": thinking_config}
+                    else:
+                        # 始终显式传递 enable_thinking，不支持该参数的模型（如 DeepSeek-R1）会直接忽略
+                        model_kwargs: Dict[str, Any] = config.extra_params.get("model_kwargs", {})
+                        model_kwargs["enable_thinking"] = config.deep_thinking
+                        if config.deep_thinking and config.thinking_budget_tokens:
+                            model_kwargs["thinking_budget"] = config.thinking_budget_tokens
+                        params["model_kwargs"] = model_kwargs
             return params
         elif provider == ModelProvider.DASHSCOPE:
             params = {
@@ -206,10 +207,15 @@ class RedBearModelFactory:
         if provider in [ModelProvider.XINFERENCE, ModelProvider.GPUSTACK]:
             return {
                 "model": config.model_name,
-                # "base_url": config.base_url,
                 "jina_api_key": config.api_key,
                 **config.extra_params
             }
+        elif provider == ModelProvider.DASHSCOPE:
+            return {
+                "model": config.model_name,
+                "dashscope_api_key": config.api_key,
+                **config.extra_params
+            }
         else:
             raise BusinessException(f"不支持的提供商: {provider}", code=BizCode.PROVIDER_NOT_SUPPORTED)
 
@@ -265,6 +271,9 @@ def get_provider_rerank_class(provider: str):
     if provider in [ModelProvider.XINFERENCE, ModelProvider.GPUSTACK]:
         from langchain_community.document_compressors import JinaRerank
         return JinaRerank
+    elif provider == ModelProvider.DASHSCOPE:
+        from langchain_community.document_compressors.dashscope_rerank import DashScopeRerank
+        return DashScopeRerank
         # elif provider == ModelProvider.OLLAMA:
     #     from langchain_ollama import OllamaEmbeddings
     #     return OllamaEmbeddings

api/app/core/models/embedding.py

@@ -36,9 +36,7 @@ class RedBearEmbeddings(Embeddings):
                 "base_url": config.base_url,
                 "api_key": config.api_key,
                 "timeout": httpx.Timeout(timeout=config.timeout, connect=60.0),
-                "max_retries": config.max_retries,
-                "check_embedding_ctx_length": False,
-                "encoding_format": "float"
+                "max_retries": config.max_retries
             }
         elif provider == ModelProvider.DASHSCOPE:
             params = {

api/app/core/models/rerank.py

@@ -76,5 +76,9 @@ class RedBearRerank(BaseDocumentCompressor):
             from langchain_community.document_compressors import JinaRerank
             model_instance: JinaRerank = self._model
             return model_instance.rerank(documents=documents, query=query, top_n=top_n)
+        elif provider == ModelProvider.DASHSCOPE:
+            from langchain_community.document_compressors.dashscope_rerank import DashScopeRerank
+            model_instance: DashScopeRerank = self._model
+            return model_instance.rerank(documents=documents, query=query, top_n=top_n)
         else:
             raise ValueError(f"不支持的模型提供商: {provider}")

api/app/core/rag/app/naive.py

@@ -672,10 +672,15 @@ def chunk(filename, binary=None, from_page=0, to_page=100000,
         excel_parser = ExcelParser()
         if parser_config.get("html4excel") and parser_config.get("html4excel").lower() == "true":
             sections = [(_, "") for _ in excel_parser.html(binary, 12) if _]
-            parser_config["chunk_token_num"] = 0
         else:
             sections = [(_, "") for _ in excel_parser(binary) if _]
-        parser_config["chunk_token_num"] = 12800
+        callback(0.8, "Finish parsing.")
+        # Excel 每行直接作为一个 chunk，不经过 naive_merge 避免被 delimiter 拆分
+        chunks = [s for s, _ in sections]
+        res.extend(tokenize_chunks(chunks, doc, is_english, None))
+        res.extend(embed_res)
+        res.extend(url_res)
+        return res
 
     elif re.search(r"\.(txt|py|js|java|c|cpp|h|php|go|ts|sh|cs|kt|sql)$", filename, re.IGNORECASE):
         callback(0.1, "Start to parse.")

api/app/core/rag/deepdoc/parser/excel_parser.py

@@ -232,14 +232,14 @@ class RAGExcelParser:
                         t = str(ti[i].value) if i < len(ti) else ""
                         t += ("：" if t else "") + str(c.value)
                         fields.append(t)
-                    line = "; ".join(fields)
+                    line = "\n".join(fields)
                     if sheetname.lower().find("sheet") < 0:
-                        line += " ——" + sheetname
+                        line += "\n——" + sheetname
                     res.append(line)
             else:
                 # 只有表头的情况
                 if header_fields:
-                    line = "; ".join(header_fields)
+                    line = "\n".join(header_fields)
                     if sheetname.lower().find("sheet") < 0:
                         line += " ——" + sheetname
                     res.append(line)

api/app/core/rag/deepdoc/parser/mineru_parser.py

@@ -292,9 +292,10 @@ class MinerUParser(RAGPdfParser):
         self.page_from = page_from
         self.page_to = page_to
         try:
-            with pdfplumber.open(fnm) if isinstance(fnm, (str, PathLike)) else pdfplumber.open(BytesIO(fnm)) as pdf:
-                self.pdf = pdf
-                self.page_images = [p.to_image(resolution=72 * zoomin, antialias=True).original for _, p in enumerate(self.pdf.pages[page_from:page_to])]
+            with sys.modules[LOCK_KEY_pdfplumber]:  # ← 加这一行，获取全局锁
+                with pdfplumber.open(fnm) if isinstance(fnm, (str, PathLike)) else pdfplumber.open(BytesIO(fnm)) as pdf:
+                    self.pdf = pdf
+                    self.page_images = [p.to_image(resolution=72 * zoomin, antialias=True).original for _, p in enumerate(self.pdf.pages[page_from:page_to])]
         except Exception as e:
             self.page_images = None
             self.total_page = 0

api/app/core/rag/llm/embedding_model.py

@@ -50,7 +50,9 @@ class OpenAIEmbed(Base):
     def encode(self, texts: list):
         # OpenAI requires batch size <=16
         batch_size = 16
-        texts = [truncate(t, 8191) for t in texts]
+        # Use 8000 instead of 8191 to leave safety margin for tokenizer differences
+        # between cl100k_base (used by truncate) and the actual embedding model
+        texts = [truncate(t, 8000) for t in texts]
         ress = []
         total_tokens = 0
         for i in range(0, len(texts), batch_size):
@@ -63,7 +65,7 @@ class OpenAIEmbed(Base):
         return np.array(ress), total_tokens
 
     def encode_queries(self, text):
-        res = self.client.embeddings.create(input=[truncate(text, 8191)], model=self.model_name, encoding_format="float",extra_body={"drop_params": True})
+        res = self.client.embeddings.create(input=[truncate(text, 8000)], model=self.model_name, encoding_format="float",extra_body={"drop_params": True})
         return np.array(res.data[0].embedding), self.total_token_count(res)
 
 
@@ -79,6 +81,7 @@ class LocalAIEmbed(Base):
 
     def encode(self, texts: list):
         batch_size = 16
+        texts = [truncate(t, 8000) for t in texts]
         ress = []
         for i in range(0, len(texts), batch_size):
             res = self.client.embeddings.create(input=texts[i : i + batch_size], model=self.model_name)
@@ -173,6 +176,7 @@ class XinferenceEmbed(Base):
 
     def encode(self, texts: list):
         batch_size = 16
+        texts = [truncate(t, 8000) for t in texts]
         ress = []
         total_tokens = 0
         for i in range(0, len(texts), batch_size):
@@ -188,7 +192,7 @@ class XinferenceEmbed(Base):
     def encode_queries(self, text):
         res = None
         try:
-            res = self.client.embeddings.create(input=[text], model=self.model_name)
+            res = self.client.embeddings.create(input=[truncate(text, 8000)], model=self.model_name)
             return np.array(res.data[0].embedding), self.total_token_count(res)
         except Exception as _e:
             log_exception(_e, res)

api/app/core/rag/nlp/search.py

@@ -28,6 +28,7 @@ from app.core.rag.common.float_utils import get_float
 from app.core.rag.common.constants import PAGERANK_FLD, TAG_FLD
 from app.core.rag.llm.chat_model import Base
 from app.core.rag.llm.embedding_model import OpenAIEmbed
+from app.services.model_service import ModelApiKeyService
 import logging
 
 logger = logging.getLogger(__name__)
@@ -114,9 +115,8 @@ def knowledge_retrieval(
         # Use the specified reranker for re-ranking
         if reranker_id:
             try:
-                return rerank(db=db, reranker_id=reranker_id, query=query, docs=all_results, top_k=reranker_top_k)
+                all_results = rerank(db=db, reranker_id=reranker_id, query=query, docs=all_results, top_k=reranker_top_k)
             except Exception as rerank_error:
-                # If reranker fails, log warning and continue with original results
                 logger.warning(
                     "Reranker failed, falling back to original results",
                     extra={
@@ -132,7 +132,10 @@ def knowledge_retrieval(
                 from app.core.rag.common.settings import kg_retriever
                 doc = kg_retriever.retrieval(question=query, workspace_ids=workspace_ids, kb_ids=kb_ids, emb_mdl=embedding_model, llm=chat_model)
                 if doc:
-                    all_results.insert(0, doc)
+                    all_results.insert(0, DocumentChunk(
+                        page_content=doc.get("page_content", ""),
+                        metadata=doc.get("metadata", {})
+                    ))
             except Exception as graph_error:
                 print(f"Failed to retrieve from knowledge graph: {str(graph_error)}")
         
@@ -198,16 +201,18 @@ def _retrieve_for_knowledge(
         workspace_ids.append(str(db_knowledge.workspace_id))
 
     if not chat_model:
+        llm_key = ModelApiKeyService.get_available_api_key(db, db_knowledge.llm_id)
         chat_model = Base(
-            key=db_knowledge.llm.api_keys[0].api_key,
-            model_name=db_knowledge.llm.api_keys[0].model_name,
-            base_url=db_knowledge.llm.api_keys[0].api_base,
+            key=llm_key.api_key,
+            model_name=llm_key.model_name,
+            base_url=llm_key.api_base,
         )
     if not embedding_model:
+        emb_key = ModelApiKeyService.get_available_api_key(db, db_knowledge.embedding_id)
         embedding_model = OpenAIEmbed(
-            key=db_knowledge.embedding.api_keys[0].api_key,
-            model_name=db_knowledge.embedding.api_keys[0].model_name,
-            base_url=db_knowledge.embedding.api_keys[0].api_base,
+            key=emb_key.api_key,
+            model_name=emb_key.model_name,
+            base_url=emb_key.api_base,
         )
 
     vector_service = ElasticSearchVectorFactory().init_vector(knowledge=db_knowledge)
@@ -248,6 +253,29 @@ def _retrieve_for_knowledge(
                     seen_ids.add(doc.metadata["doc_id"])
                     unique_rs.append(doc)
             rs = unique_rs
+            if unique_rs:
+                rs = vector_service.rerank(
+                    query=kb_config["query"],
+                    docs=unique_rs,
+                    top_k=kb_config["top_k"]
+                )
+            if kb_config["retrieve_type"] == "graph":
+                try:
+                    from app.core.rag.common.settings import kg_retriever
+                    graph_doc = kg_retriever.retrieval(
+                        question=kb_config["query"],
+                        workspace_ids=[str(db_knowledge.workspace_id)],
+                        kb_ids=[str(db_knowledge.id)],
+                        emb_mdl=embedding_model,
+                        llm=chat_model,
+                    )
+                    if graph_doc:
+                        rs.insert(0, DocumentChunk(
+                            page_content=graph_doc.get("page_content", ""),
+                            metadata=graph_doc.get("metadata", {})
+                        ))
+                except Exception as graph_error:
+                    logger.warning(f"Graph retrieval failed for kb {db_knowledge.id}: {graph_error}")
 
     results.extend(rs)
     return results, chat_model, embedding_model

api/app/core/tools/builtin/datetime_tool.py

@@ -230,7 +230,7 @@ class DateTimeTool(BuiltinTool):
     @staticmethod
     def _datetime_to_timestamp(kwargs) -> dict:
         """日期时间转时间戳"""
-        input_value = kwargs.get("input_value")
+        input_value = kwargs.get("input_value").strip()
         input_format = kwargs.get("input_format", "%Y-%m-%d %H:%M:%S")
         timezone_str = kwargs.get("from_timezone", "Asia/Shanghai")
         
@@ -253,9 +253,9 @@ class DateTimeTool(BuiltinTool):
         return {
             "datetime": input_value,
             "timezone": timezone_str,
-            "timestamp": int(dt.timestamp()),
+            "timestamp": int(dt.timestamp()) * 1000,
             "iso_format": dt.isoformat(),
-            "result_data": int(dt.timestamp())
+            "result_data": int(dt.timestamp()) * 1000
         }
 
     def _calculate_datetime(self, kwargs) -> dict:

api/app/core/tools/builtin/openclaw_tool.py

@@ -0,0 +1,300 @@
+"""OpenClaw 远程 Agent 内置工具"""
+import time
+import base64
+from io import BytesIO
+from typing import List, Dict, Any, Optional
+import aiohttp
+
+from app.core.tools.builtin.base import BuiltinTool
+from app.schemas.tool_schema import ToolParameter, ToolResult, ParameterType
+from app.core.logging_config import get_business_logger
+
+logger = get_business_logger()
+
+
+class OpenClawTool(BuiltinTool):
+    """OpenClaw 远程 Agent 工具 — 支持文本和图片多模态输入"""
+
+    def __init__(self, tool_id: str, config: Dict[str, Any]):
+        super().__init__(tool_id, config)
+        params = self.parameters_config
+
+        # 用户配置项（前端表单填写）
+        self._server_url = params.get("server_url", "")
+        self._api_key = params.get("api_key", "")
+        self._agent_id = params.get("agent_id", "main")
+
+        # 内部默认值
+        self._model = "openclaw"
+        self._session_strategy = "by_user"
+        self._timeout = 120
+
+        # 运行时上下文（通过 set_runtime_context 注入）
+        self._user_id = "anonymous"
+        self._conversation_id = None
+        self._uploaded_files = []
+
+    @property
+    def name(self) -> str:
+        return "openclaw_tool"
+
+    @property
+    def description(self) -> str:
+        return (
+            "OpenClaw 远程 Agent：将任务委托给远程 OpenClaw Agent。"
+            "具备 3D 模型生成与打印控制、设备管理、文件处理、浏览器自动化、"
+            "Shell 命令执行、网络搜索等能力。支持文本和图片多模态交互。"
+        )
+
+    def get_required_config_parameters(self) -> List[str]:
+        return ["server_url", "api_key"]
+        
+    @property
+    def parameters(self) -> List[ToolParameter]:
+        return [
+            ToolParameter(
+                name="operation",
+                type=ParameterType.STRING,
+                description="任务类型",
+                required=True,
+                enum= ["print_task", "device_query", "image_understand", "general"]
+            ),
+            ToolParameter(
+                name="message",
+                type=ParameterType.STRING,
+                description="发送给 OpenClaw Agent 的文本请求内容",
+                required=True
+            ),
+            ToolParameter(
+                name="image_url",
+                type=ParameterType.STRING,
+                description="可选，附带的图片 URL 或 base64 data URI（OpenClaw 支持图片输入）",
+                required=False
+            )
+        ]       
+            
+    # ---------- 运行时上下文注入 ----------
+    def set_runtime_context(
+        self,
+        user_id: str = "anonymous",
+        conversation_id: Optional[str] = None,
+        uploaded_files: Optional[list] = None
+    ):
+        """注入运行时上下文（由 chat service 调用）"""
+        self._user_id = user_id
+        self._conversation_id = conversation_id
+        self._uploaded_files = uploaded_files or []
+
+    # ---------- 连接测试 ----------
+    async def test_connection(self) -> Dict[str, Any]:
+        """测试 OpenClaw Gateway 连接"""
+        if not self._server_url:
+            return {"success": False, "message": "未配置 server_url"}
+        if not self._api_key:
+            return {"success": False, "message": "未配置 api_key"}
+
+        url = f"{self._server_url.rstrip('/')}/v1/responses"
+        headers = {
+            "Authorization": f"Bearer {self._api_key}",
+            "Content-Type": "application/json",
+            "x-openclaw-agent-id": self._agent_id
+        }
+        body = {
+            "model": self._model,
+            "user": "connection-test",
+            "input": "hi",
+            "stream": False
+        }
+        try:
+            timeout_cfg = aiohttp.ClientTimeout(total=30)
+            async with aiohttp.ClientSession(timeout=timeout_cfg) as session:
+                async with session.post(url, json=body, headers=headers) as resp:
+                    if resp.status < 400:
+                        return {"success": True, "message": "OpenClaw 连接成功"}
+                    error_text = await resp.text()
+                    return {
+                        "success": False,
+                        "message": f"OpenClaw HTTP {resp.status}: {error_text[:200]}"
+                    }
+        except Exception as e:
+            return {"success": False, "message": f"OpenClaw 连接失败: {str(e)}"}
+
+    # ---------- 执行 ----------
+    async def execute(self, **kwargs) -> ToolResult:
+        """执行 OpenClaw 调用"""
+        start_time = time.time()
+        try:
+            message = kwargs.get("message", "")
+            if not message:
+                return ToolResult.error_result(
+                    error="message 参数不能为空",
+                    error_code="OPENCLAW_INVALID_INPUT",
+                    execution_time=time.time() - start_time
+                )
+
+            # 提取图片：优先从用户上传文件中获取，LLM 传的 image_url 作为兜底
+            image_url = self._extract_image_from_uploads()
+            if not image_url:
+                image_url = kwargs.get("image_url")
+            if image_url and not image_url.startswith("data:"):
+                image_url = await self._download_and_encode_image(image_url)
+
+            # 构建请求
+            url = f"{self._server_url.rstrip('/')}/v1/responses"
+            headers = {
+                "Authorization": f"Bearer {self._api_key}",
+                "Content-Type": "application/json",
+                "x-openclaw-agent-id": self._agent_id
+            }
+            user_field = (
+                f"conv-{self._conversation_id}"
+                if self._session_strategy == "by_conversation" and self._conversation_id
+                else f"user-{self._user_id}"
+            )
+            input_field = self._build_input(message, image_url)
+            body = {
+                "model": self._model,
+                "user": user_field,
+                "input": input_field,
+                "stream": False
+            }
+
+            timeout_cfg = aiohttp.ClientTimeout(total=self._timeout)
+            # 打印请求日志（截断 base64 避免日志过大）
+            log_body = {**body}
+            if isinstance(log_body.get("input"), list):
+                log_body["input"] = "[multimodal input, truncated]"
+            elif isinstance(log_body.get("input"), str) and len(log_body["input"]) > 500:
+                log_body["input"] = log_body["input"][:500] + "..."
+            logger.info(
+                f"OpenClaw 请求: url={url}, agent_id={self._agent_id}, "
+                f"has_image={bool(image_url)}, body={log_body}"
+            )
+            async with aiohttp.ClientSession(timeout=timeout_cfg) as session:
+                async with session.post(url, json=body, headers=headers) as resp:
+                    execution_time = time.time() - start_time
+                    if resp.status >= 400:
+                        error_text = await resp.text()
+                        return ToolResult.error_result(
+                            error=f"OpenClaw HTTP {resp.status}: {error_text[:500]}",
+                            error_code="OPENCLAW_HTTP_ERROR",
+                            execution_time=execution_time
+                        )
+                    data = await resp.json()
+                    text = self._extract_response(data)
+                    display_text = self._format_result(text)
+                    return ToolResult.success_result(
+                        data=display_text,
+                        execution_time=execution_time
+                    )
+
+        except aiohttp.ClientError as e:
+            return ToolResult.error_result(
+                error=f"OpenClaw 网络连接失败: {str(e)}",
+                error_code="OPENCLAW_NETWORK_ERROR",
+                execution_time=time.time() - start_time
+            )
+        except Exception as e:
+            return ToolResult.error_result(
+                error=f"OpenClaw 调用失败: {str(e)}",
+                error_code="OPENCLAW_EXECUTION_ERROR",
+                execution_time=time.time() - start_time
+            )
+
+    # ---------- 私有方法 ----------
+    def _extract_image_from_uploads(self) -> Optional[str]:
+        """从用户上传文件中提取图片 URL"""
+        for f in self._uploaded_files:
+            f_type = f.get("type", "")
+            if f_type == "image":
+                source = f.get("source", {})
+                if source.get("type") == "base64":
+                    media_type = source.get("media_type", "image/jpeg")
+                    data = source.get("data", "")
+                    return f"data:{media_type};base64,{data}"
+                elif f.get("image"):
+                    return f.get("image")
+                elif f.get("url"):
+                    return f.get("url")
+            elif f_type == "image_url":
+                return f.get("image_url", {}).get("url", "")
+        return None
+
+    async def _download_and_encode_image(self, image_url: str) -> str:
+        """下载图片并转为 base64 data URI"""
+        try:
+            from PIL import Image
+            MAX_RAW_SIZE = 4 * 1024 * 1024
+
+            async with aiohttp.ClientSession() as session:
+                async with session.get(
+                    image_url, allow_redirects=True,
+                    timeout=aiohttp.ClientTimeout(total=30)
+                ) as resp:
+                    if resp.status != 200:
+                        return image_url
+                    content_type = resp.headers.get("Content-Type", "image/jpeg")
+                    if not content_type.startswith("image/"):
+                        return image_url
+                    img_bytes = await resp.read()
+
+                    if len(img_bytes) > MAX_RAW_SIZE:
+                        img = Image.open(BytesIO(img_bytes))
+                        if img.mode in ("RGBA", "P", "LA"):
+                            img = img.convert("RGB")
+                        if max(img.size) > 2048:
+                            img.thumbnail((2048, 2048), Image.LANCZOS)
+                        buf = BytesIO()
+                        img.save(buf, format="JPEG", quality=75, optimize=True)
+                        img_bytes = buf.getvalue()
+                        content_type = "image/jpeg"
+
+                    b64 = base64.b64encode(img_bytes).decode("utf-8")
+                    return f"data:{content_type};base64,{b64}"
+        except Exception as e:
+            logger.warning(f"OpenClaw 下载图片失败，使用原始 URL: {e}")
+            return image_url
+
+    def _build_input(self, message: str, image_url: Optional[str] = None):
+        """构造请求 input 字段：有图片则构造多模态结构，否则纯文本"""
+        if not image_url:
+            return message
+
+        content_parts = [{"type": "input_text", "text": message}]
+        if image_url.startswith("data:"):
+            try:
+                header, data = image_url.split(",", 1)
+                media_type = header.split(":")[1].split(";")[0]
+                content_parts.append({
+                    "type": "input_image",
+                    "source": {"type": "base64", "media_type": media_type, "data": data}
+                })
+            except (ValueError, IndexError):
+                return message
+        else:
+            content_parts.append({
+                "type": "input_image",
+                "source": {"type": "url", "url": image_url}
+            })
+
+        return [{"type": "message", "role": "user", "content": content_parts}]
+
+    def _extract_response(self, response_data: Dict[str, Any]) -> str:
+        """从 OpenClaw 响应中提取文本内容
+
+        OpenClaw /v1/responses 只返回 output_text 类型的内容。
+        图片信息（如有）由 OpenClaw Skill 以 Markdown 链接形式嵌入文本中返回。
+        """
+        output = response_data.get("output", [])
+        texts = []
+        for item in output:
+            if item.get("type") == "message":
+                for content in item.get("content", []):
+                    if content.get("type") == "output_text" and content.get("text"):
+                        texts.append(content["text"])
+        return "\n".join(texts) if texts else str(response_data)
+
+    @staticmethod
+    def _format_result(text: str) -> str:
+        """格式化结果为 LLM 可读字符串"""
+        return text or "（OpenClaw 返回了空内容）"

api/app/core/tools/builtin/operation_tool.py

@@ -11,6 +11,11 @@ class OperationTool(BaseTool):
         self.base_tool = base_tool
         self.operation = operation
         super().__init__(base_tool.tool_id, base_tool.config)
+
+    def set_runtime_context(self, **kwargs):
+        """转发运行时上下文到 base_tool"""
+        if hasattr(self.base_tool, 'set_runtime_context'):
+            self.base_tool.set_runtime_context(**kwargs)
     
     @property
     def name(self) -> str:
@@ -32,6 +37,8 @@ class OperationTool(BaseTool):
             return self._get_datetime_params()
         elif self.base_tool.name == 'json_tool':
             return self._get_json_params()
+        elif self.base_tool.name == 'openclaw_tool':
+            return self._get_openclaw_params()
         else:
             # 默认返回除operation外的所有参数
             return [p for p in self.base_tool.parameters if p.name != "operation"]
@@ -138,6 +145,29 @@ class OperationTool(BaseTool):
                     default="Asia/Shanghai"
                 )
             ]
+        elif self.operation == "datetime_to_timestamp":
+            return [
+                ToolParameter(
+                    name="input_value",
+                    type=ParameterType.STRING,
+                    description="输入值（时间字符串，如：2026-04-07 10:30:25）",
+                    required=True
+                ),
+                ToolParameter(
+                    name="input_format",
+                    type=ParameterType.STRING,
+                    description="输入时间格式（如：%Y-%m-%d %H:%M:%S）",
+                    required=False,
+                    default="%Y-%m-%d %H:%M:%S"
+                ),
+                ToolParameter(
+                    name="from_timezone",
+                    type=ParameterType.STRING,
+                    description="源时区（如：UTC, Asia/Shanghai）",
+                    required=False,
+                    default="Asia/Shanghai"
+                )
+            ]
         else:
             return []
     
@@ -209,6 +239,64 @@ class OperationTool(BaseTool):
         else:
             return base_params
     
+    def _get_openclaw_params(self) -> List[ToolParameter]:
+        """获取 openclaw_tool 特定操作的参数"""
+        if self.operation == "print_task":
+            return [
+                ToolParameter(
+                    name="message",
+                    type=ParameterType.STRING,
+                    description="发送给 OpenClaw 的打印任务描述，将用户的原始消息原封不动地传递给 OpenClaw，禁止改写、补充或润色用户的原文",
+                    required=True
+                ),
+                ToolParameter(
+                    name="image_url",
+                    type=ParameterType.STRING,
+                    description="可选，附带的设计图片或参考图，OpenClaw 可据此生成 3D 模型",
+                    required=False
+                )
+            ]
+        elif self.operation == "device_query":
+            return [
+                ToolParameter(
+                    name="message",
+                    type=ParameterType.STRING,
+                    description="发送给 OpenClaw 的设备查询指令",
+                    required=True
+                )
+            ]
+        elif self.operation == "image_understand":
+            return [
+                ToolParameter(
+                    name="message",
+                    type=ParameterType.STRING,
+                    description="发送给 OpenClaw 的图片理解任务，应描述需要对图片做什么（如描述内容、提取文字、分析信息）",
+                    required=True
+                ),
+                ToolParameter(
+                    name="image_url",
+                    type=ParameterType.STRING,
+                    description="要分析的图片 URL 或 base64 data URI",
+                    required=False
+                )
+            ]
+        else:
+            # general 及其他
+            return [
+                ToolParameter(
+                    name="message",
+                    type=ParameterType.STRING,
+                    description="发送给 OpenClaw Agent 的任务描述，应包含完整的任务需求",
+                    required=True
+                ),
+                ToolParameter(
+                    name="image_url",
+                    type=ParameterType.STRING,
+                    description="可选，附带的图片 URL 或 base64 data URI",
+                    required=False
+                )
+            ]
+
     async def execute(self, **kwargs) -> ToolResult:
         """执行特定操作"""
         # 添加operation参数

api/app/core/tools/configs/builtin/openclaw_tool.json

@@ -0,0 +1,15 @@
+{
+  "name": "openclaw_tool",
+  "description": "调用OpenClaw Agent远程服务",
+  "tool_class": "OpenClawTool",
+  "category": "agent",
+  "requires_config": true,
+  "version": "1.0.0",
+  "enabled": true,
+  "parameters": {
+    "server_url": "",
+    "api_key": "",
+    "agent_id": "main"
+  },
+  "tags": ["agent", "openclaw", "multimodal", "3d-printing", "builtin"]
+}

api/app/core/tools/configs/builtin_tools.json

@@ -30,5 +30,18 @@
     "parameters": {
       "api_key": {"type": "string", "description": "百度搜索API密钥", "sensitive": true, "required": true}
     }
+  },
+  "openclaw": {
+    "name": "OpenClaw远程Agent",
+    "description": "OpenClaw Agent远程服务",
+    "tool_class": "OpenClawTool",
+    "category": "agent",
+    "requires_config": true,
+    "version": "1.0.0",
+    "enabled": true,
+    "parameters": {
+      "server_url": {"type": "string", "description": "OpenClaw Gateway 地址", "required": true},
+      "api_key": {"type": "string", "description": "OpenClaw API Key", "sensitive": true, "required": true}
+    }
   }
 }

api/app/core/tools/custom/base.py

@@ -30,7 +30,7 @@ class CustomTool(BaseTool):
         self.auth_config = config.get("auth_config", {})
         self.base_url = config.get("base_url", "")
         self.timeout = config.get("timeout", 30)
-        
+
         # 解析schema
         self._parsed_operations = self._parse_openapi_schema()
     

api/app/core/tools/langchain_adapter.py

@@ -131,7 +131,7 @@ class LangchainAdapter:
     def _tool_supports_operations(tool: BaseTool) -> bool:
         """检查工具是否支持多操作"""
         # 内置工具中支持操作的工具
-        builtin_operation_tools = ['datetime_tool', 'json_tool']
+        builtin_operation_tools = ['datetime_tool', 'json_tool', 'openclaw_tool']
         
         # 检查内置工具
         if tool.tool_type.value == "builtin" and tool.name in builtin_operation_tools:

api/app/core/workflow/adapters/base_adapter.py

@@ -40,6 +40,7 @@ class WorkflowParserResult(BaseModel):
     edges: list[EdgeDefinition] = Field(default_factory=list)
     nodes: list[NodeDefinition] = Field(default_factory=list)
     variables: list[VariableDefinition] = Field(default_factory=list)
+    features: dict[str, Any] = Field(default_factory=dict)
     warnings: list[ExceptionDefinition] = Field(default_factory=list)
     errors: list[ExceptionDefinition] = Field(default_factory=list)
 
@@ -51,6 +52,7 @@ class WorkflowImportResult(BaseModel):
     edges: list[EdgeDefinition] = Field(default_factory=list)
     nodes: list[NodeDefinition] = Field(default_factory=list)
     variables: list[VariableDefinition] = Field(default_factory=list)
+    features: dict[str, Any] = Field(default_factory=dict)
     warnings: list[ExceptionDefinition] = Field(default_factory=list)
     errors: list[ExceptionDefinition] = Field(default_factory=list)
 

api/app/core/workflow/adapters/dify/converter.py

@@ -15,7 +15,7 @@ from app.core.workflow.adapters.errors import (
     ExceptionType
 )
 from app.core.workflow.nodes.assigner.config import AssignmentItem
-from app.core.workflow.nodes.base_config import VariableDefinition, BaseNodeConfig
+from app.core.workflow.nodes.base_config import VariableDefinition as NodeVariableDefinition, BaseNodeConfig
 from app.core.workflow.nodes.code.config import InputVariable, OutputVariable
 from app.core.workflow.nodes.configs import (
     StartNodeConfig,
@@ -36,6 +36,7 @@ from app.core.workflow.nodes.configs import (
     ListOperatorNodeConfig,
     DocExtractorNodeConfig,
 )
+from app.schemas.workflow_schema import VariableDefinition as SchemaVariableDefinition
 from app.core.workflow.nodes.cycle_graph.config import (
     ConditionDetail as LoopConditionDetail,
     ConditionsConfig,
@@ -98,6 +99,7 @@ class DifyConverter(BaseConverter):
             NodeType.CYCLE_START: lambda x: {},
             NodeType.BREAK: lambda x: {},
         }
+        self._file_vars_to_conv: list[SchemaVariableDefinition] = []
 
     def get_node_convert(self, node_type):
         func = self.CONFIG_CONVERT_MAP.get(node_type, lambda x: {})
@@ -286,19 +288,25 @@ class DifyConverter(BaseConverter):
                 )
                 continue
 
-            if var_type in ["file", "array[file]"]:
-                self.errors.append(
-                    ExceptionDefinition(
-                        type=ExceptionType.VARIABLE,
-                        node_id=node["id"],
-                        node_name=node_data["title"],
-                        name=var["variable"],
-                        detail=f"Unsupported Variable type for start node: {var_type}"
-                    )
-                )
+            if var_type in [VariableType.FILE, VariableType.ARRAY_FILE]:
+                # 开始节点不支持文件变量，转为会话变量
+                self._file_vars_to_conv.append(SchemaVariableDefinition(
+                    name=var["variable"],
+                    type=var_type.value,
+                    required=var.get("required", False),
+                    default=None,
+                    description=var.get("label", ""),
+                ))
+                self.warnings.append(ExceptionDefinition(
+                    type=ExceptionType.VARIABLE,
+                    node_id=node["id"],
+                    node_name=node_data["title"],
+                    name=var["variable"],
+                    detail=f"File variable '{var['variable']}' is not supported in start node, moved to conversation variables"
+                ))
                 continue
 
-            var_def = VariableDefinition(
+            var_def = NodeVariableDefinition(
                 name=var["variable"],
                 type=var_type,
                 required=var["required"],
@@ -837,3 +845,76 @@ class DifyConverter(BaseConverter):
         ).model_dump()
         self.config_validate(node["id"], node["data"]["title"], DocExtractorNodeConfig, result)
         return result
+
+    @staticmethod
+    def convert_features(features: dict) -> dict:
+        """Convert Dify features to MemoryBear FeaturesConfigForm format."""
+        if not features:
+            return {}
+
+        result: dict = {}
+
+        # opening_statement
+        opening = features.get("opening_statement", "")
+        suggested = features.get("suggested_questions", [])
+        result["opening_statement"] = {
+            "enabled": bool(opening),
+            "statement": opening or None,
+            "suggested_questions": suggested,
+        }
+
+        # citation (对应 Dify retriever_resource)
+        retriever = features.get("retriever_resource", {})
+        result["citation"] = {
+            "enabled": retriever.get("enabled", False) if isinstance(retriever, dict) else False,
+        }
+
+        # file_upload: Dify allowed_file_types 数组 -> 前端扁平字段
+        file_upload = features.get("file_upload", {})
+        allowed_types = file_upload.get("allowed_file_types", []) if file_upload else []
+        allowed_methods = file_upload.get("allowed_file_upload_methods", ["local_file", "remote_url"])
+        if isinstance(allowed_methods, list):
+            if len(allowed_methods) >= 2:
+                transfer_method = "both"
+            elif allowed_methods:
+                transfer_method = allowed_methods[0]
+            else:
+                transfer_method = "both"
+        else:
+            transfer_method = allowed_methods or "both"
+
+        file_config = file_upload.get("fileUploadConfig", {})
+        result["file_upload"] = {
+            "enabled": file_upload.get("enabled", False) if file_upload else False,
+            "image_enabled": "image" in allowed_types,
+            "image_max_size_mb": file_config.get("image_file_size_limit", 10) if file_config else 10,
+            "image_allowed_extensions": ["png", "jpg", "jpeg"],
+            "audio_enabled": "audio" in allowed_types,
+            "audio_max_size_mb": file_config.get("audio_file_size_limit", 50) if file_config else 50,
+            "audio_allowed_extensions": ["mp3", "wav", "m4a"],
+            "document_enabled": "document" in allowed_types,
+            "document_max_size_mb": file_config.get("file_size_limit", 100) if file_config else 100,
+            "document_allowed_extensions": ["pdf", "docx", "doc", "xlsx", "xls", "txt", "csv", "json", "md"],
+            "video_enabled": "video" in allowed_types,
+            "video_max_size_mb": file_config.get("video_file_size_limit", 100) if file_config else 100,
+            "video_allowed_extensions": ["mp4", "mov"],
+            "max_file_count": file_upload.get("number_limits", 1) if file_upload else 1,
+            "allowed_transfer_methods": transfer_method,
+        }
+
+        # text_to_speech
+        tts = features.get("text_to_speech", {})
+        result["text_to_speech"] = {
+            "enabled": tts.get("enabled", False) if isinstance(tts, dict) else False,
+            "voice": tts.get("voice") if isinstance(tts, dict) else None,
+            "language": tts.get("language") if isinstance(tts, dict) else None,
+            "autoplay": False,
+        }
+
+        # suggested_questions_after_answer
+        sqa = features.get("suggested_questions_after_answer", {})
+        result["suggested_questions_after_answer"] = {
+            "enabled": sqa.get("enabled", False) if isinstance(sqa, dict) else False,
+        }
+
+        return result

api/app/core/workflow/adapters/dify/dify_adapter.py

@@ -119,9 +119,12 @@ class DifyAdapter(BasePlatformAdapter, DifyConverter):
             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)
+        # 开始节点的文件变量合并到会话变量
+        self.conv_variables.extend(self._file_vars_to_conv)
+
+        features = self.convert_features(
+            self.config.get("workflow", {}).get("features", {})
+        )
 
         trigger = self._convert_trigger({})
         execution_config = self._convert_execution({})
@@ -135,6 +138,7 @@ class DifyAdapter(BasePlatformAdapter, DifyConverter):
             edges=self.edges,
             nodes=self.nodes,
             variables=self.conv_variables,
+            features=features,
             warnings=self.warnings,
             errors=self.errors
         )

api/app/core/workflow/engine/graph_builder.py

@@ -31,9 +31,9 @@ logger = logging.getLogger(__name__)
 # Example:
 #   "Hello {{user.name}}!" ->
 #   ["Hello ", "{{user.name}}", "!"]
-_OUTPUT_PATTERN = re.compile(r'\{\{.*?}}|[^{}]+')
+_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*}}')
+_VARIABLE_PATTERN = re.compile(r'\{\{\s*[a-zA-Z0-9_]+\.[a-zA-Z0-9_]+(?:\.[a-zA-Z0-9_]+)?\s*}}')
 
 
 class GraphBuilder:

api/app/core/workflow/engine/stream_output_coordinator.py

@@ -14,7 +14,7 @@ from app.core.workflow.engine.variable_pool import VariablePool
 logger = get_logger(__name__)
 
 SCOPE_PATTERN = re.compile(
-    r"\{\{\s*([a-zA-Z0-9_]+)\.[a-zA-Z0-9_]+\s*}}"
+    r"\{\{\s*([a-zA-Z0-9_]+)\.[a-zA-Z0-9_]+(?:\.[a-zA-Z0-9_]+)?\s*}}"
 )
 
 

api/app/core/workflow/engine/variable_pool.py

@@ -34,19 +34,22 @@ class LazyVariableDict:
             return self._cache[key]
         var_struct = self._source.get(key)
         if var_struct is None:
-            raise KeyError(key)
-        value = var_struct.instance.to_literal() if self._literal else var_struct.instance.get_value()
+            return None
+        raw = var_struct.instance.get_value()
+        # literal 模式下 dict/list 保留结构，让 Jinja2 能继续访问子字段（如 .type）
+        value = raw if (not self._literal or isinstance(raw, (dict, list))) else var_struct.instance.to_literal()
         self._cache[key] = value
         return value
 
     def get(self, key, default=None):
-        try:
-            return self._resolve(key)
-        except KeyError:
-            return default
+        value = self._resolve(key)
+        return default if value is None else value
 
     def __getitem__(self, key):
-        return self._resolve(key)
+        value = self._resolve(key)
+        if value is None:
+            raise KeyError(key)
+        return value
 
     def __getattr__(self, key):
         if key.startswith('_'):
@@ -164,7 +167,7 @@ class VariablePool:
     def transform_selector(selector):
         variable_literal = VARIABLE_PATTERN.sub(r"\1", selector).strip()
         selector = VariableSelector.from_string(variable_literal).path
-        if len(selector) != 2:
+        if len(selector) not in (2, 3):
             raise ValueError(f"Selector not valid - {selector}")
         return selector
 
@@ -196,6 +199,16 @@ class VariablePool:
             return None
         return var_instance
 
+    @staticmethod
+    def _extract_field(struct: "VariableStruct", field: str | None) -> Any:
+        """If field is given, drill into a dict/object variable's value."""
+        if field is None:
+            return struct.instance.get_value()
+        value = struct.instance.get_value()
+        if not isinstance(value, dict):
+            raise KeyError(f"Variable is not an object, cannot access field '{field}'")
+        return value.get(field)
+
     def get_instance(
             self,
             selector: str,
@@ -250,12 +263,14 @@ class VariablePool:
         Raises:
             KeyError: If strict is True and the variable does not exist.
         """
+        path = self.transform_selector(selector)
         variable_struct = self._get_variable_struct(selector)
         if variable_struct is None:
             if strict:
                 raise KeyError(f"{selector} not exist")
             return default
-
+        if len(path) == 3:
+            return self._extract_field(variable_struct, path[2])
         return variable_struct.instance.get_value()
 
     def get_literal(
@@ -282,12 +297,15 @@ class VariablePool:
         Raises:
             KeyError: If strict is True and the variable does not exist.
         """
+        path = self.transform_selector(selector)
         variable_struct = self._get_variable_struct(selector)
         if variable_struct is None:
             if strict:
                 raise KeyError(f"{selector} not exist")
             return default
-
+        if len(path) == 3:
+            value = self._extract_field(variable_struct, path[2])
+            return str(value) if value is not None else ""
         return variable_struct.instance.to_literal()
 
     async def set(
@@ -345,7 +363,14 @@ class VariablePool:
         Returns:
             变量是否存在
         """
-        return self._get_variable_struct(selector) is not None
+        path = self.transform_selector(selector)
+        struct = self._get_variable_struct(selector)
+        if struct is None:
+            return False
+        if len(path) == 3:
+            value = struct.instance.get_value()
+            return isinstance(value, dict) and path[2] in value
+        return True
 
     def lazy_namespace(self, namespace: str, literal: bool = False) -> LazyVariableDict:
         return LazyVariableDict(self.variables.get(namespace, {}), literal)

api/app/core/workflow/nodes/cycle_graph/iteration.py

@@ -28,86 +28,135 @@ class IterationRuntime:
 
     def __init__(
             self,
-            start_id: str,
             stream: bool,
-            graph: CompiledStateGraph,
             node_id: str,
             config: dict[str, Any],
             state: WorkflowState,
             variable_pool: VariablePool,
-            child_variable_pool: VariablePool,
+            cycle_nodes: list,
+            cycle_edges: list,
     ):
         """
         Initialize the iteration runtime.
 
         Args:
-            graph: Compiled workflow graph capable of async invocation.
-            node_id: Unique identifier of the loop node.
-            config: Dictionary containing iteration node configuration.
-            state: Current workflow state at the point of iteration.
+            stream:       Whether to run in streaming mode. When True, each iteration
+                          uses graph.astream and emits cycle_item events in real time.
+                          When False, graph.ainvoke is used instead.
+            node_id:      The unique identifier of the iteration node in the workflow.
+                          Also used as the variable namespace for item/index inside
+                          the subgraph (e.g. {{ node_id.item }}).
+            config:       Raw configuration dict for the iteration node, parsed into
+                          IterationNodeConfig. Controls input/output variable selectors,
+                          parallel execution settings, and output flattening.
+            state:        The parent workflow state at the point the iteration node is
+                          entered. Each task receives a copy of this state as its
+                          starting point.
+            variable_pool: The parent VariablePool containing all variables available
+                           at the time the iteration node executes, including sys.*,
+                           conv.*, and outputs from upstream nodes. Used as the source
+                           for deep-copying into each task's independent child pool.
+            cycle_nodes:  List of node config dicts belonging to this iteration's
+                          subgraph (i.e. nodes whose cycle field equals node_id).
+                          Passed to GraphBuilder when constructing each task's subgraph.
+            cycle_edges:  List of edge config dicts connecting nodes within the subgraph.
+                          Passed to GraphBuilder alongside cycle_nodes.
         """
-        self.start_id = start_id
         self.stream = stream
-        self.graph = graph
         self.state = state
         self.node_id = node_id
         self.typed_config = IterationNodeConfig(**config)
         self.looping = True
         self.variable_pool = variable_pool
-        self.child_variable_pool = child_variable_pool
+        self.cycle_nodes = cycle_nodes
+        self.cycle_edges = cycle_edges
         self.event_write = get_stream_writer()
-        self.checkpoint = RunnableConfig(
-            configurable={
-                "thread_id": uuid.uuid4()
-            }
-        )
 
         self.output_value = None
         self.result: list = []
 
-    async def _init_iteration_state(self, item, idx):
+    def _build_child_graph(self) -> tuple[CompiledStateGraph, VariablePool, str]:
         """
-        Initialize a per-iteration copy of the workflow state.
+        Build an independent compiled subgraph for a single iteration task.
 
-        Args:
-            item: Current element from the input array for this iteration.
-            idx: Index of the element in the input array.
+        Each call creates a brand-new VariablePool by deep-copying the parent pool,
+        then passes it to GraphBuilder. GraphBuilder binds this pool to every node's
+        execution closure at build time, so the pool and the subgraph always reference
+        the same object. This is the key design invariant: item/index written into the
+        pool after build will be visible to all nodes inside the subgraph.
 
         Returns:
-            A copy of the workflow state with iteration-specific variables set.
+            graph:      The compiled LangGraph subgraph ready for invocation.
+            child_pool: The VariablePool bound to this subgraph's node closures.
+                        Callers must write item/index into this pool before invoking
+                        the graph, and read output from it after invocation.
+            start_node_id: The ID of the CYCLE_START node inside the subgraph,
+                           used to set the initial activation signal in workflow state.
         """
-        loopstate = WorkflowState(
-            **self.state
+        from app.core.workflow.engine.graph_builder import GraphBuilder
+        child_pool = VariablePool()
+        child_pool.copy(self.variable_pool)
+        builder = GraphBuilder(
+            {"nodes": self.cycle_nodes, "edges": self.cycle_edges},
+            stream=self.stream,
+            variable_pool=child_pool,
+            cycle=self.node_id,
         )
-        self.child_variable_pool.copy(self.variable_pool)
-        await self.child_variable_pool.new(self.node_id, "item", item, VariableType.type_map(item), mut=True)
-        await self.child_variable_pool.new(self.node_id, "index", item, VariableType.type_map(item), mut=True)
-        loopstate["node_outputs"][self.node_id] = {
-            "item": item,
-            "index": idx,
-        }
+        graph = builder.build()
+        return graph, builder.variable_pool, builder.start_node_id
+
+    async def _init_iteration_state(self, item, idx, child_pool: VariablePool, start_id: str):
+        """
+        Initialize the workflow state for a single iteration.
+
+        Writes the current item and its index into child_pool under the iteration
+        node's namespace (e.g. iteration_xxx.item, iteration_xxx.index), making them
+        accessible to downstream nodes inside the subgraph via variable selectors.
+
+        Also prepares a copy of the parent workflow state with:
+        - node_outputs[node_id] set to {item, index} so the state snapshot is consistent
+          with the pool values.
+        - looping flag set to 1 (active) to signal the subgraph is inside a cycle.
+        - activate[start_id] set to True to trigger the CYCLE_START node.
+
+        Args:
+            item:       The current element from the input array.
+            idx:        The zero-based index of this element in the input array.
+            child_pool: The VariablePool bound to this iteration's subgraph.
+                        Must be the same object returned by _build_child_graph.
+            start_id:   The ID of the CYCLE_START node inside the subgraph.
+
+        Returns:
+            A WorkflowState instance ready to be passed to graph.ainvoke or graph.astream.
+        """
+        loopstate = WorkflowState(**self.state)
+        await child_pool.new(self.node_id, "item", item, VariableType.type_map(item), mut=True)
+        await child_pool.new(self.node_id, "index", idx, VariableType.type_map(idx), mut=True)
+        loopstate["node_outputs"][self.node_id] = {"item": item, "index": idx}
         loopstate["looping"] = 1
-        loopstate["activate"][self.start_id] = True
+        loopstate["activate"][start_id] = True
         return loopstate
 
-    def merge_conv_vars(self):
-        self.variable_pool.variables["conv"].update(
-            self.child_variable_pool.variables["conv"]
-        )
+    def _merge_conv_vars(self, child_pool: VariablePool):
+        self.variable_pool.variables["conv"].update(child_pool.variables["conv"])
 
     async def run_task(self, item, idx):
         """
         Execute a single iteration asynchronously.
+        Each task builds its own subgraph so the variable pool closure is independent.
 
-        Args:
-            item: The input element for this iteration.
-            idx: The index of this iteration.
+        Returns:
+            Tuple of (idx, output, result, child_pool, stopped)
         """
+        graph, child_pool, start_id = self._build_child_graph()
+        checkpoint = RunnableConfig(configurable={"thread_id": uuid.uuid4()})
+        init_state = await self._init_iteration_state(item, idx, child_pool, start_id)
+
         if self.stream:
-            async for event in self.graph.astream(
-                    await self._init_iteration_state(item, idx),
+            async for event in graph.astream(
+                    init_state,
                     stream_mode=["debug"],
-                    config=self.checkpoint
+                    config=checkpoint
             ):
                 if isinstance(event, tuple) and len(event) == 2:
                     mode, data = event
@@ -117,7 +166,6 @@ class IterationRuntime:
                     event_type = data.get("type")
                     payload = data.get("payload", {})
                     node_name = payload.get("name")
-
                     if node_name and node_name.startswith("nop"):
                         continue
                     if event_type == "task_result":
@@ -140,17 +188,13 @@ class IterationRuntime:
                                 "token_usage": result.get("node_outputs", {}).get(node_name, {}).get("token_usage")
                             }
                         })
-            result = self.graph.get_state(config=self.checkpoint).values
+            result = graph.get_state(config=checkpoint).values
         else:
-            result = await self.graph.ainvoke(await self._init_iteration_state(item, idx))
-        output = self.child_variable_pool.get_value(self.output_value)
-        if isinstance(output, list) and self.typed_config.flatten:
-            self.result.extend(output)
-        else:
-            self.result.append(output)
-        if result["looping"] == 2:
-            self.looping = False
-        return result
+            result = await graph.ainvoke(init_state)
+
+        output = child_pool.get_value(self.output_value)
+        stopped = result["looping"] == 2
+        return idx, output, result, child_pool, stopped
 
     def _create_iteration_tasks(self, array_obj, idx):
         """
@@ -196,16 +240,32 @@ class IterationRuntime:
                 tasks = self._create_iteration_tasks(array_obj, idx)
                 logger.info(f"Iteration node {self.node_id}: running, concurrency {len(tasks)}")
                 idx += self.typed_config.parallel_count
-                child_state.extend(await asyncio.gather(*tasks))
-                self.merge_conv_vars()
+                batch = await asyncio.gather(*tasks)
+                # Sort by idx to preserve order, then collect results
+                batch_sorted = sorted(batch, key=lambda x: x[0])
+                for _, output, result, child_pool, stopped in batch_sorted:
+                    if isinstance(output, list) and self.typed_config.flatten:
+                        self.result.extend(output)
+                    else:
+                        self.result.append(output)
+                    child_state.append(result)
+                    self._merge_conv_vars(child_pool)
+                    if stopped:
+                        self.looping = False
         else:
             # Execute iterations sequentially
             while idx < len(array_obj) and self.looping:
                 logger.info(f"Iteration node {self.node_id}: running")
                 item = array_obj[idx]
-                result = await self.run_task(item, idx)
-                self.merge_conv_vars()
+                _, output, result, child_pool, stopped = await self.run_task(item, idx)
+                if isinstance(output, list) and self.typed_config.flatten:
+                    self.result.extend(output)
+                else:
+                    self.result.append(output)
+                self._merge_conv_vars(child_pool)
                 child_state.append(result)
+                if stopped:
+                    self.looping = False
                 idx += 1
         logger.info(f"Iteration node {self.node_id}: execution completed")
         return {

api/app/core/workflow/nodes/cycle_graph/node.py

@@ -55,9 +55,9 @@ class CycleGraphNode(BaseNode):
             if config.output_type in [
                 VariableType.ARRAY_FILE,
                 VariableType.ARRAY_STRING,
-                VariableType.NUMBER,
+                VariableType.ARRAY_NUMBER,
                 VariableType.ARRAY_OBJECT,
-                VariableType.BOOLEAN
+                VariableType.ARRAY_BOOLEAN
             ]:
                 if config.flatten:
                     outputs['output'] = config.output_type
@@ -123,7 +123,7 @@ class CycleGraphNode(BaseNode):
 
         return cycle_nodes, cycle_edges
 
-    def build_graph(self):
+    def build_graph(self, variable_pool: VariablePool):
         """
         Build and compile the internal subgraph for this cycle node.
 
@@ -135,6 +135,7 @@ class CycleGraphNode(BaseNode):
         from app.core.workflow.engine.graph_builder import GraphBuilder
 
         self.child_variable_pool = VariablePool()
+        self.child_variable_pool.copy(variable_pool)
         builder = GraphBuilder(
             {
                 "nodes": self.cycle_nodes,
@@ -165,8 +166,8 @@ class CycleGraphNode(BaseNode):
         Raises:
             RuntimeError: If the node type is unsupported.
         """
-        self.build_graph()
         if self.node_type == NodeType.LOOP:
+            self.build_graph(variable_pool)
             return await LoopRuntime(
                 start_id=self.start_node_id,
                 stream=False,
@@ -179,20 +180,19 @@ class CycleGraphNode(BaseNode):
             ).run()
         if self.node_type == NodeType.ITERATION:
             return await IterationRuntime(
-                start_id=self.start_node_id,
                 stream=False,
-                graph=self.graph,
                 node_id=self.node_id,
                 config=self.config,
                 state=state,
                 variable_pool=variable_pool,
-                child_variable_pool=self.child_variable_pool
+                cycle_nodes=self.cycle_nodes,
+                cycle_edges=self.cycle_edges,
             ).run()
         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:
+            self.build_graph(variable_pool)
             yield {
                 "__final__": True,
                 "result": await LoopRuntime(
@@ -211,14 +211,13 @@ class CycleGraphNode(BaseNode):
             yield {
                 "__final__": True,
                 "result": await IterationRuntime(
-                    start_id=self.start_node_id,
                     stream=True,
-                    graph=self.graph,
                     node_id=self.node_id,
                     config=self.config,
                     state=state,
                     variable_pool=variable_pool,
-                    child_variable_pool=self.child_variable_pool
+                    cycle_nodes=self.cycle_nodes,
+                    cycle_edges=self.cycle_edges,
                 ).run()
             }
             return

api/app/core/workflow/nodes/http_request/config.py

@@ -72,8 +72,9 @@ class HttpContentTypeConfig(BaseModel):
     @classmethod
     def validate_data(cls, v, info):
         content_type = info.data.get("content_type")
-        if content_type == HttpContentType.FROM_DATA and not isinstance(v, HttpFormData):
-            raise ValueError("When content_type is 'form-data', data must be of type HttpFormData")
+        if content_type == HttpContentType.FROM_DATA and (
+                not isinstance(v, list) or not all(isinstance(item, HttpFormData) for item in v)):
+            raise ValueError("When content_type is 'form-data', data must be a list of HttpFormData")
         elif content_type in [HttpContentType.JSON] and not isinstance(v, str):
             raise ValueError("When content_type is JSON, data must be of type str")
         elif content_type in [HttpContentType.WWW_FORM] and not isinstance(v, dict):

api/app/core/workflow/nodes/http_request/node.py

@@ -260,17 +260,22 @@ class HttpRequestNode(BaseNode):
                 ))
             case HttpContentType.FROM_DATA:
                 data = {}
-                content["files"] = {}
+                files = []
                 for item in self.typed_config.body.data:
+                    key = self._render_template(item.key, variable_pool)
                     if item.type == "text":
-                        data[self._render_template(item.key, variable_pool)] = self._render_template(item.value,
-                                                                                                     variable_pool)
+                        data[key] = self._render_template(item.value, variable_pool)
                     elif item.type == "file":
-                        content["files"][self._render_template(item.key, variable_pool)] = (
-                            uuid.uuid4().hex,
-                            await variable_pool.get_instance(item.value).get_content()
-                        )
+                        file_instance = variable_pool.get_instance(item.value)
+                        if isinstance(file_instance, ArrayVariable):
+                            for v in file_instance.value:
+                                if isinstance(v, FileVariable):
+                                    files.append((key, (uuid.uuid4().hex, await v.get_content())))
+                        elif isinstance(file_instance, FileVariable):
+                            files.append((key, (uuid.uuid4().hex, await file_instance.get_content())))
                 content["data"] = data
+                if files:
+                    content["files"] = files
             case HttpContentType.BINARY:
                 content["files"] = []
                 file_instence = variable_pool.get_instance(self.typed_config.body.data)

api/app/core/workflow/nodes/knowledge/node.py

@@ -8,6 +8,8 @@ from langchain_core.documents import Document
 from app.core.error_codes import BizCode
 from app.core.exceptions import BusinessException
 from app.core.models import RedBearRerank, RedBearModelConfig
+from app.core.rag.llm.chat_model import Base
+from app.core.rag.llm.embedding_model import OpenAIEmbed
 from app.core.rag.models.chunk import DocumentChunk
 from app.core.rag.vdb.elasticsearch.elasticsearch_vector import ElasticSearchVectorFactory
 from app.core.workflow.engine.state_manager import WorkflowState
@@ -39,8 +41,9 @@ class KnowledgeRetrievalNode(BaseNode):
         if isinstance(business_result, dict) and "chunks" in business_result:
             return business_result["chunks"]
         return business_result
-
-    def _extract_citations(self, business_result: Any) -> list:
+    
+    @staticmethod
+    def _extract_citations(business_result: Any) -> list:
         if isinstance(business_result, dict):
             return business_result.get("citations", [])
         return []
@@ -230,23 +233,23 @@ class KnowledgeRetrievalNode(BaseNode):
                         }
                     )
                 )
-            case RetrieveType.HYBRID:
+            case retrieve_type if retrieve_type in (RetrieveType.HYBRID, RetrieveType.Graph):
                 rs1_task = asyncio.to_thread(
-                        vector_service.search_by_vector, **{
-                            "query": query,
-                            "top_k": kb_config.top_k,
-                            "indices": indices,
-                            "score_threshold": kb_config.vector_similarity_weight
-                        }
-                    )
+                    vector_service.search_by_vector, **{
+                        "query": query,
+                        "top_k": kb_config.top_k,
+                        "indices": indices,
+                        "score_threshold": kb_config.vector_similarity_weight
+                    }
+                )
                 rs2_task = asyncio.to_thread(
-                        vector_service.search_by_full_text, **{
-                            "query": query,
-                            "top_k": kb_config.top_k,
-                            "indices": indices,
-                            "score_threshold": kb_config.similarity_threshold
-                        }
-                    )
+                    vector_service.search_by_full_text, **{
+                        "query": query,
+                        "top_k": kb_config.top_k,
+                        "indices": indices,
+                        "score_threshold": kb_config.similarity_threshold
+                    }
+                )
                 rs1, rs2 = await asyncio.gather(rs1_task, rs2_task)
 
                 # Deduplicate hybrid retrieval results
@@ -266,6 +269,33 @@ class KnowledgeRetrievalNode(BaseNode):
                         key=lambda d: d.metadata.get("score", 0),
                         reverse=True
                     )[:kb_config.top_k])
+                if kb_config.retrieve_type == RetrieveType.Graph:
+                    from app.core.rag.common.settings import kg_retriever
+                    llm_key = self.model_balance(db_knowledge.llm)
+                    emb_key = self.model_balance(db_knowledge.embedding)
+                    chat_model = Base(
+                        key=llm_key.api_key,
+                        model_name=llm_key.model_name,
+                        base_url=llm_key.api_base
+                    )
+                    embedding_model = OpenAIEmbed(
+                        key=emb_key.api_key,
+                        model_name=emb_key.model_name,
+                        base_url=emb_key.api_base
+                    )
+                    doc = await asyncio.to_thread(
+                        kg_retriever.retrieval,
+                        question=query,
+                        workspace_ids=[str(db_knowledge.workspace_id)],
+                        kb_ids=[str(kb_config.kb_id)],
+                        emb_mdl=embedding_model,
+                        llm=chat_model
+                    )
+                    if doc:
+                        rs.insert(0, DocumentChunk(
+                            page_content=doc.get("page_content", ""),
+                            metadata=doc.get("metadata", {})
+                        ))
             case _:
                 raise RuntimeError("Unknown retrieval type")
         return rs

api/app/core/workflow/nodes/tool/node.py

@@ -15,6 +15,7 @@ from app.services.tool_service import ToolService
 logger = logging.getLogger(__name__)
 
 TEMPLATE_PATTERN = re.compile(r"\{\{.*?}}")
+PURE_VARIABLE_PATTERN = re.compile(r"^\{\{\s*([\w.]+)\s*}}$")
 
 
 class ToolNode(BaseNode):
@@ -52,13 +53,21 @@ class ToolNode(BaseNode):
         # 渲染工具参数
         rendered_parameters = {}
         for param_name, param_template in self.typed_config.tool_parameters.items():
-            if isinstance(param_template, str) and TEMPLATE_PATTERN.search(param_template):
-                try:
-                    rendered_value = self._render_template(param_template, variable_pool)
-                except Exception as e:
-                    raise ValueError(f"模板渲染失败：参数 {param_name} 的模板 {param_template} 解析错误") from e
+            if isinstance(param_template, str):
+                pure_match = PURE_VARIABLE_PATTERN.match(param_template)
+                if pure_match:
+                    # 纯单变量引用直接取原始值，保留 int/bool/float 等类型
+                    rendered_value = self.get_variable(pure_match.group(1), variable_pool, strict=False)
+                    if rendered_value is None:
+                        rendered_value = self._render_template(param_template, variable_pool)
+                elif TEMPLATE_PATTERN.search(param_template):
+                    try:
+                        rendered_value = self._render_template(param_template, variable_pool)
+                    except Exception as e:
+                        raise ValueError(f"模板渲染失败：参数 {param_name} 的模板 {param_template} 解析错误") from e
+                else:
+                    rendered_value = param_template
             else:
-                # 非模板参数（数字/布尔/普通字符串）直接保留原值
                 rendered_value = param_template
             rendered_parameters[param_name] = rendered_value
 

api/app/core/workflow/variable/variable_objects.py

@@ -84,7 +84,7 @@ class FileVariable(BaseVariable):
         total_bytes = 0
         chunks = []
 
-        async with httpx.AsyncClient() as client:
+        async with httpx.AsyncClient(follow_redirects=True) as client:
             async with client.stream("GET", self.value.url) as resp:
                 resp.raise_for_status()
                 async for chunk in resp.aiter_bytes(8192):

api/app/repositories/app_repository.py

@@ -61,3 +61,15 @@ def get_apps_by_id(db: Session, app_id: uuid.UUID) -> App:
     """根据工作空间ID查询应用"""
     repo = AppRepository(db)
     return repo.get_apps_by_id(app_id)
+
+
+def get_release_by_id(db: Session, app_id: uuid.UUID, release_id: uuid.UUID):
+    """根据发布版本ID查询发布快照（仅返回激活状态）"""
+    from app.models.app_release_model import AppRelease
+    return db.scalars(
+        select(AppRelease).where(
+            AppRelease.app_id == app_id,
+            AppRelease.id == release_id,
+            AppRelease.is_active.is_(True),
+        )
+    ).first()

api/app/repositories/implicit_emotions_storage_repository.py

@@ -5,16 +5,9 @@ Implicit Emotions Storage Repository
 事务由调用方控制，仓储层只使用 flush/refresh
 """
 import logging
-from datetime import date, datetime, timezone
+from datetime import datetime, timedelta, timezone
 from typing import Generator, Optional
 
-
-class TimeFilterUnavailableError(Exception):
-    """redis_client 不可用，无法执行时间轴筛选。
-    
-    调用方捕获此异常后可选择回退到 get_all_user_ids 进行全量处理。
-    """
-
 import redis
 from sqlalchemy import exists, not_, select
 from sqlalchemy.orm import Session
@@ -25,6 +18,13 @@ from app.models.implicit_emotions_storage_model import ImplicitEmotionsStorage
 logger = logging.getLogger(__name__)
 
 
+class TimeFilterUnavailableError(Exception):
+    """redis_client 不可用，无法执行时间轴筛选。
+    
+    调用方捕获此异常后可选择回退到 get_all_user_ids 进行全量处理。
+    """
+
+
 class ImplicitEmotionsStorageRepository:
     """隐性记忆和情绪存储仓储类"""
 
@@ -216,9 +216,7 @@ class ImplicitEmotionsStorageRepository:
         """
         from sqlalchemy import String as SAString
         from sqlalchemy import cast
-        CST = timezone(timedelta(hours=8))
-        now_cst = datetime.now(CST)
-        today_start = now_cst.replace(hour=0, minute=0, second=0, microsecond=0).astimezone(timezone.utc).replace(tzinfo=None)
+        today_start = datetime.utcnow().replace(hour=0, minute=0, second=0, microsecond=0)
         tomorrow_start = today_start + timedelta(days=1)
         offset = 0
         while True:

api/app/repositories/neo4j/cypher_queries.py

@@ -23,6 +23,7 @@ SET s += {
     end_user_id: statement.end_user_id,
     stmt_type: statement.stmt_type,
     statement: statement.statement,
+    speaker: statement.speaker,
     emotion_intensity: statement.emotion_intensity,
     emotion_target: statement.emotion_target,
     emotion_subject: statement.emotion_subject,
@@ -41,6 +42,9 @@ SET s += {
     last_access_time: statement.last_access_time,
     access_count: statement.access_count
 }
+SET s.importance_score = coalesce(s.importance_score, 0.5),
+    s.activation_value = coalesce(s.activation_value, s.importance_score, 0.5),
+    s.access_count = coalesce(s.access_count, 0)
 RETURN s.id AS uuid
 """
 
@@ -56,6 +60,7 @@ SET c += {
     expired_at: chunk.expired_at,
     dialog_id: chunk.dialog_id,
     content: chunk.content,
+    speaker: chunk.speaker,
     chunk_embedding: chunk.chunk_embedding,
     sequence_number: chunk.sequence_number,
     start_index: chunk.start_index,
@@ -91,6 +96,8 @@ SET e.name = CASE WHEN entity.name IS NOT NULL AND entity.name <> '' THEN entity
     END,
     e.statement_id = CASE WHEN entity.statement_id IS NOT NULL AND entity.statement_id <> '' THEN entity.statement_id ELSE e.statement_id END,
     e.aliases = CASE
+        // 用户实体的 aliases 由 PgSQL end_user_info 作为唯一权威源，知识抽取完全不写入
+        WHEN entity.name IN ['用户', '我', 'User', 'I'] THEN e.aliases
         WHEN entity.aliases IS NOT NULL AND size(entity.aliases) > 0
         THEN CASE 
             WHEN e.aliases IS NULL THEN entity.aliases 
@@ -116,7 +123,7 @@ SET e.name = CASE WHEN entity.name IS NOT NULL AND entity.name <> '' THEN entity
         END
     END,
     e.importance_score = CASE WHEN entity.importance_score IS NOT NULL THEN entity.importance_score ELSE coalesce(e.importance_score, 0.5) END,
-    e.activation_value = CASE WHEN entity.activation_value IS NOT NULL THEN entity.activation_value ELSE e.activation_value END,
+    e.activation_value = CASE WHEN entity.activation_value IS NOT NULL THEN entity.activation_value ELSE coalesce(e.activation_value, e.importance_score, 0.5) END,
     e.access_history = CASE WHEN entity.access_history IS NOT NULL THEN entity.access_history ELSE coalesce(e.access_history, []) END,
     e.last_access_time = CASE WHEN entity.last_access_time IS NOT NULL THEN entity.last_access_time ELSE e.last_access_time END,
     e.access_count = CASE WHEN entity.access_count IS NOT NULL THEN entity.access_count ELSE coalesce(e.access_count, 0) END,
@@ -161,6 +168,7 @@ SET e += {
 }
 // Independent weak flag，仅标记弱关系，不再维护 relations 聚合字段
 SET e.is_weak = true
+SET e.activation_value = coalesce(e.activation_value, 0.5)
 RETURN e.id AS id
 """
 
@@ -171,10 +179,12 @@ MERGE (s:ExtractedEntity {id: item.source_id, run_id: item.run_id})
 SET s += {name: item.subject, end_user_id: item.end_user_id, run_id: item.run_id}
 // Independent strong flag
 SET s.is_strong = true
+SET s.activation_value = coalesce(s.activation_value, 0.5)
 MERGE (o:ExtractedEntity {id: item.target_id, run_id: item.run_id})
 SET o += {name: item.object, end_user_id: item.end_user_id, run_id: item.run_id}
 // Independent strong flag
 SET o.is_strong = true
+SET o.activation_value = coalesce(o.activation_value, 0.5)
 """
 
 
@@ -283,7 +293,7 @@ LIMIT $limit
 """
 
 SEARCH_STATEMENTS_BY_KEYWORD = """
-CALL db.index.fulltext.queryNodes("statementsFulltext", $q) YIELD node AS s, score
+CALL db.index.fulltext.queryNodes("statementsFulltext", $query) YIELD node AS s, score
 WHERE ($end_user_id IS NULL OR s.end_user_id = $end_user_id)
 OPTIONAL MATCH (c:Chunk)-[:CONTAINS]->(s)
 OPTIONAL MATCH (s)-[:REFERENCES_ENTITY]->(e:ExtractedEntity)
@@ -307,7 +317,7 @@ LIMIT $limit
 """
 # 查询实体名称包含指定字符串的实体
 SEARCH_ENTITIES_BY_NAME = """
-CALL db.index.fulltext.queryNodes("entitiesFulltext", $q) YIELD node AS e, score
+CALL db.index.fulltext.queryNodes("entitiesFulltext", $query) YIELD node AS e, score
 WHERE ($end_user_id IS NULL OR e.end_user_id = $end_user_id)
 OPTIONAL MATCH (s:Statement)-[:REFERENCES_ENTITY]->(e)
 OPTIONAL MATCH (c:Chunk)-[:CONTAINS]->(s)
@@ -337,21 +347,21 @@ LIMIT $limit
 """
 
 SEARCH_ENTITIES_BY_NAME_OR_ALIAS = """
-CALL db.index.fulltext.queryNodes("entitiesFulltext", $q) YIELD node AS e, score
+CALL db.index.fulltext.queryNodes("entitiesFulltext", $query) YIELD node AS e, score
 WHERE ($end_user_id IS NULL OR e.end_user_id = $end_user_id)
 WITH e, score
-WITH collect({entity: e, score: score}) AS fulltextResults
+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))
+  AND ANY(alias IN ae.aliases WHERE toLower(alias) CONTAINS toLower($query))
 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
+       WHEN ANY(alias IN x.aliases WHERE toLower(alias) = toLower($query)) THEN 1.0
+       WHEN ANY(alias IN x.aliases WHERE toLower(alias) STARTS WITH toLower($query)) THEN 0.9
        ELSE 0.8
      END
 }]) AS row
@@ -384,7 +394,7 @@ LIMIT $limit
 
 
 SEARCH_CHUNKS_BY_CONTENT = """
-CALL db.index.fulltext.queryNodes("chunksFulltext", $q) YIELD node AS c, score
+CALL db.index.fulltext.queryNodes("chunksFulltext", $query) YIELD node AS c, score
 WHERE ($end_user_id IS NULL OR c.end_user_id = $end_user_id)
 OPTIONAL MATCH (c)-[:CONTAINS]->(s:Statement)
 OPTIONAL MATCH (s)-[:REFERENCES_ENTITY]->(e:ExtractedEntity)
@@ -501,7 +511,7 @@ LIMIT $limit
 """
 
 SEARCH_STATEMENTS_BY_KEYWORD_TEMPORAL = """
-CALL db.index.fulltext.queryNodes("statementsFulltext", $q) YIELD node AS s, score
+CALL db.index.fulltext.queryNodes("statementsFulltext", $query) YIELD node AS s, score
 WHERE ($end_user_id IS NULL OR s.end_user_id = $end_user_id)
   AND ((($start_date IS NULL OR (s.created_at IS NOT NULL AND datetime(s.created_at) >= datetime($start_date)))
   AND ($end_date IS NULL OR (s.created_at IS NOT NULL AND datetime(s.created_at) <= datetime($end_date))))
@@ -677,7 +687,7 @@ SET n.invalid_at = $new_invalid_at
 
 # MemorySummary keyword search using fulltext index
 SEARCH_MEMORY_SUMMARIES_BY_KEYWORD = """
-CALL db.index.fulltext.queryNodes("summariesFulltext", $q) YIELD node AS m, score
+CALL db.index.fulltext.queryNodes("summariesFulltext", $query) YIELD node AS m, score
 WHERE ($end_user_id IS NULL OR m.end_user_id = $end_user_id)
 OPTIONAL MATCH (m)-[:DERIVED_FROM_STATEMENT]->(s:Statement)
 RETURN m.id AS id,
@@ -735,7 +745,7 @@ SET m += {
     summary_embedding: summary.summary_embedding,
     config_id: summary.config_id,
     importance_score: CASE WHEN summary.importance_score IS NOT NULL THEN summary.importance_score ELSE coalesce(m.importance_score, 0.5) END,
-    activation_value: CASE WHEN summary.activation_value IS NOT NULL THEN summary.activation_value ELSE m.activation_value END,
+    activation_value: CASE WHEN summary.activation_value IS NOT NULL THEN summary.activation_value ELSE coalesce(m.activation_value, m.importance_score, 0.5) END,
     access_history: CASE WHEN summary.access_history IS NOT NULL THEN summary.access_history ELSE coalesce(m.access_history, []) END,
     last_access_time: CASE WHEN summary.last_access_time IS NOT NULL THEN summary.last_access_time ELSE m.last_access_time END,
     access_count: CASE WHEN summary.access_count IS NOT NULL THEN summary.access_count ELSE coalesce(m.access_count, 0) END
@@ -1363,7 +1373,7 @@ 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
+CALL db.index.fulltext.queryNodes("communitiesFulltext", $query) YIELD node AS c, score
 WHERE ($end_user_id IS NULL OR c.end_user_id = $end_user_id)
 RETURN c.community_id AS id,
        c.name AS name,
@@ -1451,7 +1461,7 @@ RETURN elementId(r) AS uuid
 """
 
 SEARCH_PERCEPTUAL_BY_KEYWORD = """
-CALL db.index.fulltext.queryNodes("perceptualFulltext", $q) YIELD node AS p, score
+CALL db.index.fulltext.queryNodes("perceptualFulltext", $query) YIELD node AS p, score
 WHERE p.end_user_id = $end_user_id
 RETURN p.id AS id,
        p.end_user_id AS end_user_id,

api/app/repositories/neo4j/graph_saver.py

@@ -186,6 +186,58 @@ async def save_dialog_and_statements_to_neo4j(
     Returns:
         bool: True if successful, False otherwise
     """
+    # TODO 需要在去重消歧节阶段，做以下逻辑的处理
+    # 预处理：对特殊实体（"用户"、"AI助手"）复用 Neo4j 中已有节点的 ID，
+    # 确保同一个 end_user_id 下只有一个"用户"节点和一个"AI助手"节点。
+    if entity_nodes:
+        _SPECIAL_NAMES = {"用户", "我", "user", "i", "ai助手", "助手", "ai assistant", "assistant"}
+        end_user_id = entity_nodes[0].end_user_id if entity_nodes else None
+        if end_user_id:
+            try:
+                # 查询已有的特殊实体
+                cypher = """
+                MATCH (e:ExtractedEntity)
+                WHERE e.end_user_id = $end_user_id AND toLower(e.name) IN $names
+                RETURN e.id AS id, e.name AS name
+                """
+                existing = await connector.execute_query(
+                    cypher,
+                    end_user_id=end_user_id,
+                    names=list(_SPECIAL_NAMES),
+                )
+                # 建立 name(lower) → existing_id 映射
+                existing_id_map = {}
+                for record in (existing or []):
+                    name_lower = (record.get("name") or "").strip().lower()
+                    if name_lower and record.get("id"):
+                        existing_id_map[name_lower] = record["id"]
+
+                if existing_id_map:
+                    # 替换新实体的 ID 为已有 ID，同时更新所有引用该 ID 的边
+                    for ent in entity_nodes:
+                        name_lower = (ent.name or "").strip().lower()
+                        if name_lower in existing_id_map:
+                            old_id = ent.id
+                            new_id = existing_id_map[name_lower]
+                            if old_id != new_id:
+                                ent.id = new_id
+                                # 更新 statement_entity_edges 中的引用
+                                for edge in statement_entity_edges:
+                                    if edge.target == old_id:
+                                        edge.target = new_id
+                                    if edge.source == old_id:
+                                        edge.source = new_id
+                                # 更新 entity_edges 中的引用
+                                for edge in entity_edges:
+                                    if edge.source == old_id:
+                                        edge.source = new_id
+                                    if edge.target == old_id:
+                                        edge.target = new_id
+                                logger.info(
+                                    f"特殊实体 '{ent.name}' ID 复用: {old_id[:8]}... → {new_id[:8]}..."
+                                )
+            except Exception as e:
+                logger.warning(f"特殊实体 ID 复用查询失败（不影响写入）: {e}")
 
     # 定义事务函数，将所有写操作放在一个事务中
     async def _save_all_in_transaction(tx):

api/app/repositories/neo4j/graph_search.py

@@ -2,6 +2,7 @@ import asyncio
 import logging
 from typing import Any, Dict, List, Optional
 
+from app.core.memory.utils.data.text_utils import escape_lucene_query
 from app.repositories.neo4j.cypher_queries import (
     CHUNK_EMBEDDING_SEARCH,
     COMMUNITY_EMBEDDING_SEARCH,
@@ -87,7 +88,7 @@ async def _update_activation_values_batch(
             unique_node_ids.append(node_id)
 
     if not unique_node_ids:
-        logger.warning(f"批量更新激活值：没有有效的节点ID")
+        logger.warning("批量更新激活值：没有有效的节点ID")
         return nodes
 
     # 记录去重信息（仅针对具有有效 ID 的节点）
@@ -223,7 +224,7 @@ async def _update_search_results_activation(
 
 async def search_graph(
         connector: Neo4jConnector,
-        q: str,
+        query: str,
         end_user_id: Optional[str] = None,
         limit: int = 50,
         include: List[str] = None,
@@ -234,14 +235,14 @@ async def search_graph(
     OPTIMIZED: Runs all queries in parallel using asyncio.gather()
     INTEGRATED: Updates activation values for knowledge nodes before returning results
 
-    - Statements: matches s.statement CONTAINS q
-    - Entities: matches e.name CONTAINS q
-    - Chunks: matches s.content CONTAINS q (from Statement nodes)
-    - Summaries: matches ms.content CONTAINS q
+    - Statements: matches s.statement CONTAINS query
+    - Entities: matches e.name CONTAINS query
+    - Chunks: matches s.content CONTAINS query (from Statement nodes)
+    - Summaries: matches ms.content CONTAINS query
 
     Args:
         connector: Neo4j connector
-        q: Query text
+        query: Query text for full-text search
         end_user_id: Optional group filter
         limit: Max results per category
         include: List of categories to search (default: all)
@@ -252,6 +253,9 @@ async def search_graph(
     if include is None:
         include = ["statements", "chunks", "entities", "summaries"]
 
+    # Escape Lucene special characters to prevent query parse errors
+    escaped_query = escape_lucene_query(query)
+
     # Prepare tasks for parallel execution
     tasks = []
     task_keys = []
@@ -260,7 +264,7 @@ async def search_graph(
         tasks.append(connector.execute_query(
             SEARCH_STATEMENTS_BY_KEYWORD,
             json_format=True,
-            q=q,
+            query=escaped_query,
             end_user_id=end_user_id,
             limit=limit,
         ))
@@ -270,7 +274,7 @@ async def search_graph(
         tasks.append(connector.execute_query(
             SEARCH_ENTITIES_BY_NAME_OR_ALIAS,
             json_format=True,
-            q=q,
+            query=escaped_query,
             end_user_id=end_user_id,
             limit=limit,
         ))
@@ -280,7 +284,7 @@ async def search_graph(
         tasks.append(connector.execute_query(
             SEARCH_CHUNKS_BY_CONTENT,
             json_format=True,
-            q=q,
+            query=escaped_query,
             end_user_id=end_user_id,
             limit=limit,
         ))
@@ -290,7 +294,7 @@ async def search_graph(
         tasks.append(connector.execute_query(
             SEARCH_MEMORY_SUMMARIES_BY_KEYWORD,
             json_format=True,
-            q=q,
+            query=escaped_query,
             end_user_id=end_user_id,
             limit=limit,
         ))
@@ -300,7 +304,7 @@ async def search_graph(
         tasks.append(connector.execute_query(
             SEARCH_COMMUNITIES_BY_KEYWORD,
             json_format=True,
-            q=q,
+            query=escaped_query,
             end_user_id=end_user_id,
             limit=limit,
         ))
@@ -482,7 +486,7 @@ async def search_graph_by_embedding(
         update_time = time.time() - update_start
         logger.info(f"[PERF] Activation value updates took: {update_time:.4f}s")
     else:
-        logger.info(f"[PERF] Skipping activation updates (only summaries)")
+        logger.info("[PERF] Skipping activation updates (only summaries)")
 
     return results
 
@@ -520,7 +524,7 @@ async def get_dedup_candidates_for_entities(  # 适配新版查询：使用全
                 # 全文索引按名称检索（包含 CONTAINS 语义）
                 rows = await connector.execute_query(
                     SEARCH_ENTITIES_BY_NAME,
-                    q=name,
+                    query=escape_lucene_query(name),
                     end_user_id=end_user_id,
                     limit=100,
                 )
@@ -544,7 +548,7 @@ async def get_dedup_candidates_for_entities(  # 适配新版查询：使用全
                 try:
                     rows = await connector.execute_query(
                         SEARCH_ENTITIES_BY_NAME,
-                        q=name.lower(),
+                        query=escape_lucene_query(name.lower()),
                         end_user_id=end_user_id,
                         limit=100,
                     )
@@ -593,11 +597,12 @@ async def search_graph_by_keyword_temporal(
     - Returns up to 'limit' statements
     """
     if not query_text:
-        logger.warning(f"query_text不能为空")
+        logger.warning("query_text不能为空")
         return {"statements": []}
+    escaped_query = escape_lucene_query(query_text)
     statements = await connector.execute_query(
         SEARCH_STATEMENTS_BY_KEYWORD_TEMPORAL,
-        q=query_text,
+        query=escaped_query,
         end_user_id=end_user_id,
         start_date=start_date,
         end_date=end_date,
@@ -671,7 +676,7 @@ async def search_graph_by_dialog_id(
     - Returns up to 'limit' dialogues
     """
     if not dialog_id:
-        logger.warning(f"dialog_id不能为空")
+        logger.warning("dialog_id不能为空")
         return {"dialogues": []}
 
     dialogues = await connector.execute_query(
@@ -690,7 +695,7 @@ async def search_graph_by_chunk_id(
         limit: int = 1,
 ) -> Dict[str, List[Dict[str, Any]]]:
     if not chunk_id:
-        logger.warning(f"chunk_id不能为空")
+        logger.warning("chunk_id不能为空")
         return {"chunks": []}
     chunks = await connector.execute_query(
         SEARCH_CHUNK_BY_CHUNK_ID,
@@ -968,7 +973,7 @@ async def search_graph_l_valid_at(
 
 async def search_perceptual(
         connector: Neo4jConnector,
-        q: str,
+        query: str,
         end_user_id: Optional[str] = None,
         limit: int = 10,
 ) -> Dict[str, List[Dict[str, Any]]]:
@@ -979,7 +984,7 @@ async def search_perceptual(
 
     Args:
         connector: Neo4j connector
-        q: Query text
+        query: Query text for full-text search
         end_user_id: Optional user filter
         limit: Max results
 
@@ -989,7 +994,7 @@ async def search_perceptual(
     try:
         perceptuals = await connector.execute_query(
             SEARCH_PERCEPTUAL_BY_KEYWORD,
-            q=q,
+            query=escape_lucene_query(query),
             end_user_id=end_user_id,
             limit=limit,
         )

api/app/repositories/neo4j/neo4j_connector.py

@@ -9,12 +9,15 @@ Classes:
 """
 
 from typing import Any, List, Dict
+import logging
 
 from neo4j import AsyncGraphDatabase, basic_auth
 from neo4j.time import DateTime as Neo4jDateTime, Date as Neo4jDate, Time as Neo4jTime, Duration as Neo4jDuration
 
 from app.core.config import settings
 
+logger = logging.getLogger(__name__)
+
 
 def _convert_neo4j_types(value: Any) -> Any:
     """递归将 neo4j 原生时间类型转为 Python 原生类型 / ISO 字符串，确保可被 json.dumps 序列化。"""
@@ -67,7 +70,12 @@ class Neo4jConnector:
             )
         self.driver = AsyncGraphDatabase.driver(
             uri,
-            auth=basic_auth(username, password)
+            auth=basic_auth(username, password),
+            # 抑制属性键不存在的 UNRECOGNIZED 分类通知警告（如 01N52）
+            # last_access_time 等属性在节点被检索命中后才写入，
+            # activation_value 在新节点创建后可能尚未被计算，
+            # 全新数据库或清空数据后这些属性键不存在是正常业务行为
+            notifications_disabled_classifications=["UNRECOGNIZED"],
         )
 
     async def close(self):
@@ -77,11 +85,11 @@ class Neo4jConnector:
         """
         await self.driver.close()
 
-    async def execute_query(self, query: str, json_format=False, **kwargs: Any) -> List[Dict[str, Any]]:
+    async def execute_query(self, cypher: str, json_format=False, **kwargs: Any) -> List[Dict[str, Any]]:
         """执行Cypher查询
         
         Args:
-            query: Cypher查询语句
+            cypher: Cypher查询语句
             json_format: json格式化
             **kwargs: 查询参数，将作为参数传递给Cypher查询
             
@@ -92,7 +100,7 @@ class Neo4jConnector:
 
         """
         result = await self.driver.execute_query(
-            query,
+            cypher,
             database="neo4j",
             **kwargs
         )

api/app/repositories/tool_repository.py

@@ -161,6 +161,17 @@ class BuiltinToolRepository:
             BuiltinToolConfig.id == tool_id
         ).first()
 
+    @staticmethod
+    def get_existing_tool_classes(db: Session, tenant_id: uuid.UUID) -> set:
+        """获取该租户已有的内置工具 tool_class 集合"""
+        rows = db.query(BuiltinToolConfig.tool_class).join(
+            ToolConfig, BuiltinToolConfig.id == ToolConfig.id
+        ).filter(
+            ToolConfig.tenant_id == tenant_id,
+            ToolConfig.tool_type == ToolType.BUILTIN.value
+        ).all()
+        return {row[0] for row in rows}
+
 
 class CustomToolRepository:
     """自定义工具仓储类"""

api/app/repositories/user_repository.py

@@ -23,7 +23,7 @@ class UserRepository:
         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()
+            user = self.db.query(User).options(joinedload(User.tenant)).filter(User.id == user_id, User.is_active.is_(True)).first()
             if user:
                 # 检查租户状态，租户禁用时返回 None
                 if user.tenant and not user.tenant.is_active:
@@ -297,6 +297,10 @@ def get_user_by_id(db: Session, user_id: uuid.UUID) -> Optional[User]:
     """根据ID获取用户"""
     return UserRepository(db).get_user_by_id(user_id)
 
+def get_user_by_id_regardless_active(db: Session, user_id: uuid.UUID) -> Optional[User]:
+    """根据ID获取用户（不过滤 is_active，用于启用/禁用场景）"""
+    return db.query(User).filter(User.id == user_id).first()
+
 def get_user_by_email(db: Session, email: str) -> Optional[User]:
     """根据邮箱获取用户"""
     return UserRepository(db).get_user_by_email(email)

api/app/schemas/app_schema.py

@@ -44,6 +44,8 @@ class FileInput(BaseModel):
     upload_file_id: Optional[uuid.UUID] = Field(None, description="已上传文件ID（local_file时必填）")
     url: Optional[str] = Field(None, description="远程URL（remote_url时必填）")
     file_type: Optional[str] = Field(None, description="具体文件格式（如image/jpg、audio/wav、document/docx、video/mp4）")
+    name: Optional[str] = Field(None, description="文件名")
+    size: Optional[int] = Field(None, description="文件大小（字节）")
 
     _content = None
 
@@ -616,6 +618,7 @@ class AppChatRequest(BaseModel):
     stream: bool = Field(default=False, description="是否流式返回")
     thinking: bool = Field(default=False, description="是否启用深度思考（需Agent配置支持）")
     files: List[FileInput] = Field(default_factory=list, description="附件列表（支持多文件）")
+    version: Optional[uuid.UUID] = Field(default=None, description="指定发布版本ID，不传则使用当前生效版本")
 
 
 class DraftRunRequest(BaseModel):

api/app/services/app_chat_service.py

@@ -26,6 +26,7 @@ from app.services.model_service import ModelApiKeyService
 from app.services.multi_agent_orchestrator import MultiAgentOrchestrator
 from app.services.multimodal_service import MultimodalService
 from app.services.workflow_service import WorkflowService
+from app.models.file_metadata_model import FileMetadata
 
 logger = get_business_logger()
 
@@ -165,7 +166,14 @@ class AppChatService:
             multimodal_service = MultimodalService(self.db, model_info)
             processed_files = await multimodal_service.process_files(files)
             logger.info(f"处理了 {len(processed_files)} 个文件")
-
+        # 为需要运行时上下文的工具注入上下文
+        for t in tools:
+            if hasattr(t, 'tool_instance') and hasattr(t.tool_instance, 'set_runtime_context'):
+                t.tool_instance.set_runtime_context(
+                    user_id=user_id or "anonymous",
+                    conversation_id=str(conversation_id) if conversation_id else None,
+                    uploaded_files=processed_files or []
+                )
         # 调用 Agent（支持多模态）
         result = await agent.chat(
             message=message,
@@ -211,11 +219,29 @@ class AppChatService:
             "reasoning_content": result.get("reasoning_content")
         }
         if files:
+            local_ids = [f.upload_file_id for f in files
+                         if f.transfer_method.value == "local_file" and f.upload_file_id
+                         and (not f.name or not f.size)]
+            meta_map = {}
+            if local_ids:
+                rows = self.db.query(FileMetadata).filter(
+                    FileMetadata.id.in_(local_ids),
+                    FileMetadata.status == "completed"
+                ).all()
+                meta_map = {str(r.id): r for r in rows}
             for f in files:
-                # url = await MultimodalService(self.db).get_file_url(f)
+                name, size = f.name, f.size
+                if f.transfer_method.value == "local_file" and f.upload_file_id and (not name or not size):
+                    meta = meta_map.get(str(f.upload_file_id))
+                    if meta:
+                        name = name or meta.file_name
+                        size = size or meta.file_size
                 human_meta["files"].append({
                     "type": f.type,
-                    "url": f.url
+                    "url": f.url,
+                    "name": name,
+                    "size": size,
+                    "file_type": f.file_type,
                 })
 
         if processed_files:
@@ -413,6 +439,15 @@ class AppChatService:
                 processed_files = await multimodal_service.process_files(files)
                 logger.info(f"处理了 {len(processed_files)} 个文件")
 
+            # 为需要运行时上下文的工具注入上下文
+            for t in tools:
+                if hasattr(t, 'tool_instance') and hasattr(t.tool_instance, 'set_runtime_context'):
+                    t.tool_instance.set_runtime_context(
+                        user_id=user_id or "anonymous",
+                        conversation_id=str(conversation_id) if conversation_id else None,
+                        uploaded_files=processed_files or []
+                    )
+
             # 流式调用 Agent（支持多模态），同时并行启动 TTS
             full_content = ""
             full_reasoning = ""
@@ -493,10 +528,29 @@ class AppChatService:
             }
 
             if files:
+                local_ids = [f.upload_file_id for f in files
+                             if f.transfer_method.value == "local_file" and f.upload_file_id
+                             and (not f.name or not f.size)]
+                meta_map = {}
+                if local_ids:
+                    rows = self.db.query(FileMetadata).filter(
+                        FileMetadata.id.in_(local_ids),
+                        FileMetadata.status == "completed"
+                    ).all()
+                    meta_map = {str(r.id): r for r in rows}
                 for f in files:
+                    name, size = f.name, f.size
+                    if f.transfer_method.value == "local_file" and f.upload_file_id and (not name or not size):
+                        meta = meta_map.get(str(f.upload_file_id))
+                        if meta:
+                            name = name or meta.file_name
+                            size = size or meta.file_size
                     human_meta["files"].append({
                         "type": f.type,
-                        "url": f.url
+                        "url": f.url,
+                        "name": name,
+                        "size": size,
+                        "file_type": f.file_type,
                     })
             if processed_files:
                 human_meta["history_files"] = {

api/app/services/app_dsl_service.py

@@ -73,15 +73,14 @@ class AppDslService:
             AppType.MULTI_AGENT: "multi_agent_config",
             AppType.WORKFLOW: "workflow"
         }.get(app.type, "config")
-        config_data = self._enrich_release_config(app.type, release.config or {})
+        config_data = self._enrich_release_config(app.type, release.config or {}, release.default_model_config_id)
         dsl = {**meta, "app": app_meta, config_key: config_data}
         return yaml.dump(dsl, default_flow_style=False, allow_unicode=True), f"{release.name}_v{release.version_name}.yaml"
 
-    def _enrich_release_config(self, app_type: str, cfg: dict) -> dict:
+    def _enrich_release_config(self, app_type: str, cfg: dict, default_model_config_id=None) -> dict:
         if app_type == AppType.AGENT:
             enriched = {**cfg}
-            if "default_model_config_id" in cfg:
-                enriched["default_model_config_ref"] = self._model_ref(cfg["default_model_config_id"])
+            enriched["default_model_config_ref"] = self._model_ref(default_model_config_id)
             if "knowledge_retrieval" in cfg:
                 enriched["knowledge_retrieval"] = self._enrich_knowledge_retrieval(cfg["knowledge_retrieval"])
             if "tools" in cfg:
@@ -91,8 +90,7 @@ class AppDslService:
             return enriched
         if app_type == AppType.MULTI_AGENT:
             enriched = {**cfg}
-            if "default_model_config_id" in cfg:
-                enriched["default_model_config_ref"] = self._model_ref(cfg["default_model_config_id"])
+            enriched["default_model_config_ref"] = self._model_ref(default_model_config_id)
             if "master_agent_id" in cfg:
                 enriched["master_agent_ref"] = self._release_ref(cfg["master_agent_id"])
             if "sub_agents" in cfg:
@@ -229,8 +227,11 @@ class AppDslService:
         workspace_id: uuid.UUID,
         tenant_id: uuid.UUID,
         user_id: uuid.UUID,
+        app_id: Optional[uuid.UUID] = None,
     ) -> tuple[App, list[str]]:
-        """解析 DSL，创建应用及配置，返回 (new_app, warnings)"""
+        """解析 DSL，创建或覆盖应用配置，返回 (app, warnings)。
+        app_id 不为空时：校验类型一致后覆盖配置；为空时创建新应用。
+        """
         app_meta = dsl.get("app", {})
         app_type = app_meta.get("type")
         if app_type not in (AppType.AGENT, AppType.MULTI_AGENT, AppType.WORKFLOW):
@@ -239,6 +240,9 @@ class AppDslService:
         warnings: list[str] = []
         now = datetime.datetime.now()
 
+        if app_id is not None:
+            return self._overwrite_dsl(dsl, app_id, app_type, workspace_id, tenant_id, warnings, now)
+
         new_app = App(
             id=uuid.uuid4(),
             workspace_id=workspace_id,
@@ -258,11 +262,57 @@ class AppDslService:
         self.db.add(new_app)
         self.db.flush()
 
+        self._write_config(new_app.id, app_type, dsl, workspace_id, tenant_id, warnings, now, create=True)
+
+        self.db.commit()
+        self.db.refresh(new_app)
+        return new_app, warnings
+
+    def _overwrite_dsl(
+        self,
+        dsl: dict,
+        app_id: uuid.UUID,
+        app_type: str,
+        workspace_id: uuid.UUID,
+        tenant_id: uuid.UUID,
+        warnings: list,
+        now: datetime.datetime,
+    ) -> tuple[App, list[str]]:
+        """覆盖已有应用的配置，类型不一致时抛出异常"""
+        app = self.db.query(App).filter(
+            App.id == app_id,
+            App.workspace_id == workspace_id,
+            App.is_active.is_(True)
+        ).first()
+        if not app:
+            raise ResourceNotFoundException("应用", str(app_id))
+        if app.type != app_type:
+            raise BusinessException(
+                f"YAML 类型 '{app_type}' 与应用类型 '{app.type}' 不一致，无法导入",
+                BizCode.BAD_REQUEST
+            )
+
+        self._write_config(app_id, app_type, dsl, workspace_id, tenant_id, warnings, now, create=False)
+
+        self.db.commit()
+        self.db.refresh(app)
+        return app, warnings
+
+    def _write_config(
+        self,
+        app_id: uuid.UUID,
+        app_type: str,
+        dsl: dict,
+        workspace_id: uuid.UUID,
+        tenant_id: uuid.UUID,
+        warnings: list,
+        now: datetime.datetime,
+        create: bool,
+    ) -> None:
+        """写入（新建或覆盖）应用配置"""
         if app_type == AppType.AGENT:
             cfg = dsl.get("agent_config") or {}
-            self.db.add(AgentConfig(
-                id=uuid.uuid4(),
-                app_id=new_app.id,
+            fields = dict(
                 system_prompt=cfg.get("system_prompt"),
                 model_parameters=cfg.get("model_parameters"),
                 default_model_config_id=self._resolve_model(cfg.get("default_model_config_ref"), tenant_id, warnings),
@@ -272,16 +322,21 @@ class AppDslService:
                 tools=self._resolve_tools(cfg.get("tools", []), tenant_id, warnings),
                 skills=self._resolve_skills(cfg.get("skills", {}), tenant_id, warnings),
                 features=cfg.get("features", {}),
-                is_active=True,
-                created_at=now,
                 updated_at=now,
-            ))
+            )
+            if create:
+                self.db.add(AgentConfig(id=uuid.uuid4(), app_id=app_id, is_active=True, created_at=now, **fields))
+            else:
+                existing = self.db.query(AgentConfig).filter(AgentConfig.app_id == app_id).first()
+                if existing:
+                    for k, v in fields.items():
+                        setattr(existing, k, v)
+                else:
+                    self.db.add(AgentConfig(id=uuid.uuid4(), app_id=app_id, is_active=True, created_at=now, **fields))
 
         elif app_type == AppType.MULTI_AGENT:
             cfg = dsl.get("multi_agent_config") or {}
-            self.db.add(MultiAgentConfig(
-                id=uuid.uuid4(),
-                app_id=new_app.id,
+            fields = dict(
                 orchestration_mode=cfg.get("orchestration_mode", "collaboration"),
                 master_agent_name=cfg.get("master_agent_name"),
                 model_parameters=cfg.get("model_parameters"),
@@ -291,10 +346,17 @@ class AppDslService:
                 routing_rules=self._resolve_routing_rules(cfg.get("routing_rules"), warnings),
                 execution_config=cfg.get("execution_config", {}),
                 aggregation_strategy=cfg.get("aggregation_strategy", "merge"),
-                is_active=True,
-                created_at=now,
                 updated_at=now,
-            ))
+            )
+            if create:
+                self.db.add(MultiAgentConfig(id=uuid.uuid4(), app_id=app_id, is_active=True, created_at=now, **fields))
+            else:
+                existing = self.db.query(MultiAgentConfig).filter(MultiAgentConfig.app_id == app_id).first()
+                if existing:
+                    for k, v in fields.items():
+                        setattr(existing, k, v)
+                else:
+                    self.db.add(MultiAgentConfig(id=uuid.uuid4(), app_id=app_id, is_active=True, created_at=now, **fields))
 
         elif app_type == AppType.WORKFLOW:
             adapter = MemoryBearAdapter(dsl)
@@ -306,20 +368,39 @@ class AppDslService:
             for w in result.warnings:
                 warnings.append(f"[节点警告] {w.node_name or w.node_id}: {w.detail}")
             wf = dsl.get("workflow") or {}
-            WorkflowService(self.db).create_workflow_config(
-                app_id=new_app.id,
-                nodes=[n.model_dump() for n in result.nodes],
-                edges=[e.model_dump() for e in result.edges],
-                variables=[v.model_dump() for v in result.variables],
-                execution_config=wf.get("execution_config", {}),
-                features=wf.get("features", {}),
-                triggers=wf.get("triggers", []),
-                validate=False,
-            )
-
-        self.db.commit()
-        self.db.refresh(new_app)
-        return new_app, warnings
+            wf_service = WorkflowService(self.db)
+            if create:
+                wf_service.create_workflow_config(
+                    app_id=app_id,
+                    nodes=[n.model_dump() for n in result.nodes],
+                    edges=[e.model_dump() for e in result.edges],
+                    variables=[v.model_dump() for v in result.variables],
+                    execution_config=wf.get("execution_config", {}),
+                    features=wf.get("features", {}),
+                    triggers=wf.get("triggers", []),
+                    validate=False,
+                )
+            else:
+                existing = self.db.query(WorkflowConfig).filter(WorkflowConfig.app_id == app_id).first()
+                if existing:
+                    existing.nodes = [n.model_dump() for n in result.nodes]
+                    existing.edges = [e.model_dump() for e in result.edges]
+                    existing.variables = [v.model_dump() for v in result.variables]
+                    existing.execution_config = wf.get("execution_config", {})
+                    existing.features = wf.get("features", {})
+                    existing.triggers = wf.get("triggers", [])
+                    existing.updated_at = now
+                else:
+                    wf_service.create_workflow_config(
+                        app_id=app_id,
+                        nodes=[n.model_dump() for n in result.nodes],
+                        edges=[e.model_dump() for e in result.edges],
+                        variables=[v.model_dump() for v in result.variables],
+                        execution_config=wf.get("execution_config", {}),
+                        features=wf.get("features", {}),
+                        triggers=wf.get("triggers", []),
+                        validate=False,
+                    )
 
     def _unique_app_name(self, name: str, workspace_id: uuid.UUID, app_type: AppType) -> str:
         """生成唯一应用名称，同时检查本空间自有应用和共享到本空间的应用"""

api/app/services/app_service.py

@@ -411,6 +411,7 @@ class AppService:
             edges=[edge.model_dump() for edge in data.edges] if data.edges else [],
             variables=[var.model_dump() for var in data.variables] if data.variables else [],
             execution_config=data.execution_config.model_dump() if data.execution_config else {},
+            features=data.features if data.features else {},
             triggers=[trigger.model_dump() for trigger in data.triggers] if data.triggers else [],
             is_active=True,
             created_at=now,
@@ -619,6 +620,28 @@ class AppService:
         self._validate_app_accessible(app, workspace_id)
         return app
 
+    def get_release_by_id(self, app_id: uuid.UUID, release_id: uuid.UUID) -> AppRelease:
+        """按发布版本ID获取发布快照
+
+        Args:
+            app_id: 应用ID
+            release_id: 发布版本ID
+
+        Returns:
+            AppRelease: 发布快照
+
+        Raises:
+            BusinessException: 版本不存在或已下线
+        """
+        from app.repositories.app_repository import get_release_by_id
+        release = get_release_by_id(self.db, app_id, release_id)
+        if not release:
+            raise BusinessException(
+                f"版本 {release_id} 不存在或已下线",
+                BizCode.RELEASE_NOT_FOUND,
+            )
+        return release
+
     def create_app(
             self,
             *,

api/app/services/draft_run_service.py

@@ -640,7 +640,14 @@ class AgentRunService:
                 multimodal_service = MultimodalService(self.db, model_info)
                 processed_files = await multimodal_service.process_files(files)
                 logger.info(f"处理了 {len(processed_files)} 个文件，provider={provider}")
-
+            # 为需要运行时上下文的工具注入上下文
+            for t in tools:
+                if hasattr(t, 'tool_instance') and hasattr(t.tool_instance, 'set_runtime_context'):
+                    t.tool_instance.set_runtime_context(
+                        user_id=user_id or "anonymous",
+                        conversation_id=str(conversation_id) if conversation_id else None,
+                        uploaded_files=processed_files or []
+                    )
             # 7. 知识库检索
             context = None
 
@@ -890,7 +897,14 @@ class AgentRunService:
                 multimodal_service = MultimodalService(self.db, model_info)
                 processed_files = await multimodal_service.process_files(files)
                 logger.info(f"处理了 {len(processed_files)} 个文件，provider={provider}")
-
+            # 为需要运行时上下文的工具注入上下文
+            for t in tools:
+                if hasattr(t, 'tool_instance') and hasattr(t.tool_instance, 'set_runtime_context'):
+                    t.tool_instance.set_runtime_context(
+                        user_id=user_id or "anonymous",
+                        conversation_id=str(conversation_id) if conversation_id else None,
+                        uploaded_files=processed_files or []
+                    )
             # 7. 知识库检索
             context = None
 
@@ -1285,10 +1299,30 @@ class AgentRunService:
                 "history_files": {}
             }
             if files:
+                from app.models.file_metadata_model import FileMetadata
+                local_ids = [f.upload_file_id for f in files
+                             if f.transfer_method.value == "local_file" and f.upload_file_id
+                             and (not f.name or not f.size)]
+                meta_map = {}
+                if local_ids:
+                    rows = self.db.query(FileMetadata).filter(
+                        FileMetadata.id.in_(local_ids),
+                        FileMetadata.status == "completed"
+                    ).all()
+                    meta_map = {str(r.id): r for r in rows}
                 for f in files:
+                    name, size = f.name, f.size
+                    if f.transfer_method.value == "local_file" and f.upload_file_id and (not name or not size):
+                        meta = meta_map.get(str(f.upload_file_id))
+                        if meta:
+                            name = name or meta.file_name
+                            size = size or meta.file_size
                     human_meta["files"].append({
                         "type": f.type,
-                        "url": f.url
+                        "url": f.url,
+                        "file_type": f.file_type,
+                        "name": name,
+                        "size": size
                     })
 
             # 保存 history_files，包含 provider 和 is_omni 信息

api/app/services/emotion_analytics_service.py

@@ -679,9 +679,9 @@ class EmotionAnalyticsService:
 
             # 查询用户的实体和标签
             query = """
-            MATCH (e:Entity)
+            MATCH (e:ExtractedEntity)
             WHERE e.end_user_id = $end_user_id
-            RETURN e.name as name, e.type as type
+            RETURN e.name as name, e.entity_type as type
             ORDER BY e.created_at DESC
             LIMIT 20
             """

api/app/services/implicit_memory_service.py

@@ -34,6 +34,7 @@ from app.schemas.implicit_memory_schema import (
     UserMemorySummary,
 )
 from app.schemas.memory_config_schema import MemoryConfig
+from app.services.memory_base_service import MIN_MEMORY_SUMMARY_COUNT
 from sqlalchemy.orm import Session
 
 logger = logging.getLogger(__name__)
@@ -379,12 +380,59 @@ class ImplicitMemoryService:
             raise
     
 
+    def _build_empty_profile(self) -> dict:
+        """构建 MemorySummary 不足时返回的固定空白画像数据"""
+        now_ms = int(datetime.utcnow().timestamp() * 1000)
+        insufficient = "Insufficient data for analysis"
+
+        def _empty_dimension(name: str) -> dict:
+            return {
+                "evidence": [insufficient],
+                "reasoning": f"No clear evidence found for {name} dimension",
+                "percentage": 0.0,
+                "dimension_name": name,
+                "confidence_level": 20,
+            }
+
+        def _empty_category(name: str) -> dict:
+            return {
+                "evidence": [insufficient],
+                "percentage": 25.0,
+                "category_name": name,
+                "trending_direction": None,
+            }
+
+        return {
+            "habits": [],
+            "portrait": {
+                "aesthetic": _empty_dimension("aesthetic"),
+                "creativity": _empty_dimension("creativity"),
+                "literature": _empty_dimension("literature"),
+                "technology": _empty_dimension("technology"),
+                "historical_trends": None,
+                "analysis_timestamp": now_ms,
+                "total_summaries_analyzed": 0,
+            },
+            "preferences": [],
+            "interest_areas": {
+                "art": _empty_category("art"),
+                "tech": _empty_category("tech"),
+                "music": _empty_category("music"),
+                "lifestyle": _empty_category("lifestyle"),
+                "analysis_timestamp": now_ms,
+                "total_summaries_analyzed": 0,
+            },
+        }
+
     async def generate_complete_profile(
         self,
         user_id: str
     ) -> dict:
         """生成完整的用户画像（包含所有4个模块）
         
+        需要该用户的 MemorySummary 节点数量 >= 5 才会真正调用 LLM 生成画像，
+        否则返回固定的空白画像数据。
+        
         Args:
             user_id: 用户ID
             
@@ -394,6 +442,16 @@ class ImplicitMemoryService:
         logger.info(f"生成完整用户画像: user={user_id}")
         
         try:
+            # 前置检查：查询该用户有效的 MemorySummary 节点数量（排除孤立节点）
+            from app.services.memory_base_service import MemoryBaseService
+            base_service = MemoryBaseService()
+            memory_summary_count = await base_service.get_valid_memory_summary_count(user_id)
+            logger.info(f"用户 MemorySummary 节点数量: {memory_summary_count} (user={user_id})")
+
+            if memory_summary_count < MIN_MEMORY_SUMMARY_COUNT:
+                logger.info(f"MemorySummary 数量不足 {MIN_MEMORY_SUMMARY_COUNT}（当前 {memory_summary_count}），返回空白画像: user={user_id}")
+                return self._build_empty_profile()
+
             # 并行调用4个分析方法
             preferences, portrait, interest_areas, habits = await asyncio.gather(
                 self.get_preference_tags(user_id=user_id),

api/app/services/memory_base_service.py

@@ -265,12 +265,50 @@ async def Translation_English(modid, text, fields=None):
     # 其他类型（数字、布尔值、None等）：原样返回
     else:
         return text
+# 隐性记忆画像生成所需的最低 MemorySummary 节点数量
+MIN_MEMORY_SUMMARY_COUNT = 5
+
+
 class MemoryBaseService:
     """记忆服务基类，提供共享的辅助方法"""
     
     def __init__(self):
         self.neo4j_connector = Neo4jConnector()
     
+    async def get_valid_memory_summary_count(
+        self,
+        end_user_id: str
+    ) -> int:
+        """获取用户有效的 MemorySummary 节点数量（排除孤立节点）。
+
+        只统计存在 DERIVED_FROM_STATEMENT 关系的 MemorySummary 节点。
+
+        Args:
+            end_user_id: 终端用户ID
+
+        Returns:
+            有效 MemorySummary 节点数量
+        """
+        try:
+            query = """
+            MATCH (n:MemorySummary)-[:DERIVED_FROM_STATEMENT]->(:Statement)
+            WHERE n.end_user_id = $end_user_id
+            RETURN count(DISTINCT n) as count
+            """
+            result = await self.neo4j_connector.execute_query(
+                query, end_user_id=end_user_id
+            )
+            count = result[0]["count"] if result and len(result) > 0 else 0
+            logger.debug(
+                f"有效 MemorySummary 节点数量: {count} (end_user_id={end_user_id})"
+            )
+            return count
+        except Exception as e:
+            logger.error(
+                f"获取有效 MemorySummary 数量失败: {str(e)}", exc_info=True
+            )
+            return 0
+    
     @staticmethod
     def parse_timestamp(timestamp_value) -> Optional[int]:
         """

api/app/services/memory_dashboard_service.py

@@ -803,7 +803,6 @@ def get_rag_content(
                 "page": {
                     "page": page,
                     "pagesize": pagesize,
-                    "total": 0,
                     "hasnext": False,
                 },
                 "items": []
@@ -897,13 +896,12 @@ def get_rag_content(
             "page": {
                 "page": page,
                 "pagesize": pagesize,
-                "total": global_total,
                 "hasnext": offset_end < global_total,
             },
             "items": conversations
         }
         
-        business_logger.info(f"成功获取RAG内容: total={global_total}, page={page}, 返回={len(conversations)} 条对话")
+        business_logger.info(f"成功获取RAG内容: page={page}, 返回={len(conversations)} 条对话")
         return result
         
     except Exception as e:

api/app/services/prompt_optimizer_service.py

@@ -227,10 +227,20 @@ class PromptOptimizerService:
             content = getattr(chunk, "content", chunk)
             if not content:
                 continue
-            buffer += content
+            if isinstance(content, str):
+                buffer += content
+            elif isinstance(content, list):
+                for _ in content:
+                    buffer += _["text"]
+            else:
+                logger.error(f"Unsupported content type - {content}")
+                raise Exception("Unsupported content type")
             cache = buffer[:-20]
+            last_idx = 19
+            while cache and cache[-1] == '\\' and last_idx > 0:
+                cache = buffer[:-last_idx]
+                last_idx -= 1
 
-            # 尝试找到 "prompt": " 开始位置
             if prompt_finished:
                 continue
 
@@ -272,7 +282,7 @@ class PromptOptimizerService:
     def parser_prompt_variables(prompt: str):
         try:
             pattern = r'\{\{\s*([a-zA-Z_][a-zA-Z0-9_]*)\s*\}\}'
-            matches = re.findall(pattern, prompt)
+            matches = re.findall(pattern, str(prompt))
             variables = list(set(matches))
             return variables
         except Exception as e:

api/app/services/tool_service.py

@@ -34,7 +34,8 @@ BUILTIN_TOOLS = {
     "JsonTool": "app.core.tools.builtin.json_tool",
     "BaiduSearchTool": "app.core.tools.builtin.baidu_search_tool",
     "MinerUTool": "app.core.tools.builtin.mineru_tool",
-    "TextInTool": "app.core.tools.builtin.textin_tool"
+    "TextInTool": "app.core.tools.builtin.textin_tool",
+    "OpenClawTool": "app.core.tools.builtin.openclaw_tool",
 }
 
 
@@ -340,18 +341,18 @@ class ToolService:
             return {"success": False, "message": f"测试失败: {str(e)}"}
 
     def ensure_builtin_tools_initialized(self, tenant_id: uuid.UUID):
-        """确保内置工具已初始化"""
-        existing = self.tool_repo.exists_builtin_for_tenant(self.db, tenant_id)
-
-        if existing:
+        """确保内置工具已初始化（支持增量补充新工具）"""
+        builtin_config = self._load_builtin_config()
+        if not builtin_config:
             return
 
-        # 从配置文件加载内置工具定义
-        builtin_config = self._load_builtin_config()
+        existing_classes = self.builtin_repo.get_existing_tool_classes(self.db, tenant_id)
 
+        added = False
         for tool_key, tool_info in builtin_config.items():
+            if tool_info['tool_class'] in existing_classes:
+                continue
             try:
-                # 创建工具配置
                 initial_status = self._determine_initial_status(tool_info)
                 tool_config = ToolConfig(
                     name=tool_info['name'],
@@ -367,7 +368,6 @@ class ToolService:
                 self.db.add(tool_config)
                 self.db.flush()
 
-                # 创建内置工具配置
                 builtin_config_obj = BuiltinToolConfig(
                     id=tool_config.id,
                     tool_class=tool_info['tool_class'],
@@ -375,12 +375,14 @@ class ToolService:
                     requires_config=tool_info.get('requires_config', False)
                 )
                 self.db.add(builtin_config_obj)
+                added = True
 
             except Exception as e:
                 logger.error(f"初始化内置工具失败: {tool_key}, {e}")
 
-        self.db.commit()
-        logger.info(f"租户 {tenant_id} 内置工具初始化完成")
+        if added:
+            self.db.commit()
+            logger.info(f"租户 {tenant_id} 内置工具增量初始化完成")
 
     async def get_tool_methods(self, tool_id: str, tenant_id: uuid.UUID) -> Optional[List[Dict[str, Any]]]:
         """获取工具的所有方法
@@ -458,6 +460,9 @@ class ToolService:
         # 对于json_tool，根据操作类型返回相关参数
         elif hasattr(tool_instance, 'name') and tool_instance.name == 'json_tool':
             return self._get_json_tool_params(operation)
+        # 对于openclaw_tool，根据操作类型返回不同描述的参数
+        elif hasattr(tool_instance, 'name') and tool_instance.name == 'openclaw_tool':
+            return self._get_openclaw_tool_params(operation)
         
         # 其他工具的默认处理：返回除operation外的所有参数
         return [{
@@ -574,6 +579,29 @@ class ToolService:
                     "default": "Asia/Shanghai"
                 }
             ]
+        elif operation == "datetime_to_timestamp":
+            return [
+                {
+                    "name": "input_value",
+                    "type": "string",
+                    "description": "输入值（时间字符串，如：2026-04-07 10:30:25）",
+                    "required": True
+                },
+                {
+                    "name": "input_format",
+                    "type": "string",
+                    "description": "输入时间格式（如：%Y-%m-%d %H:%M:%S）",
+                    "required": False,
+                    "default": "%Y-%m-%d %H:%M:%S"
+                },
+                {
+                    "name": "from_timezone",
+                    "type": "string",
+                    "description": "源时区（如：UTC, Asia/Shanghai）",
+                    "required": False,
+                    "default": "Asia/Shanghai"
+                }
+            ]
         else:
             # 默认返回所有参数（除了operation）
             return [
@@ -687,6 +715,65 @@ class ToolService:
         
         return base_params
 
+    @staticmethod
+    def _get_openclaw_tool_params(operation: str) -> List[Dict[str, Any]]:
+        """获取 openclaw_tool 特定操作的参数"""
+        if operation == "print_task":
+            return [
+                {
+                    "name": "message",
+                    "type": "string",
+                    "description": "发送给 OpenClaw 的打印任务描述，将用户的原始消息原封不动地传递给 OpenClaw，禁止改写、补充或润色用户的原文",
+                    "required": True
+                },
+                {
+                    "name": "image_url",
+                    "type": "string",
+                    "description": "可选，附带的设计图片或参考图，OpenClaw 可据此生成 3D 模型",
+                    "required": False
+                }
+            ]
+        elif operation == "device_query":
+            return [
+                {
+                    "name": "message",
+                    "type": "string",
+                    "description": "发送给 OpenClaw 的设备查询指令",
+                    "required": True
+                }
+            ]
+        elif operation == "image_understand":
+            return [
+                {
+                    "name": "message",
+                    "type": "string",
+                    "description": "发送给 OpenClaw 的图片理解任务，应描述需要对图片做什么（如描述内容、提取文字、分析信息）",
+                    "required": True
+                },
+                {
+                    "name": "image_url",
+                    "type": "string",
+                    "description": "要分析的图片 URL 或 base64 data URI",
+                    "required": False
+                }
+            ]
+        else:
+            # general 及其他
+            return [
+                {
+                    "name": "message",
+                    "type": "string",
+                    "description": "发送给 OpenClaw Agent 的任务描述，应包含完整的任务需求",
+                    "required": True
+                },
+                {
+                    "name": "image_url",
+                    "type": "string",
+                    "description": "可选，附带的图片 URL 或 base64 data URI",
+                    "required": False
+                }
+            ]
+
     async def _get_custom_tool_methods(self, config: ToolConfig) -> List[Dict[str, Any]]:
         """获取自定义工具的方法"""
         custom_config = self.custom_repo.find_by_tool_id(self.db, config.id)

api/app/services/user_memory_service.py

@@ -14,6 +14,7 @@ from pydantic import BaseModel, Field
 from sqlalchemy.orm import Session
 
 from app.core.logging_config import get_logger
+from app.core.memory.storage_services.extraction_engine.deduplication.deduped_and_disamb import _USER_PLACEHOLDER_NAMES
 from app.core.memory.utils.llm.llm_utils import MemoryClientFactory
 from app.db import get_db_context
 from app.repositories.conversation_repository import ConversationRepository
@@ -21,7 +22,7 @@ from app.repositories.end_user_repository import EndUserRepository
 from app.repositories.neo4j.cypher_queries import Graph_Node_query
 from app.repositories.neo4j.neo4j_connector import Neo4jConnector
 from app.schemas.memory_episodic_schema import EmotionSubject, EmotionType, type_mapping
-from app.services.memory_base_service import MemoryBaseService
+from app.services.memory_base_service import MemoryBaseService, MIN_MEMORY_SUMMARY_COUNT
 from app.services.memory_config_service import MemoryConfigService
 from app.services.memory_perceptual_service import MemoryPerceptualService
 from app.services.memory_short_service import ShortService
@@ -398,12 +399,16 @@ class UserMemoryService:
                 }
             
             # 构建响应数据（转换时间为毫秒时间戳）
+            # 将 meta_data 中的 profile、knowledge_tags、behavioral_hints 平铺到顶层
+            meta = end_user_info_record.meta_data or {}
             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,
+                "profile": meta.get("profile"),
+                "knowledge_tags": meta.get("knowledge_tags"),
+                "behavioral_hints": meta.get("behavioral_hints"),
                 "created_at": datetime_to_timestamp(end_user_info_record.created_at),
                 "updated_at": datetime_to_timestamp(end_user_info_record.updated_at)
             }
@@ -473,7 +478,7 @@ class UserMemoryService:
             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'}
+            _user_placeholder_names = _USER_PLACEHOLDER_NAMES
             
             # 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:
@@ -1500,7 +1505,7 @@ async def analytics_memory_types(
     2. 工作记忆 (WORKING_MEMORY) = 会话数量（通过 ConversationRepository.get_conversation_by_user_id 获取）
     3. 短期记忆 (SHORT_TERM_MEMORY) = /short_term 接口返回的问答对数量
     4. 显性记忆 (EXPLICIT_MEMORY) = 情景记忆 + 语义记忆（通过 MemoryBaseService.get_explicit_memory_count 获取）
-    5. 隐性记忆 (IMPLICIT_MEMORY) = Statement 节点数量的三分之一
+    5. 隐性记忆 (IMPLICIT_MEMORY) = MemorySummary 节点数量（需 >= MIN_MEMORY_SUMMARY_COUNT 才显示，否则为 0）
     6. 情绪记忆 (EMOTIONAL_MEMORY) = 情绪标签统计总数（通过 MemoryBaseService.get_emotional_memory_count 获取）
     7. 情景记忆 (EPISODIC_MEMORY) = memory_summary（通过 MemoryBaseService.get_episodic_memory_count 获取）
     8. 遗忘记忆 (FORGET_MEMORY) = 激活值低于阈值的节点数（通过 MemoryBaseService.get_forget_memory_count 获取）
@@ -1557,23 +1562,15 @@ async def analytics_memory_types(
             logger.warning(f"获取会话数量失败，工作记忆数量设为0: {str(e)}")
             work_count = 0
     
-    # 获取隐性记忆数量（基于 Statement 节点数量的三分之一）
+    # 获取隐性记忆数量（基于有关联关系的 MemorySummary 节点数量，需 >= MIN_MEMORY_SUMMARY_COUNT 才计入）
     implicit_count = 0
     if end_user_id:
         try:
-            # 查询 Statement 节点数量
-            query = """
-            MATCH (n:Statement)
-            WHERE n.end_user_id = $end_user_id
-            RETURN count(n) as count
-            """
-            result = await _neo4j_connector.execute_query(query, end_user_id=end_user_id)
-            statement_count = result[0]["count"] if result and len(result) > 0 else 0
-            # 取三分之一作为隐性记忆数量
-            implicit_count = round(statement_count / 3)
-            logger.debug(f"隐性记忆数量（Statement数量的1/3）: {implicit_count} (Statement总数={statement_count}, end_user_id={end_user_id})")
+            memory_summary_count = await base_service.get_valid_memory_summary_count(end_user_id)
+            implicit_count = memory_summary_count if memory_summary_count >= MIN_MEMORY_SUMMARY_COUNT else 0
+            logger.debug(f"隐性记忆数量（有效MemorySummary节点数）: {implicit_count} (有效MemorySummary总数={memory_summary_count}, end_user_id={end_user_id})")
         except Exception as e:
-            logger.warning(f"获取Statement数量失败，隐性记忆数量设为0: {str(e)}")
+            logger.warning(f"获取MemorySummary数量失败，隐性记忆数量设为0: {str(e)}")
             implicit_count = 0
     
     # 原有的基于行为习惯的统计方式（已注释）
@@ -1639,7 +1636,7 @@ async def analytics_memory_types(
         "WORKING_MEMORY": work_count,                             # 工作记忆（基于会话数量）
         "SHORT_TERM_MEMORY": short_term_count,                    # 短期记忆（基于问答对数量）
         "EXPLICIT_MEMORY": explicit_count,                        # 显性记忆（情景记忆 + 语义记忆）
-        "IMPLICIT_MEMORY": implicit_count,                        # 隐性记忆（Statement数量的1/3）
+        "IMPLICIT_MEMORY": implicit_count,                        # 隐性记忆（MemorySummary节点数，需>=MIN_MEMORY_SUMMARY_COUNT）
         "EMOTIONAL_MEMORY": emotion_count,                        # 情绪记忆（使用情绪标签统计）
         "EPISODIC_MEMORY": episodic_count,                        # 情景记忆
         "FORGET_MEMORY": forget_count                             # 遗忘记忆（激活值低于阈值）

api/app/services/user_service.py

@@ -285,7 +285,7 @@ def activate_user(db: Session, user_id_to_activate: uuid.UUID, current_user: Use
     try:
         # 查找用户
         business_logger.debug(f"查找待激活用户: {user_id_to_activate}")
-        db_user = user_repository.get_user_by_id(db, user_id=user_id_to_activate)
+        db_user = user_repository.get_user_by_id_regardless_active(db, user_id=user_id_to_activate)
         if not db_user:
             business_logger.warning(f"用户不存在: {user_id_to_activate}")
             raise BusinessException("用户不存在", code=BizCode.USER_NOT_FOUND)

api/app/services/workflow_import_service.py

@@ -69,6 +69,7 @@ class WorkflowImportService:
             edges=workflow_config.edges,
             nodes=workflow_config.nodes,
             variables=workflow_config.variables,
+            features=workflow_config.features,
             warnings=workflow_config.warnings,
             errors=workflow_config.errors
         )
@@ -95,7 +96,8 @@ class WorkflowImportService:
                 workflow_config=WorkflowConfigCreate(
                     nodes=config["nodes"],
                     edges=config["edges"],
-                    variables=config["variables"]
+                    variables=config["variables"],
+                    features=config.get("features", {})
                 )
             )
         )

api/app/services/workflow_service.py

@@ -957,7 +957,10 @@ class WorkflowService:
                                     for file in message["content"]:
                                         human_meta["files"].append({
                                             "type": file.get("type"),
-                                            "url": file.get("url")
+                                            "url": file.get("url"),
+                                            "file_type": file.get("origin_file_type"),
+                                            "name": file.get("name"),
+                                            "size": file.get("size")
                                         })
                             if message["role"] == "assistant":
                                 assistant_message = message["content"]

api/app/version_info.json

@@ -1,4 +1,36 @@
 {
+  "v0.2.10": {
+    "introduction": {
+      "codeName": "炼剑",
+      "releaseDate": "2026-4-8",
+      "upgradePosition": "🐻 全面强化工作流引擎、引入 Agent 深度思考模式与多模态记忆读取，百炼成锋，剑指生产就绪",
+      "coreUpgrades": [
+        "1. 工作流引擎增强<br>* 会话变量文件格式支持：支持文件类型值及本地/远程默认值配置<br>* 列表操作节点：新增专用列表操作节点<br>* 模板转换支持 HTML：扩展富内容渲染能力<br>* 表单返回与提交：工作流返回交互式表单，前端支持提交<br>* HTTP 节点 XML 响应：拓宽企业级 API 集成兼容性<br>* 开场白与文件引用：支持配置开场白及附件引用<br>* 模板转换三级变量：支持深层嵌套变量访问<br>* 节点连线添加按钮：连线处新增内联添加按钮",
+        "2. Agent 智能 🧠<br>* Agent 深度思考模式：支持更充分的推理以产出高质量回答<br>* 模型深度思考特性开关：模型级特性标识与应用级开关控制",
+        "3. 记忆系统升级 📚<br>* 用户记忆库分页：支持大规模记忆集合分页浏览<br>* RAG 用户记忆数据结构刷新：后端 API 数据结构重新设计<br>* 多模态记忆读取：支持检索图像、音频等非文本记忆<br>* 语义剪枝阈值提示文案：显示描述性区间标签",
+        "4. 前端与体验 🎨<br>* 技能工具删除状态展示：工具列表显示删除状态标识<br>* 仪表盘日环比数据：关键指标增加与昨日对比数据",
+        "5. 稳健性与缺陷修复 🔧<br>* 参数提取空值处理：优雅处理缺失数据<br>* Token 消耗展示优化：确保用量报告准确<br>* 模型参数负值修复：明确参数范围定义<br>* 应用共享删除同步：正确更新所有共享记录<br>* 记忆写入任务排序：按时间戳顺序执行<br>* 多模态模型缺失优雅处理：不再中断感知记忆写入<br>* 自定义工具 Number 变量传递：解决类型转换问题<br>* 集群子代理保存后显示：修复未反显问题<br>* 记忆开启后流式输出修复：解决字符串序列化问题",
+        "<br>",
+        "v0.2.10 标志着平台向生产成熟度迈出的重要一步。深度思考、交互式表单工作流与多模态记忆的结合展现了平台从记忆存储向综合认知基础设施的演进。我们期待 4 月 17 日 v0.3.0 发布会，届时将带来更深层的 Agent 推理能力、多智能体协作功能及记忆智能管线的进一步优化。剑已炼成，只待出鞘。",
+        "MemoryBear — 百炼成锋 🐻✨"
+      ]
+    },
+    "introduction_en": {
+      "codeName": "LianJian",
+      "releaseDate": "2026-4-8",
+      "upgradePosition": "🐻 Comprehensive workflow engine enhancements, Agent deep thinking mode, and multimodal memory reading — forging the blade for production readiness",
+      "coreUpgrades": [
+        "1. Workflow Engine Enhancements<br>* Session Variable File Support: File-type values with local/remote defaults<br>* List Operation Node: Dedicated node for array manipulation<br>* Template Conversion HTML Support: Rich-content rendering<br>* Form Return & Submission: Interactive forms in workflow conversations<br>* HTTP Node XML Response: Enterprise API integration compatibility<br>* Opening Remarks & File References: Configurable conversation openers<br>* Template Conversion Three-Level Variables: Deep nested variable access<br>* Node Connection Add Button: Inline add button on connections",
+        "2. Agent Intelligence 🧠<br>* Agent Deep Thinking Mode: Thorough reasoning for complex queries<br>* Model Deep Thinking Feature Toggle: Model-level flag with per-app control",
+        "3. Memory System Upgrades 📚<br>* User Memory Pagination: Paginated browsing for large collections<br>* RAG User Memory Data Structure Refresh: Redesigned backend API contracts<br>* Multimodal Memory Reading: Retrieval of image, audio, and non-text memory<br>* Semantic Pruning Threshold Hints: Descriptive range labels for configuration",
+        "4. Frontend & Usability 🎨<br>* Skill Tool Deletion Status Display: Deletion indicators in tool list<br>* Dashboard Day-over-Day Comparison: Key metrics with yesterday comparison",
+        "5. Robustness & Bug Fixes 🔧<br>* Parameter Extraction Null Handling: Graceful handling of missing data<br>* Token Consumption Display Optimization: Accurate usage reporting<br>* Model Parameter Negative Value Fix: Clear parameter range definitions<br>* App Share Deletion Sync: Correct update of all share records<br>* Memory Write Task Ordering: Chronological execution per end_user<br>* Multimodal Model Missing Graceful Handling: No more interrupted writes<br>* Custom Tool Number Variable Pass-through: Type coercion fix<br>* Cluster Sub-Agent Display After Save: Fixed UI reflection<br>* Memory-Enabled Streaming Output Fix: String serialization resolved",
+        "<br>",
+        "v0.2.10 marks a significant step toward production maturity. The combination of deep thinking, interactive form workflows, and multimodal memory demonstrates the platform's evolution from memory storage to comprehensive cognitive infrastructure. We look forward to the v0.3.0 launch on April 17, bringing deeper agent reasoning, multi-agent collaboration, and further memory intelligence refinements. The blade has been forged — now it's time to wield it.",
+        "MemoryBear — Forging the Blade 🐻✨"
+      ]
+    }
+  },
   "v0.2.8": {
     "introduction": {
       "codeName": "景玉",

web/src/App.tsx

@@ -16,7 +16,7 @@ import {
   ConfigProvider,
   App as AntdApp
 } from 'antd';
-import { useTranslation } from 'react-i18next';
+import i18n from 'i18next';
 
 import { lightTheme } from './styles/antdThemeConfig.ts'
 import router from './routes';
@@ -29,11 +29,58 @@ import 'dayjs/plugin/utc'
 import { cookieUtils } from './utils/request';
 import { useUser } from '@/store/user';
 
+import menuJson from '@/store/menu.json';
+
+type MenuEntry = { path: string; i18nKey: string };
+
+function flattenMenuEntries(list: any[]): MenuEntry[] {
+  const result: MenuEntry[] = [];
+  for (const item of list) {
+    if (item.path && item.i18nKey && item.type !== 'group') result.push({ path: item.path, i18nKey: item.i18nKey });
+    if (item.subs?.length) result.push(...flattenMenuEntries(item.subs));
+  }
+  return result;
+}
+
+const menuEntries: MenuEntry[] = flattenMenuEntries([...menuJson.manage, ...menuJson.space]);
+
+function pathMatches(pattern: string, path: string): boolean {
+  if (pattern === path) return true;
+  if (pattern.includes(':')) {
+    return new RegExp('^' + pattern.replace(/:[\w-]+/g, '[^/]+') + '$').test(path);
+  }
+  return false;
+}
+
+function getPageTitle(pathname: string): string {
+  const appName = i18n.t('memoryBear');
+  const entry = menuEntries.find(e => pathMatches(e.path, pathname));
+  if (!entry) return appName;
+  return `${i18n.t(entry.i18nKey)} - ${appName}`;
+}
+
+const SKIP_TITLE_PATTERNS = [
+  '/user-memory/detail/:id/:type',
+  '/forgetting-engine/:id',
+  '/memory-extraction-engine/:id',
+  '/emotion-engine/:id',
+  '/reflection-engine/:id',
+];
+
+
+
 
 function App() {
-  const { t } = useTranslation();
   const { locale, language, timeZone } = useI18n()
   const { checkJump } = useUser();
+  useEffect(() => {
+    const unsubscribe = router.subscribe(({ location }) => {
+      if (SKIP_TITLE_PATTERNS.some(p => pathMatches(p, location.pathname))) return;
+      document.title = getPageTitle(location.pathname);
+    });
+    return () => unsubscribe();
+  }, [])
+
   useEffect(() => {
     const authToken = cookieUtils.get('authToken')
     if (!authToken && !window.location.hash.includes('#/login') && !window.location.hash.includes('#/conversation/') && !window.location.hash.includes('#/jump') && !window.location.hash.includes('#/invite-register')) {
@@ -44,7 +91,9 @@ function App() {
   }, [])
 
   useEffect(() => {
-    document.title = t('memoryBear')
+    if (!SKIP_TITLE_PATTERNS.some(p => pathMatches(p, router.state.location.pathname))) {
+      document.title = getPageTitle(router.state.location.pathname)
+    }
     dayjs.locale(language)
     localStorage.setItem('language', language)
   }, [language])

web/src/api/package.ts

@@ -0,0 +1,14 @@
+import { request } from '@/utils/request'
+
+import type { Package } from '@/views/Package/types'
+
+export const SYS_API_PREFIX = '/sys';
+// 套餐列表
+export const getPackageListUrl = `${SYS_API_PREFIX}/package-plans`
+export const getPackageList = (query: { category: Package['category']; status: boolean; }) => {
+  return request.get(getPackageListUrl, query)
+}
+// 获取套餐详情
+export const getPackageDetail = (package_plan_id: string) => {
+  return request.get(`${SYS_API_PREFIX}/package-plans/${package_plan_id}`)
+}

web/src/assets/images/workflow/checkList.svg

@@ -0,0 +1,16 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<svg width="16px" height="16px" viewBox="0 0 16 16" version="1.1" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+    <title>参与</title>
+    <g id="空间里层页面优化" stroke="none" stroke-width="1" fill="none" fill-rule="evenodd">
+        <g id="应用管理-工作流-配置-开始" transform="translate(-1173, -24)" fill="#000000" fill-rule="nonzero">
+            <g id="编组-11" transform="translate(1166, 17)">
+                <g id="参与" transform="translate(7, 7)">
+                    <g id="编组" transform="translate(1.5, 1)">
+                        <path d="M9.66581309,0 C11.5071324,0 12.9999203,1.50297946 12.9999203,3.35712964 L12.9999203,6.99997709 C12.9999203,7.34514783 12.7220975,7.62497504 12.3793991,7.62497504 C12.0367007,7.62497504 11.7588778,7.34514783 11.7588778,6.99997709 L11.7588778,3.35712964 C11.7588778,2.19344595 10.8218287,1.24999591 9.66581309,1.24999591 L3.33410726,1.24999591 C2.17807615,1.24999591 1.24104252,2.19344595 1.24104252,3.35712964 L1.24104252,10.6428245 C1.24104252,11.8065082 2.17809167,12.7499583 3.33410726,12.7499583 L6.04769325,12.7499583 C6.39040715,12.7499583 6.66821451,13.0297855 6.66821451,13.3749562 C6.66821451,13.720127 6.39040715,13.9999542 6.04769325,13.9999542 L3.33410726,13.9999542 C1.49278799,13.9999542 0,12.4969747 0,10.6428245 L0,3.35712964 C0,1.50297946 1.49278799,0 3.33410726,0 L9.66581309,0 Z" id="路径"></path>
+                        <path d="M11.8585646,8.937002 C12.0448761,8.6472842 12.4290718,8.56453447 12.7167144,8.75215885 C13.0043726,8.93981449 13.0865296,9.3267976 12.9002336,9.6165154 L10.2649729,13.7147051 C10.0576723,14.0370947 9.61342558,14.0966257 9.3296457,13.8400641 L7.8566058,12.5082872 C7.60157156,12.2777254 7.58041179,11.8825705 7.80932208,11.6256963 C8.03824788,11.3688222 8.43057245,11.3475097 8.68560669,11.5780715 L9.61814154,12.4211937 L11.8585646,8.93698637 L11.8585646,8.937002 Z M9.21354617,4.09820534 C9.55624455,4.09820534 9.83406743,4.37801692 9.83406743,4.72320329 C9.83406743,5.06837404 9.55624455,5.34820125 9.21354617,5.34820125 L3.78637417,5.34820125 C3.4436758,5.34820125 3.16585292,5.06837404 3.16585292,4.72320329 C3.16585292,4.37801692 3.4436758,4.09820534 3.78637417,4.09820534 L9.21354617,4.09820534 Z M9.21354617,7.74105279 C9.55624455,7.74105279 9.83406743,8.02086437 9.83406743,8.36605074 C9.83406743,8.71122149 9.55624455,8.9910487 9.21354617,8.9910487 L3.78637417,8.9910487 C3.4436758,8.9910487 3.16585292,8.71122149 3.16585292,8.36605074 C3.16585292,8.02086437 3.4436758,7.74105279 3.78637417,7.74105279 L9.21354617,7.74105279 Z" id="形状"></path>
+                    </g>
+                </g>
+            </g>
+        </g>
+    </g>
+</svg>

web/src/assets/images/workflow/features.svg

@@ -1,12 +1,14 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <svg width="16px" height="16px" viewBox="0 0 16 16" version="1.1" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
     <title>参与</title>
-    <g id="空间里层页面优化" stroke="none" stroke-width="1" fill="none" fill-rule="evenodd">
-        <g id="应用管理-工作流-配置-开始" transform="translate(-1211, -24)" fill="#171719" fill-rule="nonzero">
+    <g id="空间里层页面优化" stroke="none" stroke-width="1" fill="none" fill-rule="evenodd" stroke-linecap="round">
+        <g id="应用管理-工作流-配置-开始" transform="translate(-1211, -24)" stroke="#171719" stroke-width="1.2">
             <g id="编组-11" transform="translate(1204, 17)">
                 <g id="参与" transform="translate(7, 7)">
-                    <g id="编组-35" transform="translate(0.5, 1.5)">
-                        <path d="M13.3524137,3.04473843 C13.7876396,3.04473843 14.1979604,3.21975634 14.507738,3.53746403 C14.8173619,3.85501246 14.9923132,4.28005597 15,4.73408333 L15,10.2997805 C15,10.7391566 14.8365789,11.1556006 14.5400225,11.472512 C14.2665266,11.7647393 13.9083222,11.9416683 13.5224454,11.9771815 L13.5155273,13.3373525 C13.5155273,13.6047366 13.3547197,13.8526919 13.1155068,13.9536577 C13.021728,13.9861451 12.9450138,14 12.8673773,14 C12.6896587,14 12.521318,13.9261071 12.40494,13.797113 L10.6609614,11.9676263 L8.48098801,11.9676263 C8.12370606,11.9676263 7.83314543,11.6666401 7.83314543,11.2965385 C7.83314543,10.926437 8.12370606,10.6254507 8.48098801,10.6254507 L10.9635134,10.6254507 C11.1415394,10.6459942 11.2911243,10.7176576 11.3904376,10.8283378 L12.2272215,11.7015163 L12.2272215,11.2966978 C12.2272215,10.9371068 12.5239315,10.6334133 12.8750641,10.6334133 L13.3674798,10.6334133 C13.5491954,10.6334133 13.6969355,10.4803722 13.6969355,10.2921364 L13.6969355,4.72819101 C13.6969355,4.53995518 13.5491954,4.38691404 13.3674798,4.38691404 C13.0060469,4.38691404 12.7121041,4.08592781 12.7121041,3.71582623 C12.7121041,3.34588391 12.9994363,3.04473843 13.3524137,3.04473843 Z M10.4203649,0 C11.3164907,0 12.0455058,0.755172845 12.0455058,1.68345258 L12.0455058,8.25976271 C12.0455058,9.18804245 11.3131085,9.94305605 10.4129855,9.94305605 L5.13154658,9.94305605 L2.58091627,12.7683453 C2.45792764,12.9048242 2.29081685,12.9799911 2.11017731,12.9799911 C2.03346315,12.9799911 1.94675618,12.9634289 1.87234806,12.9344451 C1.62268115,12.8328423 1.462181,12.5875943 1.462181,12.3089033 L1.46940658,9.93604896 C1.08614328,9.89719148 0.730552424,9.71962553 0.459055037,9.42946844 C0.163113662,9.11319403 0,8.69770563 0,8.25960346 L0,1.68345258 C0,0.755172845 0.729015066,0 1.62514092,0 L10.4203649,0 Z M10.4208261,1.33453151 L1.62560213,1.33453151 C1.44388644,1.33453151 1.29614636,1.48757266 1.29614636,1.67580849 L1.29614636,8.25976271 C1.29614636,8.44799854 1.44388644,8.60103969 1.62560213,8.60103969 L2.14030952,8.60103969 C2.31449216,8.60103969 2.48329405,8.67588811 2.60320795,8.80663398 C2.72066209,8.93451331 2.78092651,9.10061312 2.77323973,9.27467552 C2.77323973,9.29060072 2.77293225,9.31273675 2.76570667,9.34156135 L2.75924977,10.5990149 L4.38623552,8.79532709 C4.40698985,8.76475071 4.43450856,8.74850701 4.45295685,8.73879264 L4.47217382,8.72557473 C4.48293533,8.71872689 4.49246695,8.71171981 4.50276724,8.70407571 C4.52905606,8.68464697 4.56180178,8.66028142 4.60638516,8.64483398 L4.65634929,8.63161607 L4.67326022,8.63161607 C4.70446859,8.62078693 4.73721431,8.61362059 4.76857641,8.60788752 L4.78302757,8.59291784 L10.4208261,8.59291784 C10.6025418,8.59291784 10.7502818,8.43987669 10.7502818,8.25164086 L10.7502818,1.67580849 C10.7502818,1.48757266 10.6025418,1.33453151 10.4208261,1.33453151 Z M3.1173004,4.58263471 C3.34559803,4.58263471 3.55975197,4.70939928 3.67612996,4.9132418 C3.85062007,5.22505716 4.19252844,5.51999181 4.63943835,5.74469634 C5.09157528,5.9719489 5.59213898,6.09712095 6.04904171,6.09712095 C6.99467049,6.09712095 8.05882956,5.56983768 8.42149226,4.92168216 C8.53833145,4.71688412 8.7524854,4.59011955 8.98078303,4.59011955 C9.1026955,4.59011955 9.21599877,4.62181069 9.30870145,4.68184869 C9.45797889,4.76513747 9.56497899,4.908942 9.61033104,5.08714495 C9.64815005,5.27283275 9.62339859,5.4460989 9.53976632,5.59356622 C8.85594957,6.7876375 7.23096239,7.43165247 6.04904171,7.43165247 C4.85989546,7.43165247 3.22968125,6.78477096 2.55078405,5.58608138 C2.46745926,5.43256248 2.4448601,5.24862645 2.48775238,5.07966011 C2.53125961,4.90830499 2.63549247,4.76816326 2.78892077,4.67468235 C2.89146254,4.61273333 2.99877011,4.58263471 3.1173004,4.58263471 Z" id="形状结合"></path>
+                    <g id="编组-35" transform="translate(1, 2)">
+                        <path d="M1.5,0 L9.5,0 C10.3284271,-2.22044605e-16 11,0.671572875 11,1.5 L11,7.5 C11,8.32842712 10.3284271,9 9.5,9 L4.16268077,9 L4.16268077,9 L1.61845449,11.7671841 L1.61845449,9 L1.5,9 C0.671572875,9 2.22044605e-16,8.32842712 2.22044605e-16,7.5 L0,1.5 C0,0.671572875 0.671572875,-2.22044605e-16 1.5,-2.22044605e-16 Z" id="矩形" stroke-linejoin="round"></path>
+                        <path d="M14,10.7913467 L14,10.7913467 L11.889042,10.7913467 L9.95430953,12.7671841 L9.95430953,10.7913467 L9.33585504,10.7913467 C8.78357029,10.7913467 8.33585504,10.3436314 8.33585504,9.79134668 L8.33585504,4.21772464 C8.33585504,3.66543989 8.78357029,3.21772464 9.33585504,3.21772464 L9.35694395,3.21772464 L9.35694395,3.21772464" id="路径" stroke-linejoin="round" transform="translate(11.1679, 7.9925) scale(-1, 1) translate(-11.1679, -7.9925)"></path>
+                        <path d="M2.57312498,4.7624569 C3.0405519,5.21866557 5.40029607,7.88187743 8.56554211,4.7624569" id="路径-22"></path>
                     </g>
                 </g>
             </g>

web/src/components/Chat/ChatContent.tsx

@@ -2,7 +2,7 @@
  * @Author: ZhaoYing 
  * @Date: 2025-12-10 16:46:17 
  * @Last Modified by: ZhaoYing
- * @Last Modified time: 2026-04-08 11:23:18
+ * @Last Modified time: 2026-04-14 10:13:56
  */
 import { type FC, useRef, useEffect, useState } from 'react'
 import clsx from 'clsx'
@@ -174,6 +174,7 @@ const ChatContent: FC<ChatContentProps> = ({
                       )
                     }
 
+                    const documentType = (file.file_type || file.type)?.split('/')
                     return (
                       <Flex
                         key={file.url || file.uid}
@@ -208,7 +209,7 @@ const ChatContent: FC<ChatContentProps> = ({
                         ></div>
                         <div className="rb:flex-1 rb:w-32.5">
                           <div className="rb:leading-4 rb:text-ellipsis rb:overflow-hidden rb:whitespace-nowrap">{file.name}</div>
-                          <div className="rb:leading-3.5 rb:mt-0.5 rb:text-[#5B6167] rb:text-ellipsis rb:overflow-hidden rb:whitespace-nowrap">{file.type?.split('/')[file.type?.split('/').length - 1]} · {file.size}</div>
+                          <div className="rb:leading-3.5 rb:mt-0.5 rb:text-[#5B6167] rb:text-ellipsis rb:overflow-hidden rb:whitespace-nowrap">{documentType?.[documentType.length - 1]} · {file.size}</div>
                         </div>
                       </Flex>
                     )
@@ -217,7 +218,7 @@ const ChatContent: FC<ChatContentProps> = ({
                 {/* Message bubble */}
                 <div className={clsx('rb:text-left rb:leading-5 rb:inline-block rb:wrap-break-word rb:relative', item.role === 'user' ? contentClassNames : '', {
                   // Error message style (content is null and not assistant message)
-                  'rb:bg-[rgba(255,93,52,0.08)] rb:text-[#FF5D34]': (item.status && item.status !== 'completed') || (errorDesc && item.role === 'assistant' && item.content === null && !renderRuntime),
+                  'rb:text-[#FF5D34]': (item.status && item.status !== 'completed') || (errorDesc && item.role === 'assistant' && item.content === null && !renderRuntime),
                   // Assistant message style
                   'rb:bg-[#E3EBFD] rb:p-[10px_12px_2px_12px] rb:rounded-lg rb:max-w-130': item.role === 'user',
                   'rb:max-w-full rb:w-full': item.role === 'assistant',

web/src/components/Header/index.module.css

@@ -12,6 +12,14 @@
   font-weight: 500;
   font-style: normal;
 }
+.breadcrumbTitle {
+  display: inline-block;
+  max-width: 200px;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+  vertical-align: bottom;
+}
 .header :global(.ant-breadcrumb) {
   line-height: 31px;
 }

web/src/components/Header/index.tsx

@@ -2,7 +2,7 @@
  * @Author: ZhaoYing 
  * @Date: 2026-02-02 15:07:49 
  * @Last Modified by: ZhaoYing
- * @Last Modified time: 2026-04-07 12:18:58
+ * @Last Modified time: 2026-04-16 10:31:21
  */
 /**
  * AppHeader Component
@@ -14,7 +14,7 @@
  */
 
 import { type FC, useRef, useState } from 'react';
-import { Layout, Dropdown, Breadcrumb, Flex } from 'antd';
+import { Layout, Dropdown, Breadcrumb, Flex, Tooltip } from 'antd';
 import type { MenuProps, BreadcrumbProps } from 'antd';
 import { useTranslation } from 'react-i18next';
 import { useLocation } from 'react-router-dom';
@@ -31,7 +31,7 @@ const { Header } = Layout;
 /**
  * @param source - Breadcrumb source type ('space' or 'manage'), defaults to 'manage'
  */
-const AppHeader: FC<{source?: 'space' | 'manage';}> = ({source = 'manage'}) => {
+const AppHeader: FC<{ source?: 'space' | 'manage'; }> = ({ source = 'manage' }) => {
   const { t } = useTranslation();
   const location = useLocation();
   const settingModalRef = useRef<SettingModalRef>(null)
@@ -39,7 +39,7 @@ const AppHeader: FC<{source?: 'space' | 'manage';}> = ({source = 'manage'}) => {
 
   const { user, logout } = useUser();
   const { allBreadcrumbs } = useMenu();
-  
+
   /**
    * Dynamically select breadcrumb source based on current route
    * - Knowledge base list: uses 'space' breadcrumb
@@ -48,24 +48,24 @@ const AppHeader: FC<{source?: 'space' | 'manage';}> = ({source = 'manage'}) => {
    */
   const getBreadcrumbSource = () => {
     const pathname = location.pathname;
-    
+
     // Knowledge base list page uses default space breadcrumb
     if (pathname === '/knowledge-base') {
       return 'space';
     }
-    
+
     // Knowledge base detail pages use independent breadcrumb
     if (pathname.includes('/knowledge-base/') && pathname !== '/knowledge-base') {
       return 'space-detail';
     }
-    
+
     // Other pages use the passed source
     return source;
   };
-  
+
   const breadcrumbSource = getBreadcrumbSource();
   const breadcrumbs = allBreadcrumbs[breadcrumbSource] || [];
-  
+
 
   /** Handle user logout */
   const handleLogout = () => {
@@ -76,9 +76,11 @@ const AppHeader: FC<{source?: 'space' | 'manage';}> = ({source = 'manage'}) => {
   const userMenuItems: MenuProps['items'] = [
     {
       key: '1',
-      icon: <Flex align="center" justify="center" className="rb:size-10 rb:rounded-xl rb:bg-[#155EEF] rb:text-white">
-        {/[\u4e00-\u9fa5]/.test(user.username) ? user.username.slice(0, 2) : user.username?.[0]}
-      </Flex>,
+      icon: user.username
+        ? <Flex align="center" justify="center" className="rb:size-10 rb:rounded-xl rb:bg-[#155EEF] rb:text-white">
+          {/[\u4e00-\u9fa5]/.test(user.username) ? user.username.slice(-2) : user.username[0]}
+        </Flex>
+        : null,
       label: (<>
         <div className="rb:text-[#212332] rb:leading-5">{user.username}</div>
         <div className="rb:text-[12px] rb:text-[#7B8085] rb:leading-4.5 rb:mt-0.5 rb:mr-2">{user.email}</div>
@@ -127,7 +129,7 @@ const AppHeader: FC<{source?: 'space' | 'manage';}> = ({source = 'manage'}) => {
       onClick: handleLogout,
     },
   ];
-  
+
   /**
    * Format breadcrumb items with proper titles, paths, and click handlers
    * - Translates i18n keys to display text
@@ -135,32 +137,34 @@ const AppHeader: FC<{source?: 'space' | 'manage';}> = ({source = 'manage'}) => {
    * - Disables navigation for the last breadcrumb item
    */
   const formatBreadcrumbNames = () => {
-    return breadcrumbs.filter(item => item.type !== 'group').map((menu, index) => {
+    const filtered = breadcrumbs.filter(item => item.type !== 'group');
+    return filtered.map((menu, index) => {
+      const label = menu.i18nKey ? t(menu.i18nKey) : menu.label;
+      const isLast = index === filtered.length - 1;
       const item: any = {
-        title: menu.i18nKey ? t(menu.i18nKey) : menu.label,
+        title: (
+          <Tooltip title={label} placement="bottom">
+            <span className={styles.breadcrumbTitle}>{label}</span>
+          </Tooltip>
+        ),
       };
-      
-      // If it's the last item, don't set path
-      if (index === breadcrumbs.length - 1) {
-        return item;
+
+      if (!isLast) {
+        if ((menu as any).onClick) {
+          item.onClick = (e: React.MouseEvent) => {
+            e.preventDefault();
+            (menu as any).onClick(e);
+          };
+          item.href = '#';
+        } else if (menu.path && menu.path !== '#') {
+          item.path = menu.path;
+        }
       }
-      
-      // If has custom onClick, use onClick and set href to '#' to show pointer cursor
-      if ((menu as any).onClick) {
-        item.onClick = (e: React.MouseEvent) => {
-          e.preventDefault();
-          (menu as any).onClick(e);
-        };
-        item.href = '#';
-      } else if (menu.path && menu.path !== '#') {
-        // Only set path when path is not '#'
-        item.path = menu.path;
-      }
-      
+
       return item;
     });
   }
-  
+
   const [open, setOpen] = useState(false);
   const handleOpenChange = (open: boolean) => {
     setOpen(open);
@@ -179,9 +183,9 @@ const AppHeader: FC<{source?: 'space' | 'manage';}> = ({source = 'manage'}) => {
           overlayClassName={styles.userDropdown}
         >
           <Flex align="center" className="rb:cursor-pointer rb:font-medium">
-            <Flex align="center" justify="center" className="rb:size-8 rb:rounded-xl rb:bg-[#155EEF] rb:text-white rb:mr-2!">
-              {/[\u4e00-\u9fa5]/.test(user.username) ? user.username.slice(user.username.length, -2) : user.username[0]}
-            </Flex>
+            {user.username && <Flex align="center" justify="center" className="rb:size-8 rb:rounded-xl rb:bg-[#155EEF] rb:text-white rb:mr-2!">
+              {/[\u4e00-\u9fa5]/.test(user.username) ? user.username.slice(-2) : user.username[0]}
+            </Flex>}
             <span className="rb:text-[#212332] rb:text-[12px] rb:leading-4 rb:mr-1">{user.username}</span>
             <div className={clsx("rb:size-3 rb:bg-cover rb:bg-[url('@/assets/images/common/arrow_up.svg')]", {
               'rb:rotate-180': !open,

web/src/hooks/useBreadcrumbManager.ts

@@ -1,8 +1,8 @@
 /*
  * @Author: ZhaoYing 
  * @Date: 2026-02-02 16:24:44 
- * @Last Modified by:   ZhaoYing 
- * @Last Modified time: 2026-02-02 16:24:44 
+ * @Last Modified by: ZhaoYing
+ * @Last Modified time: 2026-04-14 16:52:43
  */
 /**
  * useBreadcrumbManager Hook
@@ -18,8 +18,10 @@
 
 import { useCallback } from 'react';
 import { useNavigate } from 'react-router-dom';
+import { useTranslation } from 'react-i18next'
 import { useMenu } from '@/store/menu';
 import type { MenuItem } from '@/store/menu';
+import { useI18n } from '@/store/locale'
 
 /** Breadcrumb item interface */
 export interface BreadcrumbItem {
@@ -53,6 +55,8 @@ export interface BreadcrumbOptions {
 export const useBreadcrumbManager = (options?: BreadcrumbOptions) => {
   const { allBreadcrumbs, setCustomBreadcrumbs } = useMenu();
   const navigate = useNavigate();
+  const { t } = useTranslation()
+  const { language } = useI18n()
 
   /** Update breadcrumbs based on current path and type */
   const updateBreadcrumbs = useCallback((breadcrumbPath: BreadcrumbPath) => {
@@ -336,10 +340,10 @@ export const useBreadcrumbManager = (options?: BreadcrumbOptions) => {
     /** Use different keys based on breadcrumb type to implement independent breadcrumb paths */
     const breadcrumbKey = breadcrumbType === 'list' ? 'space' : 'space-detail';
     
-
-    
+    const lastMenu = customBreadcrumbs[customBreadcrumbs.length - 1]
+    document.title = `${lastMenu.i18nKey ? t(lastMenu.i18nKey) : lastMenu.label} - ${t('memoryBear') }`;
     setCustomBreadcrumbs(customBreadcrumbs, breadcrumbKey);
-  }, [setCustomBreadcrumbs, navigate, options?.breadcrumbType, options?.onKnowledgeBaseMenuClick, options?.onKnowledgeBaseFolderClick]);
+  }, [setCustomBreadcrumbs, navigate, options?.breadcrumbType, options?.onKnowledgeBaseMenuClick, options?.onKnowledgeBaseFolderClick, language]);
 
   return {
     updateBreadcrumbs,

web/src/i18n/en.ts

@@ -116,7 +116,7 @@ export const en = {
       prompt: 'Prompt Engineering',
       skills: 'Skill Library',
       workbench: 'Workbench',
-      memoryRelated: 'Memory-Related',
+      memoryRelated: 'Memory Hub',
       advancedSettings: 'Advanced Settings',
       promptHistory: 'My history',
       platformManagement: 'Platform Management',
@@ -1396,6 +1396,43 @@ export const en = {
       pleaseUploadFile: 'Please upload file', 
       setting: 'Settings',
       features: 'Conversation Features',
+      checkList: 'Check List',
+      checkListDesc: 'Ensure all issues are resolved before publishing',
+      checkListEmpty: 'No issues found',
+      notConnected: 'This node is not connected to other nodes',
+      goto: 'Go to',
+      cannotBeEmpty: 'cannot be empty',
+      checkListErrors: {
+        'llm.model_id': 'Model',
+        'llm.messages': 'Messages',
+        'end.output': 'Output',
+        'knowledge-retrieval.knowledge_retrieval': 'Knowledge bases',
+        'parameter-extractor.model_id': 'Model',
+        'parameter-extractor.text': 'Input variable',
+        'parameter-extractor.params': 'Params',
+        'memory-read.message': 'Message',
+        'memory-read.config_id': 'Memory config',
+        'memory-read.search_switch': 'Search mode',
+        'memory-write.messages': 'Messages',
+        'memory-write.config_id': 'Memory config',
+        'if-else.cases': 'Condition',
+        'question-classifier.model_id': 'Model',
+        'question-classifier.input_variable': 'Input variable',
+        'question-classifier.categories': 'Categories',
+        'iteration.input': 'Input variable',
+        'iteration.output': 'Output variable',
+        'var-aggregator.group_variables': 'Variables',
+        'assigner.assignments': 'Variables',
+        'http-request.url': 'API URL',
+        'http-request.body.data': 'Binary file variable',
+        'code.input_variables': 'Input variables',
+        'code.code': 'Code',
+        'code.output_variables': 'Output variables',
+        'jinja-render.mapping': 'Input variables',
+        'jinja-render.template': 'Template',
+        'document-extractor.file_selector': 'File variable',
+        'list-operator.input_list': 'Input list',
+      },
       file_upload: 'File Upload',
       file_upload_desc: 'The chat input box supports file uploads. Types include images, documents, and other types',
       settings: 'File Upload Settings',
@@ -1459,6 +1496,34 @@ export const en = {
       resetFeaturesTip: 'Please reconfigure the [Conversation Features - File Upload] settings',
       logTitle: 'Description',
       range: 'Range',
+      body: 'BODY Parameter Example',
+      bodyRequestExample: `{
+  "message": "user message content",
+  // string, required, the conversation content entered by the user;
+
+  "conversation_id": "conversation_id",
+  // string, optional, session ID; for multi-turn conversations, pass the conversation_id from the previous response; omit on first request;
+
+  "user_id": "user_id",
+  // string, optional, end-user identifier to distinguish memory and sessions across users; recommended to pass your business system user ID;
+
+  "variables": {},
+  // object, optional (requires application configuration to take effect);
+
+  "stream": false,
+  // boolean, optional, whether to stream the response; defaults to false; when true, returns an SSE event stream;
+
+  "thinking": false,
+  // boolean, optional, whether to enable deep thinking; defaults to false (requires application configuration when true);
+
+  "files": [],
+  // array, optional, list of multimodal attachments (requires application configuration to take effect);
+
+  "version":"app_release_id"
+  // string, optional, application version ID; specify a historical release version ID, or omit to use the currently active version;
+}`,
+      uploadCover: 'Import and Overwrite',
+      refresh: 'Refresh Current Page',
     },
     userMemory: {
       userMemory: 'User Memory',
@@ -1526,6 +1591,11 @@ export const en = {
       created_at: 'Created At',
       updated_at: 'Last Updated',
       fullScreen: 'Full Screen',
+      role: 'Role',
+      domain: 'Domain',
+      expertise: 'Expertise',
+      interests: 'Interests',
+      knowledge_tags: 'Knowledge Tags',
 
       memoryWindow: "{{name}}'s Memory Overview",
       memory_insight: 'Overall Overview',
@@ -2113,6 +2183,19 @@ Memory Bear: After the rebellion, regional warlordism intensified for several re
       return_text_position_enable: 'Return Text Position Info',
       return_text_position_enable_desc: 'Whether to return coordinate positions of recognized text',
 
+      OpenClawTool_desc: 'OpenClaw Remote Agent',
+      OpenClawTool_features: 'OpenClaw Remote Agent — 3D Printing and Device Management',
+      OpenClawTool_config_desc: 'Configure OpenClaw Gateway connection. Server URL and API Key are required.',
+      OpenClawTool_server_url_desc: 'OpenClaw Gateway server URL, e.g. http://xxx.xxx.xxx.xx:xxx',
+      OpenClawTool_api_key_desc: 'OpenClaw API Key, created in OpenClaw admin console',
+      OpenClawTool_agent_id_desc: 'Target Agent ID, defaults to main, usually no need to change',
+      OpenClawTool_enable: 'Enable OpenClaw',
+      agent_id: 'Agent ID',
+      '3dPrinting': '3D Printing',
+      deviceManagement: 'Device Management',
+      multimodalInteraction: 'Multimodal Interaction',
+      remoteAgent: 'Remote Agent',
+
       addCustom: 'Add Custom Tool',
       editCustom: 'Edit Custom Tool',
       schema: 'Schema',
@@ -2189,6 +2272,7 @@ Memory Bear: After the rebellion, regional warlordism intensified for several re
       addvariable: 'Chat Variables',
       addChatVariable: 'Add Chat Variable',
       editChatVariable: 'Edit Chat Variable',
+      invalidJSON: 'Invalid JSON format',
 
       config: {
         llm: {
@@ -2291,6 +2375,11 @@ Memory Bear: After the rebellion, regional warlordism intensified for several re
             "eq": 'Is',
             "ne": 'Is Not',
           },
+          file: {
+            "empty": 'Not Exist',
+            "not_empty": 'Exists',
+            eq: 'All Are'
+          },
           else_desc: 'Used to define the logic that should be executed when the if condition is not met.',
           unset: 'Condition Not Set',
           set: 'Set',
@@ -2403,7 +2492,8 @@ Memory Bear: After the rebellion, regional warlordism intensified for several re
         value: 'Value',
         addCase: 'Add Condition',
         addVariable: 'Add Variables',
-        output: 'Output Variable'
+        output: 'Output Variable',
+        duplicateName: 'Variable name cannot be duplicated',
       },
 
       clear: 'Clear',
@@ -2429,7 +2519,49 @@ Memory Bear: After the rebellion, regional warlordism intensified for several re
         iteration: 'Iteration',
         input_cycle_vars: 'Initial Loop Variables',
         output_cycle_vars: 'Final Loop Variables',
-      }
+      },
+      sureReplace: 'Confirm Replace',
+      checkList: 'Check List',
+      checkListDesc: 'Ensure all issues are resolved before publishing',
+      checkListEmpty: 'No issues found',
+      notConnected: 'This node is not connected to other nodes',
+      goto: 'Go to',
+      cannotBeEmpty: 'cannot be empty',
+      checkListErrors: {
+        'llm.model_id': 'Model',
+        'llm.messages': 'Messages',
+        'end.output': 'Output',
+        'knowledge-retrieval.knowledge_retrieval': 'Knowledge bases',
+        'parameter-extractor.model_id': 'Model',
+        'parameter-extractor.text': 'Input variable',
+        'parameter-extractor.params': 'Params',
+        'memory-read.message': 'Message',
+        'memory-read.config_id': 'Memory config',
+        'memory-read.search_switch': 'Search mode',
+        'memory-write.messages': 'Messages',
+        'memory-write.config_id': 'Memory config',
+        'if-else.cases': 'Condition',
+        'question-classifier.model_id': 'Model',
+        'question-classifier.input_variable': 'Input variable',
+        'question-classifier.categories': 'Categories',
+        'iteration.input': 'Input variable',
+        'iteration.output': 'Output variable',
+        'var-aggregator.group_variables': 'Variables',
+        'assigner.assignments': 'Variables',
+        'http-request.url': 'API URL',
+        'http-request.body.data': 'Binary file variable',
+        'code.input_variables': 'Input variables',
+        'code.code': 'Code',
+        'code.output_variables': 'Output variables',
+        'jinja-render.mapping': 'Input variables',
+        'jinja-render.template': 'Template',
+        'document-extractor.file_selector': 'File variable',
+        'list-operator.input_list': 'Input list',
+      },
+      checkListHasErrors: 'Please resolve all issues in the checklist before publishing',
+      variableSelect: {
+        empty: 'No variables available',
+      },
     },
     emotionEngine: {
       emotionEngineConfig: 'Emotion Engine Configuration',
@@ -2887,5 +3019,69 @@ Memory Bear: After the rebellion, regional warlordism intensified for several re
       apply: 'Apply',
       tools: 'Tools',
     },
+    package: {
+      package: 'Package Management',
+      saas_personal: 'SaaS Personal',
+      commercial_deployment: 'Commercial Deployment',
+      noCommercialPackages: 'No commercial deployment packages available',
+
+      addPackage: 'Add Plan',
+      packageName: 'Plan Name',
+      packageNameZh: 'Plan Name (中文)',
+      packageNameEn: 'Plan Name (English)',
+      packageNamePlaceholder: '中文, 例如：记忆体验版',
+      packageNamePlaceholderEn: 'English, e.g. Memory Trial Plan',
+      packageCategory: 'Package Category',
+      price: 'Price',
+      pricePlaceholder: 'e.g. 0, 19, 299 or Contact Us',
+      billingPeriod: 'Billing Period',
+      monthly: 'Monthly',
+      yearly: 'Yearly',
+      permanent_free: 'Permanent Free',
+      local_deployment: 'Local Deployment',
+      coreValue: 'Core Value',
+      coreValueZh: 'Core Value (中文)',
+      coreValueEn: 'Core Value (English)',
+      coreValuePlaceholder: '中文, 一句话描述核心价值',
+      coreValuePlaceholderEn: 'EngLish, describe the core value in one sentence',
+      tech_support: 'Technical Support',
+      tech_support_zh: 'Technical Support (中文)',
+      tech_support_en: 'Technical Support (English)',
+      technicalSupportPlaceholder: '中文, 例如：社群交流、工单支持',
+      technicalSupportPlaceholderEn: 'English, e.g. Community support, ticket support',
+      sla: 'SLA & Compliance',
+      slaZh: 'SLA & Compliance (中文)',
+      slaEn: 'SLA & Compliance (English)',
+      slaPlaceholder: '中文, 例如：无、验证力加强+审计日志',
+      slaPlaceholderEn: 'English, e.g. None, dedicated compute pool + audit logs',
+      customPage: 'Chat Page Customization',
+      customPageZh: 'Chat Page Customization (中文)',
+      customPageEn: 'Chat Page Customization (English)',
+      customPagePlaceholder: '中文, 例如：LOGO定制',
+      customPagePlaceholderEn: 'English, e.g. Logo customization',
+      primaryColor: 'Primary Color',
+      status: 'Status',
+      active: 'Active',
+      inactive: 'Inactive',
+      api_ops_rate_limit: 'API OPS Rate Limit',
+      ops: 'req/s',
+      pcs: 'pcs',
+      GB: 'GB',
+      tier_level: 'Tier Level',
+      numberPlaceholder: 'e.g. 10',
+
+      packageDetail: 'Package Detail',
+      basicInfo: 'Basic Info',
+      featureConfig: 'Billing Unit Quota',
+      workspace_quota: 'Workspace Quota',
+      skill_quota: 'Skill Library Quota',
+      app_quota: 'App Quota',
+      knowledge_capacity_quota: 'Knowledge Base Capacity',
+      memory_engine_quota: 'Memory Engine Quota',
+      end_user_quota: 'Memorable End Users',
+      ontology_project_quota: 'Ontology Project',
+      model_quota: 'Model Quota',
+      editPackage: 'Edit Package',
+    },
   },
 };

web/src/i18n/zh.ts

@@ -116,7 +116,7 @@ export const zh = {
       prompt: '提示词工程',
       skills: '技能库',
       workbench: '工作台',
-      memoryRelated: '记忆相关',
+      memoryRelated: '记忆中枢',
       advancedSettings: '高级设置',
       promptHistory: '我的历史',
       platformManagement: '平台管理',
@@ -831,6 +831,34 @@ export const zh = {
       resetFeaturesTip: '请重新配置【对话功能-文件上传】功能',
       logTitle: '描述',
       range: '范围',
+      body: 'BODY 参数示例',
+      bodyRequestExample: `{
+  "message": "用户消息内容",
+  // string，必填，用户输入的对话内容；
+
+  "conversation_id": "conversation_id",
+  // string，可选，会话ID，多轮对话时传上一次返回的conversation_id，首次不传；
+
+  "user_id": "user_id",
+  // string，可选，终端用户标识，用于区分不同用户的记忆和会话，建议传业务系统的用户ID；
+
+  "variables": {},
+  // object，可选（需要应用配置才支持生效）；
+
+  "stream": false,
+  // boolean，可选，是否流式返回；默认 false，true时返回SSE事件流；
+
+  "thinking": false,
+  // boolean，可选，是否启用深度思考；默认 false（true时需要应用配置才支持生效）；
+
+  "files": [],
+  // array，可选，多模态附件列表（需要应用配置才支持生效）；
+
+  "version":"app_release_id"
+  //string，可选，应用版本ID；指定历史发布版本ID，不传则使用当前生效版本；
+}`,
+      uploadCover: '导入并覆盖',
+      refresh: '刷新当前页',
     },
     table: {
       totalRecords: '共 {{total}} 条记录'
@@ -1524,6 +1552,11 @@ export const zh = {
       created_at: '创建时间',
       updated_at: '最后更新时间',
       fullScreen: '全屏',
+      role: '角色',
+      domain: '领域',
+      expertise: '专业擅长',
+      interests: '兴趣爱好',
+      knowledge_tags: '知识标签',
 
       memoryWindow: "{{name}} 的记忆之窗",
       memory_insight: '总体概述',
@@ -2109,6 +2142,21 @@ export const zh = {
       return_text_position_enable: '返回文本位置信息',
       return_text_position_enable_desc: '是否返回识别文字的坐标位置',
 
+      OpenClawTool_desc: 'OpenClaw远程Agent',
+      OpenClawTool_features: 'OpenClaw远程Agent —3D打印控制、设备管理等',
+      OpenClawTool_config_desc: '配置OpenClaw Gateway连接信息，需要提供服务地址和API Key。',
+      OpenClawTool_server_url_desc: 'OpenClaw Gateway 服务地址，如 http://xxx.xxx.xxx.xx:xxx',
+      OpenClawTool_api_key_desc: 'OpenClaw API Key，在 OpenClaw 管理后台创建',
+      OpenClawTool_agent_id_desc: '目标 Agent ID，默认为 main，通常无需修改',
+      OpenClawTool_enable: '启用 OpenClaw',
+      agent_id: 'Agent ID',
+      '3dPrinting': '3D 打印',
+      deviceManagement: '设备管理',
+      multimodalInteraction: '多模态交互',
+      remoteAgent: '远程 Agent',
+
+
+
       addCustom: '添加自定义工具',
       editCustom: '编辑自定义工具',
       schema: 'Schema',
@@ -2185,6 +2233,7 @@ export const zh = {
       addvariable: '会话变量',
       addChatVariable: '添加会话变量',
       editChatVariable: '编辑会话变量',
+      invalidJSON: 'JSON 格式不正确',
 
       config: {
         llm: {
@@ -2287,6 +2336,11 @@ export const zh = {
             "eq": '是',
             "ne": '不是',
           },
+          file: {
+            "empty": '不存在',
+            "not_empty": '存在',
+            eq: '全都是'
+          },
           else_desc: '用于定义当 if 条件不满足时应执行的逻辑。',
           unset: '条件未设置',
           set: '已设置',
@@ -2402,7 +2456,8 @@ export const zh = {
         value: '值',
         addCase: '添加条件',
         addVariable: '添加变量',
-        output: '输出变量'
+        output: '输出变量',
+        duplicateName: '变量名不能重复',
       },
 
       clear: '清空',
@@ -2430,6 +2485,47 @@ export const zh = {
         output_cycle_vars: '最终循环变量',
       },
       sureReplace: '确认替换',
+      checkList: '检查清单',
+      checkListDesc: '发布前确保所有问题均已解决',
+      checkListEmpty: '没有发现问题',
+      notConnected: '此节点尚未连接到其他节点',
+      goto: '转到',
+      cannotBeEmpty: '不能为空',
+      checkListErrors: {
+        'llm.model_id': '模型',
+        'llm.messages': '提示词',
+        'end.output': '回复',
+        'knowledge-retrieval.knowledge_retrieval': '知识库',
+        'parameter-extractor.model_id': '模型',
+        'parameter-extractor.text': '输入变量',
+        'parameter-extractor.params': '提取参数',
+        'memory-read.message': '消息',
+        'memory-read.config_id': '记忆配置',
+        'memory-read.search_switch': '检索模式',
+        'memory-write.messages': '消息',
+        'memory-write.config_id': '记忆配置',
+        'if-else.cases': '条件',
+        'question-classifier.model_id': '模型',
+        'question-classifier.input_variable': '输入变量',
+        'question-classifier.categories': '分类',
+        'iteration.input': '输入变量',
+        'iteration.output': '输出变量',
+        'var-aggregator.group_variables': '变量',
+        'assigner.assignments': '变量',
+        'http-request.url': 'API URL',
+        'http-request.body.data': 'binary文件类型变量',
+        'code.input_variables': '输入变量',
+        'code.code': '代码',
+        'code.output_variables': '输出变量',
+        'jinja-render.mapping': '输入变量',
+        'jinja-render.template': '模板',
+        'document-extractor.file_selector': '文件变量',
+        'list-operator.input_list': '输入变量',
+      },
+      checkListHasErrors: '发布前确认检查清单中所有问题均已解决',
+      variableSelect: {
+        empty: '暂无变量',
+      },
     },
     emotionEngine: {
       emotionEngineConfig: '情感引擎配置',
@@ -2887,5 +2983,69 @@ export const zh = {
       apply: '应用',
       tools: '工具',
     },
+    package: {
+      package: '套餐管理',
+      saas_personal: 'SaaS 个人版',
+      commercial_deployment: '商业化部署',
+      noCommercialPackages: '暂无商业化部署套餐',
+
+      addPackage: '添加套餐',
+      packageName: '套餐名称',
+      packageNameZh: '套餐名称 (中文)',
+      packageNameEn: '套餐名称 (English)',
+      packageNamePlaceholder: '中文, 例如：记忆体验版',
+      packageNamePlaceholderEn: 'English, e.g. Memory Trial Plan',
+      packageCategory: '套餐分类',
+      price: '价格',
+      pricePlaceholder: '例如: 0, 19, 299 或联系我们',
+      billingPeriod: '计费周期',
+      monthly: '月',
+      yearly: '年',
+      permanent_free: '永久免费',
+      local_deployment: '本地化部署',
+      coreValue: '核心价值',
+      coreValueZh: '核心价值 (中文)',
+      coreValueEn: '核心价值 (English)',
+      coreValuePlaceholder: '中文, 一句话描述核心价值',
+      coreValuePlaceholderEn: 'EngLish, describe the core value in one sentence',
+      tech_support: '技术支持',
+      tech_support_zh: '技术支持 (中文)',
+      tech_support_en: '技术支持 (English)',
+      technicalSupportPlaceholder: '中文, 例如：社群交流、工单支持',
+      technicalSupportPlaceholderEn: 'English, e.g. Community support, ticket support',
+      sla: 'SLA与合规',
+      slaZh: 'SLA与合规 (中文)',
+      slaEn: 'SLA与合规 (English)',
+      slaPlaceholder: '中文, 例如：无、验证力加强+审计日志',
+      slaPlaceholderEn: 'English, e.g. None, dedicated compute pool + audit logs',
+      customPage: '对应页面个性化配置',
+      customPageZh: '对应页面个性化配置 (中文)',
+      customPageEn: '对应页面个性化配置 (English)',
+      customPagePlaceholder: '中文, 例如：LOGO定制',
+      customPagePlaceholderEn: 'English, e.g. Logo customization',
+      primaryColor: '主题色',
+      status: '状态',
+      active: '启用',
+      inactive: '停用',
+      api_ops_rate_limit: 'API OPS 频次',
+      ops: '次/秒',
+      pcs: '个',
+      GB: 'GB',
+      tier_level: '层级',
+      numberPlaceholder: '如: 10',
+
+      packageDetail: '套餐详情',
+      basicInfo: '基础信息',
+      featureConfig: '计费单元配额',
+      workspace_quota: '空间数量',
+      skill_quota: '技能库数量',
+      app_quota: '应用数量',
+      knowledge_capacity_quota: '知识库容量',
+      memory_engine_quota: '记忆引擎数量',
+      end_user_quota: '可记忆终端用户数',
+      ontology_project_quota: '本体工程',
+      model_quota: '可负载模型数量',
+      editPackage: '编辑套餐',
+    },
   },
 }

web/src/routes/index.tsx

@@ -2,7 +2,7 @@
  * @Author: ZhaoYing 
  * @Date: 2026-02-02 16:33:11 
  * @Last Modified by: ZhaoYing
- * @Last Modified time: 2026-02-04 18:11:34
+ * @Last Modified time: 2026-04-13 16:53:15
  */
 /**
  * Route Configuration
@@ -76,13 +76,12 @@ const componentMap: Record<string, LazyExoticComponent<ComponentType<object>>> =
   SpaceManagement: lazy(() => import('@/views/SpaceManagement')),
   ApiKeyManagement: lazy(() => import('@/views/ApiKeyManagement')),
   EmotionEngine: lazy(() => import('@/views/EmotionEngine')),
-  StatementDetail: lazy(() => import('@/views/UserMemoryDetail/pages/StatementDetail')),
   ForgetDetail: lazy(() => import('@/views/UserMemoryDetail/pages/ForgetDetail')),
   MemoryNodeDetail: lazy(() => import('@/views/UserMemoryDetail/pages/index')),
   SelfReflectionEngine: lazy(() => import('@/views/SelfReflectionEngine')),
   OrderPayment: lazy(() => import('@/views/OrderPayment')),
   OrderHistory: lazy(() => import('@/views/OrderHistory')),
-  Pricing: lazy(() => import('@/views/Pricing')),
+  Package: lazy(() => import('@/views/Package')),
   ToolManagement: lazy(() => import('@/views/ToolManagement')),
   SpaceConfig: lazy(() => import('@/views/SpaceConfig')),
   Ontology: lazy(() => import('@/views/Ontology')),

web/src/routes/routes.json

@@ -7,7 +7,7 @@
       { "path": "/model", "element": "ModelManagement" },
       { "path": "/space", "element": "SpaceManagement" },
       { "path": "/tool", "element": "ToolManagement" },
-      { "path": "/pricing", "element": "Pricing" },
+      { "path": "/pricing", "element": "Package" },
       { "path": "/order-pay", "element": "OrderPayment" },
       { "path": "/orders", "element": "OrderHistory" },
       { "path": "/skills", "element": "Skills" },
@@ -48,7 +48,6 @@
       { "path": "/application/config/:id", "element": "ApplicationConfig" },
       { "path": "/application/config/:id/:source", "element": "ApplicationConfig" },
       { "path": "/user-memory/neo4j/:id", "element": "Neo4jUserMemoryDetail" },
-      { "path": "/statement/:id", "element": "StatementDetail" },
       { "path": "/user-memory/detail/:id/:type", "element": "MemoryNodeDetail" },
       { "path": "/ontology/:id", "element": "OntologyDetail" }
     ]

web/src/store/menu.json

@@ -6,7 +6,7 @@
       "code": "workbench",
       "label": "workbench",
       "i18nKey": "menu.workbench",
-      "path": "/",
+      "path": null,
       "enable": true,
       "display": true,
       "level": 1,
@@ -174,7 +174,7 @@
       "code": "workbench",
       "label": "workbench",
       "i18nKey": "menu.workbench",
-      "path": "/",
+      "path": null,
       "enable": true,
       "display": true,
       "level": 1,
@@ -425,15 +425,14 @@
                 {
                   "id": 2211,
                   "parent": 221,
-                  "code": "statementDetail",
+                  "code": "userMemoryDetail",
                   "label": "记忆详情",
-                  "i18nKey": "menu.statementDetail",
-                  "path": "/statement/:id",
+                  "i18nKey": "menu.userMemoryDetail",
+                  "path": "/user-memory/detail/:id/:type",
                   "enable": true,
                   "display": false,
-                  "level": 4,
-                  "sort": 0,
-                  "subs": null
+                  "level": 3,
+                  "sort": 0
                 }
               ]
             },

web/src/store/workflow.ts

@@ -0,0 +1,21 @@
+/*
+ * @Author: ZhaoYing 
+ * @Date: 2026-04-10 18:11:19 
+ * @Last Modified by:   ZhaoYing 
+ * @Last Modified time: 2026-04-10 18:11:19 
+ */
+import { create } from 'zustand'
+import type { NodeCheckResult } from '@/views/Workflow/components/CheckList'
+
+interface WorkflowState {
+  checkResults: Record<string, NodeCheckResult[]>
+  setCheckResults: (appId: string, results: NodeCheckResult[]) => void
+  getCheckResults: (appId: string) => NodeCheckResult[]
+}
+
+export const useWorkflowStore = create<WorkflowState>((set, get) => ({
+  checkResults: {},
+  setCheckResults: (appId, results) =>
+    set(state => ({ checkResults: { ...state.checkResults, [appId]: results } })),
+  getCheckResults: (appId) => get().checkResults[appId] ?? [],
+}))

web/src/styles/index.css

@@ -420,4 +420,7 @@ body {
 .ant-picker-outlined:focus,
 .ant-picker-outlined:focus-within {
   box-shadow: none;
+}
+.ͼ1.cm-focused {
+  outline: none;
 }

web/src/utils/apiKeyReplacer.ts

@@ -43,7 +43,8 @@ export const maskApiKeys = (text: string): string => {
     result = result.replace(API_KEY_PATTERNS[key as keyof typeof API_KEY_PREFIX], (match) => {
       const prefixLength = API_KEY_PREFIX[key].length
       const prefix = match.substring(0, prefixLength)
-      return prefix + '*'.repeat(match.length - prefixLength)
+      const suffix = match.slice(-4)
+      return prefix + '*'.repeat(match.length - prefixLength - 4) + suffix
     })
   })
 

web/src/utils/request.ts

@@ -2,7 +2,7 @@
  * @Author: ZhaoYing 
  * @Date: 2026-02-02 16:35:15 
  * @Last Modified by: ZhaoYing
- * @Last Modified time: 2026-03-06 10:39:00
+ * @Last Modified time: 2026-04-14 14:43:54
  */
 /**
  * HTTP Request Utility Module
@@ -23,6 +23,7 @@ import { clearAuthData } from './auth';
 import { message } from 'antd';
 import { refreshTokenUrl, refreshToken, loginUrl, logoutUrl } from '@/api/user'
 import i18n from '@/i18n'
+import { SYS_API_PREFIX } from '@/api/package'
 
 /**
  * Standard API response structure
@@ -74,6 +75,10 @@ let requests: RequestQueueItem[] = [];
 // Request interceptor
 service.interceptors.request.use(
   (config) => {
+    console.log('config', config, config.url?.startsWith(SYS_API_PREFIX))
+    if (config.url?.startsWith(SYS_API_PREFIX)) {
+      config.baseURL = '';
+    }
     if (!config.headers.Authorization) {
       const token = cookieUtils.get('authToken');
       if (token) {

web/src/views/ApplicationConfig/Agent.tsx

@@ -229,7 +229,11 @@ const Agent = forwardRef<AgentRef, { onFeaturesLoad?: (features: FeaturesConfigF
         ...knowledgeRest,
         knowledge_bases: knowledge_bases.map(item => ({
           kb_id: item.kb_id || item.id,
-          ...(item.config || {})
+          retrieve_type: item.retrieve_type,
+          top_k: item.top_k,
+          similarity_threshold: item.similarity_threshold,
+          vector_similarity_weight: item.vector_similarity_weight,
+          // ...(item.config || {})
         }))
       } as KnowledgeConfig : null,
       tools: tools.map(vo => {

web/src/views/ApplicationConfig/Api.tsx

@@ -2,7 +2,7 @@
  * @Author: ZhaoYing 
  * @Date: 2026-02-03 16:29:29 
  * @Last Modified by: ZhaoYing
- * @Last Modified time: 2026-03-26 15:31:36
+ * @Last Modified time: 2026-04-10 18:09:56
  */
 import { type FC, useState, useRef, useEffect } from 'react';
 import clsx from 'clsx';
@@ -18,6 +18,7 @@ import ApiKeyConfigModal from './components/ApiKeyConfigModal';
 import { getApiKeyList, getApiKeyStats, deleteApiKey } from '@/api/apiKey';
 import { maskApiKeys } from '@/utils/apiKeyReplacer'
 import RbCard from '@/components/RbCard/Card';
+import CodeMirrorEditor from '@/components/CodeMirrorEditor'
 
 /**
  * API configuration page component
@@ -155,6 +156,21 @@ const Api: FC<{ application: Application | null }> = ({ application }) => {
               {t('common.copy')}
             </Button>
           </Flex>
+
+          <div className="rb:font-medium rb:mt-4!">
+            {t('application.body')}
+          </div>
+          <Flex align="start" justify="space-between" className="rb:text-[#5B6167] rb:mt-3! rb:py-2! rb:px-4! rb:bg-white rb-border rb:rounded-lg rb:leading-5">
+            <CodeMirrorEditor readOnly={true} value={t('application.bodyRequestExample')} />
+
+            <Button className="rb:px-2! rb:h-7! rb:group" onClick={() => handleCopy(t('application.bodyRequestExample'))}>
+              <div
+                className="rb:w-4 rb:h-4 rb:cursor-pointer rb:bg-cover rb:bg-[url('@/assets/images/copy.svg')] rb:group-hover:bg-[url('@/assets/images/copy_active.svg')]"
+              ></div>
+              {t('common.copy')}
+            </Button>
+          </Flex>
+
         </RbCard>
         <RbCard
           title={() => (<Flex align="center">

web/src/views/ApplicationConfig/ReleasePage.tsx

@@ -2,12 +2,13 @@
  * @Author: ZhaoYing 
  * @Date: 2026-02-03 16:29:41 
  * @Last Modified by: ZhaoYing
- * @Last Modified time: 2026-03-26 15:24:41
+ * @Last Modified time: 2026-04-10 17:02:07
  */
 import { type FC, useState, useEffect, useRef } from 'react';
 import { useTranslation } from 'react-i18next';
 import clsx from 'clsx';
 import { Space, Input, Form, App, Flex } from 'antd';
+import copy from 'copy-to-clipboard';
 
 import Tag, { type TagProps } from './components/Tag'
 import RbCard from '@/components/RbCard/Card'
@@ -17,6 +18,7 @@ import ReleaseShareModal from './components/ReleaseShareModal'
 import AppSharingModal from './components/AppSharingModal'
 import type { Release, ReleaseModalRef, ReleaseShareModalRef, AppSharingModalRef } from './types'
 import type { Application } from '@/views/ApplicationManagement/types'
+import { useWorkflowStore } from '@/store/workflow'
 import Empty from '@/components/Empty'
 import { formatDateTime } from '@/utils/format';
 import Markdown from '@/components/Markdown'
@@ -40,6 +42,7 @@ const heightClass = 'rb:max-h-[calc(100vh-140px)]'
 const ReleasePage: FC<{data: Application; refresh: () => void}> = ({data, refresh}) => {
   const { t } = useTranslation();
   const { message } = App.useApp()
+  const { getCheckResults } = useWorkflowStore()
   const releaseModalRef = useRef<ReleaseModalRef>(null)
   const releaseShareModalRef = useRef<ReleaseShareModalRef>(null)
   const appSharingModalRef = useRef<AppSharingModalRef>(null)
@@ -75,6 +78,10 @@ const ReleasePage: FC<{data: Application; refresh: () => void}> = ({data, refres
     if (!selectedVersion) return
     appExport(data.id, data.name, { release_id: selectedVersion.id})
   }
+  const handleCopy = (id: string) => {
+    copy(id)
+    message.success(t('common.copySuccess'))
+  }
   return (
     <Flex gap={12}>
       <div className="rb:w-101 rb:h-full">
@@ -102,7 +109,7 @@ const ReleasePage: FC<{data: Application; refresh: () => void}> = ({data, refres
                         </Tag>}
                       </>}
                       className={clsx("rb:hover:shadow-[0px_2px_8px_0px_rgba(0,0,0,0.2)]! rb:cursor-pointer rb:bg-white", {
-                        'rb:border-[#171719]!': version.id === selectedVersion.id,
+                        'rb:border! rb:border-[#171719]!': version.id === selectedVersion.id,
                         'rb:border-[#DFE4ED] ': version.id !== selectedVersion.id
                       })}
                       headerType="borderless"
@@ -140,13 +147,30 @@ const ReleasePage: FC<{data: Application; refresh: () => void}> = ({data, refres
                 <RbButton type="primary" ghost onClick={() => releaseShareModalRef.current?.handleOpen()}>{t('application.share')}</RbButton>
                 {data?.type !== 'multi_agent' && <RbButton type="primary" ghost onClick={() => appSharingModalRef.current?.handleOpen()}>{t('application.sharing')}</RbButton>}
               </>}
-              <RbButton type="primary" onClick={() => releaseModalRef.current?.handleOpen()}>{t('application.release')}</RbButton>
+              <RbButton type="primary" onClick={async () => {
+                if (data?.type === 'workflow') {
+                  const errors = getCheckResults(data.id)
+                  if (errors.length) {
+                    message.error(t('workflow.checkListHasErrors'))
+                    return
+                  }
+                }
+                releaseModalRef.current?.handleOpen()
+              }}>{t('application.release')}</RbButton>
             </Space>
           </Flex>
           {selectedVersion && 
             <Flex gap={16} vertical className={`${heightClass} rb:overflow-y-auto`}>
               <RbCard
-                title={t('application.VersionInformation')}
+                title={() => <Flex>{t('application.VersionInformation')}
+                  <Flex align="center" className="rb:text-[#5B6167] rb:text-[12px]">
+                    (ID: {selectedVersion.id}
+                      <div className="rb:size-4.5 rb:ml-1 rb:cursor-pointer rb:bg-cover rb:bg-[url('@/assets/images/common/copy_dark.svg')]"
+                        onClick={() => handleCopy(selectedVersion.id)}
+                      ></div>
+                    )
+                  </Flex>
+                </Flex>}
                 headerType="borderless"
               >
                 <div className="rb:grid rb:grid-cols-3 rb:gap-4">

web/src/views/ApplicationConfig/components/Chat.tsx

@@ -2,7 +2,7 @@
  * @Author: ZhaoYing 
  * @Date: 2026-02-03 16:27:39 
  * @Last Modified by: ZhaoYing
- * @Last Modified time: 2026-03-31 15:02:07
+ * @Last Modified time: 2026-04-10 18:51:43
  */
 /**
  * Chat debugging component for application testing
@@ -291,7 +291,6 @@ const Chat: FC<ChatProps> = ({
         addAssistantMessage()
 
         const handleStreamMessage = (data: SSEMessage[]) => {
-          setCompareLoading(false)
 
           data.map(item => {
             const { model_config_id, conversation_id, content, message_length, audio_url, citations } = item.data as {
@@ -306,12 +305,21 @@ const Chat: FC<ChatProps> = ({
             
             switch (item.event) {
               case 'model_reasoning':
+                if (compareLoading) {
+                  setCompareLoading(false)
+                }
                 updateAssistantReasoningMessage(content, model_config_id, conversation_id)
                 break;
               case 'model_message':
+                if (compareLoading) {
+                  setCompareLoading(false)
+                }
                 updateAssistantMessage(content, model_config_id, conversation_id, audio_url)
                 break;
               case 'model_end':
+                if (compareLoading) {
+                  setCompareLoading(false)
+                }
                 const idToPoll = `${model_config_id}_${audio_url}`
                 if (audio_url && !audioStatusMap[idToPoll]) {
                   setAudioStatusMap(prev => ({
@@ -352,6 +360,9 @@ const Chat: FC<ChatProps> = ({
                 updateErrorAssistantMessage(message_length, model_config_id)
                 break;
               case 'compare_end':
+                if (compareLoading) {
+                  setCompareLoading(false)
+                }
                 setLoading(false);
                 break;
             }
@@ -473,7 +484,6 @@ const Chat: FC<ChatProps> = ({
         addClusterAssistantMessage()
 
         const handleStreamMessage = (data: SSEMessage[]) => {
-          setCompareLoading(false)
 
           data.map(item => {
             const { conversation_id, content, message_length } = item.data as { conversation_id: string, content: string, message_length: number };
@@ -485,15 +495,24 @@ const Chat: FC<ChatProps> = ({
                 }
                 break
               case 'message':
+                if (compareLoading) {
+                  setCompareLoading(false)
+                }
                 updateClusterAssistantMessage(content)
                 if (conversation_id && conversationId !== conversation_id) {
                   setConversationId(conversation_id);
                 }
                 break;
               case 'model_end':
+                if (compareLoading) {
+                  setCompareLoading(false)
+                }
                 updateClusterErrorAssistantMessage(message_length)
                 break;
               case 'compare_end':
+                if (compareLoading) {
+                  setCompareLoading(false)
+                }
                 setLoading(false);
                 break;
             }