refactor(memory): reorganize imports and move MemoryClientFactory to utils

- Move MemoryClientFactory from app.core.memory.client_factory to app.core.memory.utils.llm.llm_utils
- Update all evaluation modules to import MemoryClientFactory from new location (locomo, longmemeval, memsciqa)
- Move GenerateCacheRequest from memory_storage_service to memory_storage_schema
- Update memory_storage_controller imports to reflect schema reorganization
- Add analytics_user_summary import to memory_storage_controller from user_memory_service
- Consolidate utility imports across evaluation test files for consistency
- Improve code organization by centralizing LLM utilities in dedicated utils module

api/app/core/memory/evaluation/locomo/locomo_benchmark.py

@@ -23,7 +23,6 @@ except ImportError:
     def load_dotenv():
         pass
 
-from app.core.memory.client_factory import MemoryClientFactory
 from app.core.memory.evaluation.common.metrics import (
     avg_context_tokens,
     bleu1,
@@ -51,6 +50,7 @@ from app.core.memory.utils.definitions import (
     SELECTED_GROUP_ID,
     SELECTED_LLM_ID,
 )
+from app.core.memory.utils.llm.llm_utils import MemoryClientFactory
 from app.core.models.base import RedBearModelConfig
 from app.db import get_db_context
 from app.repositories.neo4j.neo4j_connector import Neo4jConnector

api/app/core/memory/evaluation/locomo/locomo_test.py

@@ -435,12 +435,12 @@ async def run_enhanced_evaluation():
             return None
      
     # 修正导入路径：使用 app.core.memory.src 前缀
-    from app.core.memory.client_factory import MemoryClientFactory
     from app.core.memory.llm_tools.openai_embedder import OpenAIEmbedderClient
     from app.core.memory.utils.config.definitions import (
         SELECTED_EMBEDDING_ID,
         SELECTED_LLM_ID,
     )
+    from app.core.memory.utils.llm.llm_utils import MemoryClientFactory
     from app.core.models.base import RedBearModelConfig
     from app.db import get_db_context
     from app.repositories.neo4j.graph_search import search_graph_by_embedding

api/app/core/memory/evaluation/locomo/qwen_search_eval.py

@@ -15,7 +15,6 @@ except Exception:
 
 import re
 
-from app.core.memory.client_factory import MemoryClientFactory
 from app.core.memory.evaluation.common.metrics import (
     avg_context_tokens,
     bleu1,
@@ -34,6 +33,7 @@ from app.core.memory.utils.config.definitions import (
     SELECTED_GROUP_ID,
     SELECTED_LLM_ID,
 )
+from app.core.memory.utils.llm.llm_utils import MemoryClientFactory
 from app.core.models.base import RedBearModelConfig
 from app.db import get_db_context
 from app.repositories.neo4j.graph_search import search_graph, search_graph_by_embedding

api/app/core/memory/evaluation/longmemeval/qwen_search_eval.py

@@ -34,7 +34,6 @@ try:
     )
 except Exception:
     ingest_contexts_via_full_pipeline = None  # 在运行时做兜底检查
-from app.core.memory.client_factory import MemoryClientFactory
 from app.core.memory.evaluation.common.metrics import (
     avg_context_tokens,
     jaccard,
@@ -48,6 +47,7 @@ from app.core.memory.utils.config.definitions import (
     SELECTED_EMBEDDING_ID,
     SELECTED_LLM_ID,
 )
+from app.core.memory.utils.llm.llm_utils import MemoryClientFactory
 from app.core.models.base import RedBearModelConfig
 from app.db import get_db_context
 from app.repositories.neo4j.graph_search import search_graph, search_graph_by_embedding

api/app/core/memory/evaluation/longmemeval/test_eval.py

@@ -15,7 +15,6 @@ except Exception:
         return None
 
 # 与现有评估脚本保持一致的导入方式
-from app.core.memory.client_factory import MemoryClientFactory
 from app.core.memory.evaluation.common.metrics import (
     avg_context_tokens,
     jaccard,
@@ -29,6 +28,7 @@ from app.core.memory.utils.config.definitions import (
     SELECTED_EMBEDDING_ID,
     SELECTED_LLM_ID,
 )
+from app.core.memory.utils.llm.llm_utils import MemoryClientFactory
 from app.core.models.base import RedBearModelConfig
 from app.db import get_db_context
 from app.repositories.neo4j.graph_search import search_graph, search_graph_by_embedding

api/app/core/memory/evaluation/memsciqa/evaluate_qa.py

@@ -15,7 +15,6 @@ except Exception:
     def load_dotenv():
         return None
 
-from app.core.memory.client_factory import MemoryClientFactory
 from app.core.memory.evaluation.common.metrics import (
     avg_context_tokens,
     exact_match,
@@ -31,6 +30,7 @@ from app.core.memory.utils.config.definitions import (
     SELECTED_GROUP_ID,
     SELECTED_LLM_ID,
 )
+from app.core.memory.utils.llm.llm_utils import MemoryClientFactory
 from app.db import get_db_context
 from app.repositories.neo4j.neo4j_connector import Neo4jConnector
 

api/app/core/memory/evaluation/memsciqa/memsciqa-test.py

@@ -24,7 +24,6 @@ for _p in (_SRC_DIR, _PROJECT_ROOT):
         sys.path.insert(0, _p)
 
 # 对齐 locomo_test 的检索逻辑：直接使用 graph_search 与 Neo4jConnector/Embedder1
-from app.core.memory.client_factory import MemoryClientFactory
 from app.core.memory.evaluation.common.metrics import (
     avg_context_tokens,
     exact_match,
@@ -37,6 +36,7 @@ from app.core.memory.utils.config.definitions import (
     SELECTED_GROUP_ID,
     SELECTED_LLM_ID,
 )
+from app.core.memory.utils.llm.llm_utils import MemoryClientFactory
 from app.core.models.base import RedBearModelConfig
 from app.db import get_db_context
 from app.repositories.neo4j.graph_search import search_graph, search_graph_by_embedding

api/app/core/memory/storage_services/reflection_engine/self_reflexion.py

@@ -19,7 +19,6 @@ from typing import Any, Dict, List, Optional
 
 from app.core.memory.llm_tools.openai_client import OpenAIClient
 from app.core.memory.utils.config import definitions as config_defs
-from app.core.memory.utils.config import get_model_config
 from app.core.memory.utils.config.get_data import (
     extract_and_process_changes,
     get_data,
@@ -169,12 +168,18 @@ class ReflectionEngine:
                 self.llm_client = factory.get_llm_client(config_defs.SELECTED_LLM_ID)
         elif isinstance(self.llm_client, str):
             # 如果 llm_client 是字符串（model_id），则用它初始化客户端
-            # from app.core.memory.utils.llm.llm_utils import MemoryClientFactory
+            from app.core.memory.utils.llm.llm_utils import MemoryClientFactory
             from app.db import get_db_context
-            # model_id = self.llm_client
+            from app.services.memory_config_service import MemoryConfigService
+            model_id = self.llm_client
             with get_db_context() as db:
                 factory = MemoryClientFactory(db)
                 # self.llm_client = factory.get_llm_client(model_id)
+                
+                # Use MemoryConfigService to get model config
+                config_service = MemoryConfigService(db)
+                model_config = config_service.get_model_config(model_id)
+                
             extra_params={
                     "temperature": 0.2,  # 降低温度提高响应速度和一致性
                     "max_tokens": 600,  # 限制最大token数
@@ -182,7 +187,6 @@ class ReflectionEngine:
                     "stream": False,  # 确保非流式输出以获得最快响应
                 }
 
-            model_config = get_model_config(self.llm_client)
             self.llm_client  = OpenAIClient(RedBearModelConfig(
                 model_name=model_config.get("model_name"),
                 provider=model_config.get("provider"),