[changes] Community Clustering Retrieval Module

api/app/repositories/neo4j/graph_saver.py

@@ -1,5 +1,5 @@
 import asyncio
-from typing import List
+from typing import List, Optional
 
 # 使用新的仓储层
 from app.repositories.neo4j.neo4j_connector import Neo4jConnector
@@ -156,10 +156,13 @@ async def save_dialog_and_statements_to_neo4j(
         entity_edges: List[EntityEntityEdge],
         statement_chunk_edges: List[StatementChunkEdge],
         statement_entity_edges: List[StatementEntityEdge],
-        connector: Neo4jConnector
+        connector: Neo4jConnector,
 ) -> bool:
     """Save dialogue nodes, chunk nodes, statement nodes, entities, and all relationships to Neo4j using graph models.
 
+    只负责数据写入，不触发聚类。聚类由调用方在写入成功后通过
+    schedule_clustering_after_write() 显式触发。
+
     Args:
         dialogue_nodes: List of DialogueNode objects to save
         chunk_nodes: List of ChunkNode objects to save
@@ -290,13 +293,6 @@ async def save_dialog_and_statements_to_neo4j(
         logger.info("Transaction completed. Summary: %s", summary)
         logger.debug("Full transaction results: %r", results)
 
-        # 写入成功后，触发聚类
-        if entity_nodes:
-            end_user_id = entity_nodes[0].end_user_id
-            new_entity_ids = [e.id for e in entity_nodes]
-            logger.info(f"[Clustering] 准备触发聚类，实体数: {len(new_entity_ids)}, end_user_id: {end_user_id}")
-            await _trigger_clustering(new_entity_ids, end_user_id)
-
         return True
 
     except Exception as e:
@@ -306,9 +302,38 @@ async def save_dialog_and_statements_to_neo4j(
         return False
 
 
+def schedule_clustering_after_write(
+    entity_nodes: List,
+    config_id: Optional[str] = None,
+    llm_model_id: Optional[str] = None,
+    embedding_model_id: Optional[str] = None,
+) -> None:
+    """
+    写入 Neo4j 成功后，调度后台聚类任务。
+
+    可通过环境变量 CLUSTERING_ENABLED=false 禁用（用于基准测试对比）。
+    使用 asyncio.create_task 异步触发，不阻塞写入响应。
+    """
+    if not entity_nodes:
+        return
+
+    clustering_enabled = os.getenv("CLUSTERING_ENABLED", "true").lower() != "false"
+    if not clustering_enabled:
+        logger.info("[Clustering] 聚类已禁用（CLUSTERING_ENABLED=false），跳过聚类触发")
+        return
+
+    end_user_id = entity_nodes[0].end_user_id
+    new_entity_ids = [e.id for e in entity_nodes]
+    logger.info(f"[Clustering] 准备触发聚类，实体数: {len(new_entity_ids)}, end_user_id: {end_user_id}")
+    asyncio.create_task(_trigger_clustering(new_entity_ids, end_user_id, config_id=config_id, llm_model_id=llm_model_id, embedding_model_id=embedding_model_id))
+
+
 async def _trigger_clustering(
     new_entity_ids: List[str],
     end_user_id: str,
+    config_id: Optional[str] = None,
+    llm_model_id: Optional[str] = None,
+    embedding_model_id: Optional[str] = None,
 ) -> None:
     """
     聚类触发函数，自动判断全量初始化还是增量更新。
@@ -318,7 +343,7 @@ async def _trigger_clustering(
         from app.core.memory.storage_services.clustering_engine import LabelPropagationEngine
         logger.info(f"[Clustering] 开始聚类，end_user_id={end_user_id}, 实体数={len(new_entity_ids)}")
         connector = Neo4jConnector()
-        engine = LabelPropagationEngine(connector)
+        engine = LabelPropagationEngine(connector, config_id=config_id, llm_model_id=llm_model_id, embedding_model_id=embedding_model_id)
         await engine.run(end_user_id=end_user_id, new_entity_ids=new_entity_ids)
         logger.info(f"[Clustering] 聚类完成，end_user_id={end_user_id}")
     except Exception as e: