Merge pull request #659 from SuanmoSuanyangTechnology/feature/rich-summary

[add] Introduce examples and triples to enrich the community summaries

api/app/core/memory/storage_services/clustering_engine/label_propagation.py

@@ -439,15 +439,17 @@ class LabelPropagationEngine:
 
     @staticmethod
     def _build_entity_lines(members: List[Dict]) -> List[str]:
-        """将实体列表格式化为 prompt 行，包含 name、aliases、description。"""
+        """将实体列表格式化为 prompt 行，包含 name、aliases、description、example。"""
         lines = []
         for m in members:
             m_name = m.get("name", "")
             aliases = m.get("aliases") or []
             description = m.get("description") or ""
+            example = m.get("example") or ""
             aliases_str = f"（别名：{'、'.join(aliases)}）" if aliases else ""
             desc_str = f"：{description}" if description else ""
-            lines.append(f"- {m_name}{aliases_str}{desc_str}")
+            example_str = f"（示例：{example}）" if example else ""
+            lines.append(f"- {m_name}{aliases_str}{desc_str}{example_str}")
         return lines
 
     async def _generate_community_metadata(
@@ -481,11 +483,24 @@ class LabelPropagationEngine:
             core_entities = [m["name"] for m in sorted_members[:CORE_ENTITY_LIMIT] if m.get("name")]
 
             entity_list_str = "\n".join(self._build_entity_lines(members))
+
+            # 方案四：注入社区内实体间关系三元组
+            relationships = await self.repo.get_community_relationships(cid, end_user_id)
+            rel_lines = [
+                f"- {r['subject']} → {r['predicate']} → {r['object']}"
+                for r in relationships
+                if r.get("subject") and r.get("predicate") and r.get("object")
+            ]
+            rel_section = (
+                f"\n实体间关系：\n" + "\n".join(rel_lines)
+                if rel_lines else ""
+            )
+
             prompt = (
-                f"以下是一组语义相关的实体：\n{entity_list_str}\n\n"
+                f"以下是一组语义相关的实体：\n{entity_list_str}{rel_section}\n\n"
                 f"请为这组实体所代表的主题：\n"
                 f"1. 起一个简洁的中文名称（不超过10个字）\n"
-                f"2. 写一句话摘要（不超过50个字）\n\n"
+                f"2. 写一句话摘要（不超过80个字）\n\n"
                 f"严格按以下格式输出，不要有其他内容：\n"
                 f"名称：<名称>\n摘要：<摘要>"
             )

api/app/repositories/neo4j/community_repository.py

@@ -17,6 +17,7 @@ from app.repositories.neo4j.cypher_queries import (
     GET_ALL_ENTITY_IDS_FOR_USER,
     GET_ENTITIES_PAGE,
     GET_COMMUNITY_MEMBERS,
+    GET_COMMUNITY_RELATIONSHIPS,
     GET_ALL_COMMUNITY_MEMBERS_BATCH,
     GET_ALL_ENTITY_NEIGHBORS_BATCH,
     GET_ENTITY_NEIGHBORS_BATCH_FOR_IDS,
@@ -177,7 +178,7 @@ class CommunityRepository:
     async def get_community_members(
         self, community_id: str, end_user_id: str
     ) -> List[Dict]:
-        """查询社区成员列表。"""
+        """查询社区成员列表（含 example 字段）。"""
         try:
             return await self.connector.execute_query(
                 GET_COMMUNITY_MEMBERS,
@@ -188,6 +189,20 @@ class CommunityRepository:
             logger.error(f"get_community_members failed: {e}")
             return []
 
+    async def get_community_relationships(
+        self, community_id: str, end_user_id: str
+    ) -> List[Dict]:
+        """查询社区内实体间的关系三元组（subject, predicate, object）。"""
+        try:
+            return await self.connector.execute_query(
+                GET_COMMUNITY_RELATIONSHIPS,
+                community_id=community_id,
+                end_user_id=end_user_id,
+            )
+        except Exception as e:
+            logger.error(f"get_community_relationships failed: {e}")
+            return []
+
     async def get_all_community_members_batch(
         self, community_ids: List[str], end_user_id: str
     ) -> Dict[str, List[Dict]]:

api/app/repositories/neo4j/cypher_queries.py

@@ -1137,10 +1137,20 @@ MATCH (e:ExtractedEntity {end_user_id: $end_user_id})-[:BELONGS_TO_COMMUNITY]->(
 RETURN e.id AS id, e.name AS name, e.entity_type AS entity_type,
        e.importance_score AS importance_score, e.activation_value AS activation_value,
        e.name_embedding AS name_embedding,
-       e.aliases AS aliases, e.description AS description
+       e.aliases AS aliases, e.description AS description,
+       e.example AS example
 ORDER BY coalesce(e.activation_value, 0) DESC
 """
 
+GET_COMMUNITY_RELATIONSHIPS = """
+MATCH (e1:ExtractedEntity {end_user_id: $end_user_id})-[:BELONGS_TO_COMMUNITY]->(c:Community {community_id: $community_id})
+MATCH (e2:ExtractedEntity {end_user_id: $end_user_id})-[:BELONGS_TO_COMMUNITY]->(c)
+MATCH (e1)-[r:EXTRACTED_RELATIONSHIP]->(e2)
+RETURN e1.name AS subject, r.predicate AS predicate, e2.name AS object
+ORDER BY e1.name, r.predicate, e2.name
+LIMIT 20
+"""
+
 GET_ALL_COMMUNITY_MEMBERS_BATCH = """
 MATCH (e:ExtractedEntity {end_user_id: $end_user_id})-[:BELONGS_TO_COMMUNITY]->(c:Community)
 RETURN c.community_id AS community_id,

api/app/repositories/neo4j/graph_saver.py

@@ -1,4 +1,5 @@
 import asyncio
+import os
 from typing import List, Optional
 
 # 使用新的仓储层