refactor(memory): use MemorySummary node count for implicit memory metrics

- Replace Statement-based implicit memory count (count/3) with actual
  MemorySummary node count filtered by DERIVED_FROM_STATEMENT relationship
- Add minimum threshold of 5 MemorySummary nodes before reporting data
- Add _build_empty_profile() to return structured empty profile when
  insufficient data exists, skipping unnecessary LLM calls

api/app/services/implicit_memory_service.py

@@ -379,12 +379,59 @@ 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.now().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 +441,20 @@ 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）
+            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=user_id)
+            memory_summary_count = result[0]["count"] if result and len(result) > 0 else 0
+            logger.info(f"用户 MemorySummary 节点数量: {memory_summary_count} (user={user_id})")
+
+            if memory_summary_count < 5:
+                logger.info(f"MemorySummary 数量不足 5（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),