'commit'

dsLightRag/Volcengine/Kit/VikingDBMemoryService.py

@@ -212,70 +212,6 @@ class VikingDBMemoryService(Service):
         res = self.json("AddSession", {}, json.dumps(params))
         return json.loads(res)
 
-    def handle_conversation_turn(self, llm_client, user_id, user_message, conversation_history):
-        """处理一轮对话，包括记忆搜索和LLM响应。"""
-        logger.info("\n" + "=" * 60)
-        logger.info(f"用户: {user_message}")
-
-        # 修复：调用正确的search_relevant_memories方法
-        relevant_memories = self.search_relevant_memories(MEMORY_COLLECTION_NAME, user_id, user_message)
-
-        system_prompt = "你是一个富有同情心、善于倾听的AI伙伴，拥有长期记忆能力。你的目标是为用户提供情感支持和温暖的陪伴。"
-        if relevant_memories:
-            memory_context = "\n".join(
-                [f"- {json.dumps(mem['memory_info'], ensure_ascii=False)}" for mem in relevant_memories])
-            system_prompt += f"\n\n这是我们过去的一些对话记忆，请参考：\n{memory_context}\n\n请利用这些信息来更好地理解和回应用户。"
-
-        logger.info("AI正在思考...")
-
-        try:
-            messages = [{"role": "system", "content": system_prompt}] + conversation_history + [
-                {"role": "user", "content": user_message}]
-            completion = llm_client.chat.completions.create(
-                model="doubao-seed-1-6-flash-250715",
-                messages=messages
-            )
-            assistant_reply = completion.choices[0].message.content
-        except Exception as e:
-            logger.info(f"LLM调用失败: {e}")
-            assistant_reply = "抱歉，我现在有点混乱，无法回应。我们可以稍后再聊吗？"
-
-        logger.info(f"伙伴: {assistant_reply}")
-
-        conversation_history.extend([
-            {"role": "user", "content": user_message},
-            {"role": "assistant", "content": assistant_reply}
-        ])
-        return assistant_reply
-
-    def archive_conversation(self, user_id, assistant_id, conversation_history, topic_name):
-        """将对话历史归档到记忆数据库。"""
-        if not conversation_history:
-            logger.info("没有对话可以归档。")
-            return False
-
-        logger.info(f"\n正在归档关于 '{topic_name}' 的对话...")
-        session_id = f"{topic_name}_{int(time.time())}"
-        metadata = {
-            "default_user_id": user_id,
-            "default_assistant_id": assistant_id,
-            "time": int(time.time() * 1000)
-        }
-
-        try:
-            self.add_session(
-                collection_name=MEMORY_COLLECTION_NAME,
-                session_id=session_id,
-                messages=conversation_history,
-                metadata=metadata
-            )
-            logger.info(f"对话已成功归档，会话ID: {session_id}")
-            logger.info("正在等待记忆索引更新...")
-            return True
-        except Exception as e:
-            logger.info(f"归档对话失败: {e}")
-            return False
-
     def wait_for_collection_ready(self, timeout=300, interval=10):
         """
         等待集合准备就绪
@@ -352,52 +288,6 @@ class VikingDBMemoryService(Service):
                 logger.info(f"检查集合时出错: {e}")
                 raise
 
-    def search_relevant_memories(self, collection_name, user_id, query, limit=3):
-        """搜索与用户查询相关的记忆，并在索引构建中时重试。"""
-        logger.info(f"正在搜索与 '{query}' 相关的记忆...")
-        retry_attempt = 0
-        while True:
-            try:
-                filter_params = {
-                    "user_id": user_id,  # 修正为字符串类型
-                    "memory_type": ["sys_event_v1", "sys_profile_v1"]
-                }
-                response = self.search_memory(
-                    collection_name=collection_name,
-                    query=query,
-                    filter=filter_params,
-                    limit=limit
-                )
-
-                memories = []
-                if response.get('data', {}).get('count', 0) > 0:
-                    for result in response['data']['result_list']:
-                        if 'memory_info' in result and result['memory_info']:
-                            memories.append({
-                                'memory_info': result['memory_info'],
-                                'score': result['score']
-                            })
-
-                if memories:
-                    if retry_attempt > 0:
-                        logger.info("重试后搜索成功。")
-                    logger.info(f"找到 {len(memories)} 条相关记忆:")
-                    for i, memory in enumerate(memories, 1):
-                        logger.info(
-                            f"  {i}. (相关度: {memory['score']:.3f}): {json.dumps(memory['memory_info'], ensure_ascii=False, indent=2)}")
-                else:
-                    logger.info("未找到相关记忆。")
-                return memories
-
-            except Exception as e:
-                error_message = str(e)
-                if "1000023" in error_message:
-                    retry_attempt += 1
-                    logger.info(f"记忆索引正在构建中。将在60秒后重试... (尝试次数 {retry_attempt})")
-                    time.sleep(60)
-                else:
-                    logger.info(f"搜索记忆时出错 (不可重试): {e}")
-                    return []
 
     def setup_memory_collection(self):
         """独立封装记忆体创建逻辑，返回memory_service供测试使用"""
@@ -441,7 +331,7 @@ def initialize_services():
     return memory_service, llm_client
 
 
-def search_relevant_memories(memory_service, collection_name, user_id, query):
+def search_relevant_memories(memory_service, collection_name, user_id, assistant_id, query):
     """搜索与用户查询相关的记忆，并在索引构建中时重试。"""
     logger.info(f"正在搜索与 '{query}' 相关的记忆...")
     retry_attempt = 0
@@ -449,6 +339,7 @@ def search_relevant_memories(memory_service, collection_name, user_id, query):
         try:
             filter_params = {
                 "user_id": [user_id],
+                "assistant_id": assistant_id,  # 添加assistant_id过滤条件
                 "memory_type": ["sys_event_v1", "sys_profile_v1"]
             }
             response = memory_service.search_memory(
@@ -489,12 +380,12 @@ def search_relevant_memories(memory_service, collection_name, user_id, query):
                 return []
 
 
-def handle_conversation_turn(memory_service, llm_client, collection_name, user_id, user_message, conversation_history):
+def handle_conversation_turn(memory_service, llm_client, collection_name, user_id, assistant_id, user_message, conversation_history):
     """处理一轮对话，包括记忆搜索和LLM响应。"""
     logger.info("\n" + "=" * 60)
     logger.info(f"用户: {user_message}")
 
-    relevant_memories = search_relevant_memories(memory_service, collection_name, user_id, user_message)
+    relevant_memories = search_relevant_memories(memory_service, collection_name, user_id, assistant_id, user_message)
 
     system_prompt = "你是一个富有同情心、善于倾听的AI伙伴，拥有长期记忆能力。你的目标是为用户提供情感支持和温暖的陪伴。"
     if relevant_memories:

dsLightRag/Volcengine/T3_ChatWithMemory.py

@@ -22,10 +22,13 @@ def main():
     try:
         # 使用initialize_services函数初始化服务和LLM客户端
         memory_service, llm_client = initialize_services()
-
+        # 集合名称【数据库名】
         collection_name = MEMORY_COLLECTION_NAME
-        user_id = "liming"  # 用户李明
-        assistant_id = "assistant"  # 助手ID:助手
+
+        # 用户李明
+        user_id = "liming"
+        # 助手ID:助手
+        assistant_id = "assistant"
 
         # 告知大模型用户信息
         logger.info("告知大模型用户信息...")
@@ -39,12 +42,13 @@ def main():
 
         # 使用正确的handle_conversation_turn方法参数
         response = handle_conversation_turn(
-            memory_service=memory_service,
-            llm_client=llm_client,
-            collection_name=collection_name,
-            user_id=user_id,
-            user_message=f"请记住以下用户信息：{user_info}",
-            conversation_history=conversation_history
+            memory_service=memory_service,# 内存记忆服务
+            llm_client=llm_client,#  大模型客户端
+            collection_name=collection_name,# 集合名称
+            user_id=user_id,# 用户ID
+            assistant_id=assistant_id,# 助手ID
+            user_message=f"请记住以下用户信息：{user_info}",# 要记忆的信息
+            conversation_history=conversation_history # 对话历史
         )
 
         logger.info(f"模型回复: {response}")
@@ -72,6 +76,7 @@ def main():
             llm_client=llm_client,
             collection_name=collection_name,
             user_id=user_id,
+            assistant_id=assistant_id,  # 添加这一行
             user_message="请告诉我李明的个人信息",
             conversation_history=test_conversation_history
         )