This commit is contained in:
@@ -3,142 +3,151 @@ Mem0 记忆层客户端封装模块
|
||||
负责 Mem0 的初始化、检索和存储
|
||||
"""
|
||||
|
||||
import os
|
||||
import asyncio
|
||||
from typing import Optional, List, Dict, Any
|
||||
from mem0 import AsyncMemory
|
||||
|
||||
# 本地模块
|
||||
from app.config import QDRANT_URL, QDRANT_COLLECTION_NAME, VLLM_EMBEDDING_URL
|
||||
from app.config import QDRANT_URL, QDRANT_COLLECTION_NAME, LLAMACPP_EMBEDDING_URL, LLAMACPP_API_KEY
|
||||
from app.logger import info, warning, error
|
||||
|
||||
|
||||
class Mem0Client:
|
||||
"""Mem0 异步客户端封装类"""
|
||||
|
||||
|
||||
def __init__(self, llm_instance):
|
||||
"""
|
||||
初始化 Mem0 客户端
|
||||
|
||||
|
||||
Args:
|
||||
llm_instance: LangChain LLM 实例（用于事实提取）
|
||||
"""
|
||||
self.llm = llm_instance
|
||||
self.mem0: Optional[AsyncMemory] = None
|
||||
self._initialized = False
|
||||
|
||||
|
||||
async def initialize(self):
|
||||
"""异步初始化 Mem0 客户端"""
|
||||
"""异步初始化 Mem0 客户端，并进行实际连接测试"""
|
||||
if self._initialized:
|
||||
return
|
||||
|
||||
try:
|
||||
# 检查 Qdrant 是否可达 (可选)
|
||||
import requests
|
||||
try:
|
||||
resp = requests.get(f"{QDRANT_URL}/collections", timeout=2)
|
||||
if resp.status_code == 200:
|
||||
info(f"✅ Qdrant 服务正常: {QDRANT_URL}")
|
||||
except Exception:
|
||||
warning(f"⚠️ 无法连接到 Qdrant: {QDRANT_URL}，Mem0 将尝试自动连接")
|
||||
|
||||
try:
|
||||
# Mem0 配置
|
||||
config = {
|
||||
# 向量存储：复用 Qdrant 实例
|
||||
"vector_store": {
|
||||
"provider": "qdrant",
|
||||
"config": {
|
||||
"url": QDRANT_URL, # 直接使用完整 URL
|
||||
"collection_name": QDRANT_COLLECTION_NAME,
|
||||
"host": QDRANT_URL.split("://")[1].split(":")[0] if "://" in QDRANT_URL else "localhost",
|
||||
"port": int(QDRANT_URL.split(":")[-1]) if ":" in QDRANT_URL.split("://")[-1] else 6333,
|
||||
"embedding_model_dims": 768, # embeddinggemma-300m 输出 768 维
|
||||
"embedding_model_dims": 768,
|
||||
}
|
||||
},
|
||||
# 事实提取 LLM：直接复用传入的 LangChain 实例
|
||||
"llm": {
|
||||
"provider": "langchain",
|
||||
"config": {
|
||||
"model": self.llm # 直接传入 LangChain 模型实例
|
||||
"model": self.llm
|
||||
}
|
||||
},
|
||||
# Embedding：指向 vLLM 服务
|
||||
"embedder": {
|
||||
"provider": "openai",
|
||||
"embedding_dims": 768, # 关键：将维度参数提升到顶层
|
||||
"config": {
|
||||
"model": "google/embeddinggemma-300m",
|
||||
"api_key": "EMPTY",
|
||||
"api_base": VLLM_EMBEDDING_URL,
|
||||
# 注意：不要在此处传递 dimensions 参数，避免与 vLLM v0.7.2 不兼容
|
||||
}
|
||||
"model": "embeddinggemma-300M-Q8_0",
|
||||
"api_key": LLAMACPP_API_KEY,
|
||||
"openai_base_url": LLAMACPP_EMBEDDING_URL,
|
||||
},
|
||||
},
|
||||
"version": "v1.1"
|
||||
}
|
||||
|
||||
self.mem0 = AsyncMemory.from_config(config)
|
||||
self._initialized = True
|
||||
info(f"✅ Mem0 初始化成功 (Embedding: vLLM@8002, Vector: Qdrant, LLM: 复用现有实例)")
|
||||
info("✅ Mem0 配置加载成功，开始连接测试...")
|
||||
|
||||
except Exception as e:
|
||||
error(f"❌ Mem0 初始化失败: {e}")
|
||||
import traceback
|
||||
traceback.print_exc()
|
||||
# 实际连接测试：调用一次 search 确保 Qdrant 和 Embedding 都可达
|
||||
await asyncio.wait_for(
|
||||
self.mem0.search("ping", user_id="test", limit=1),
|
||||
timeout=60.0
|
||||
)
|
||||
info("✅ Mem0 实际连接测试成功，初始化完成")
|
||||
self._initialized = True
|
||||
|
||||
except asyncio.TimeoutError:
|
||||
error("❌ Mem0 连接测试超时 (10s)，请检查 Qdrant 或 Embedding 服务响应")
|
||||
self.mem0 = None
|
||||
|
||||
self._initialized = False
|
||||
except Exception as e:
|
||||
error(f"❌ Mem0 初始化或连接测试失败: {e}")
|
||||
import traceback
|
||||
error(f"详细错误信息:\n{traceback.format_exc()}")
|
||||
self.mem0 = None
|
||||
self._initialized = False
|
||||
|
||||
async def search_memories(self, query: str, user_id: str, limit: int = 5) -> List[str]:
|
||||
"""
|
||||
检索相关记忆
|
||||
|
||||
|
||||
Args:
|
||||
query: 查询文本
|
||||
user_id: 用户 ID
|
||||
limit: 返回结果数量限制
|
||||
|
||||
|
||||
Returns:
|
||||
List[str]: 记忆事实列表
|
||||
"""
|
||||
if not self.mem0:
|
||||
warning("⚠️ Mem0 未初始化，跳过记忆检索")
|
||||
return []
|
||||
|
||||
|
||||
try:
|
||||
memories = await self.mem0.search(query, user_id=user_id, limit=limit)
|
||||
|
||||
memories = await asyncio.wait_for(
|
||||
self.mem0.search(query, user_id=user_id, limit=limit),
|
||||
timeout=30.0
|
||||
)
|
||||
|
||||
if memories and "results" in memories:
|
||||
facts = [m["memory"] for m in memories["results"] if m.get("memory")]
|
||||
if facts:
|
||||
info(f"🔍 [记忆检索] Mem0 返回 {len(facts)} 条记忆")
|
||||
return facts
|
||||
|
||||
|
||||
info("🔍 [记忆检索] 未找到相关记忆")
|
||||
return []
|
||||
|
||||
|
||||
except asyncio.TimeoutError:
|
||||
warning("⚠️ Mem0 检索超时 (30s)，跳过本次记忆检索")
|
||||
return []
|
||||
except Exception as e:
|
||||
warning(f"⚠️ Mem0 检索失败: {e}")
|
||||
return []
|
||||
|
||||
|
||||
async def add_memories(self, messages: List[Dict[str, str]], user_id: str) -> bool:
|
||||
"""
|
||||
添加记忆（自动提取事实并存储）
|
||||
|
||||
|
||||
Args:
|
||||
messages: 消息列表，格式为 [{"role": "user/assistant/system", "content": "..."}]
|
||||
user_id: 用户 ID
|
||||
|
||||
|
||||
Returns:
|
||||
bool: 是否成功
|
||||
"""
|
||||
if not self.mem0:
|
||||
warning("⚠️ Mem0 未初始化，跳过记忆添加")
|
||||
return False
|
||||
|
||||
|
||||
try:
|
||||
result = await self.mem0.add(
|
||||
messages,
|
||||
user_id=user_id,
|
||||
metadata={"type": "conversation"}
|
||||
await asyncio.wait_for(
|
||||
self.mem0.add(
|
||||
messages,
|
||||
user_id=user_id,
|
||||
metadata={"type": "conversation"}
|
||||
),
|
||||
timeout=60.0
|
||||
)
|
||||
info(f"📝 [记忆添加] 已提交给 Mem0 进行事实提取")
|
||||
info("📝 [记忆添加] 已提交给 Mem0 进行事实提取")
|
||||
return True
|
||||
|
||||
|
||||
except asyncio.TimeoutError:
|
||||
error("❌ Mem0 记忆添加超时 (60s)")
|
||||
return False
|
||||
except Exception as e:
|
||||
error(f"❌ Mem0 记忆添加失败: {e}")
|
||||
return False
|
||||
return False
|
||||
Reference in New Issue
Block a user