fix: 修复 LangGraph Postgres checkpointer 初始化与 Windows 事件循环兼容

使用 psycopg_pool 作为 LangGraph checkpointer 连接池，并在启动时执行 setup() 确保表存在；Windows 下切换为 SelectorEventLoopPolicy 以兼容 psycopg 异步模式。

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "yuxi-know"
-version = "0.5.0.dev"
+version = "0.5.2"
 description = "基于大模型的智能知识库与知识图谱智能体开发平台，融合了 RAG 技术与知识图谱技术，基于 LangGraph v1 + Vue.js + FastAPI + LightRAG 架构构建"
 readme = "README.md"
 requires-python = ">=3.12,<3.14"
@@ -72,6 +72,7 @@ dependencies = [
     "redis>=5.2.0",
     "aioboto3>=13.0.0",
     "wcmatch>=8.0.0",
+    "psycopg[binary,pool]>=3.3.3",
 ]
 [tool.ruff]
 line-length = 120  # 代码最大行宽

server/main.py

@@ -1,5 +1,16 @@
 import asyncio
-import time
+import os
+import sys
+
+# ==============================================================================
+# 解决 Windows 下 psycopg 异步模式不支持 ProactorEventLoop 的问题
+# 注意：这段代码必须放在应用的极早期，最好在导入 FastAPI 或初始化数据库之前
+# ==============================================================================
+if sys.platform == "win32":
+    # 把当前文件 (main.py) 的上一级的上一级 (即根目录 Yuxi-Know) 加入到 sys.path
+    sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+    asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())
+
 from collections import defaultdict, deque
 
 import uvicorn
@@ -124,4 +135,12 @@ app.add_middleware(LoginRateLimitMiddleware)
 app.add_middleware(AuthMiddleware)
 
 if __name__ == "__main__":
-    uvicorn.run(app, host="0.0.0.0", port=5050, threads=10, workers=10, reload=True)
+    # uvicorn.run(app, host="0.0.0.0", port=5050, threads=10, workers=10, reload=True)
+
+    uvicorn.run(
+        "server.main:app",
+        host="0.0.0.0",
+        port=5050,
+        reload=True,
+        reload_dirs=["server", "src"],
+    )

server/utils/lifespan.py

@@ -1,10 +1,11 @@
 from contextlib import asynccontextmanager
 
 from fastapi import FastAPI
+from langgraph.checkpoint.postgres.aio import AsyncPostgresSaver
 
-from src.services.task_service import tasker
 from src.services.mcp_service import init_mcp_servers
 from src.services.run_queue_service import close_queue_clients, get_redis_client
+from src.services.task_service import tasker
 from src.storage.postgres.manager import pg_manager
 from src.knowledge import knowledge_base
 from src.sandbox import init_sandbox_provider, shutdown_sandbox_provider
@@ -47,6 +48,13 @@ async def lifespan(app: FastAPI):
     except Exception as e:
         logger.error(f"Failed to initialize sandbox provider during startup: {e}")
 
+    # =========================================================
+    # 2. 核心修复：在这里执行一次 setup()，建完表就拉倒
+    # =========================================================
+    checkpointer = AsyncPostgresSaver(pg_manager.langgraph_pool)
+    await checkpointer.setup()
+    print("LangGraph Checkpoint tables verified/created!")
+
     await tasker.start()
     yield
     await tasker.shutdown()

server/worker_main.py

@@ -1,5 +1,14 @@
 """ARQ worker entrypoint."""
+import asyncio
+import os
+import sys
+
+# 必须放在最顶层！
+if sys.platform == "win32":
+    # 把当前文件 (main.py) 的上一级的上一级 (即根目录 Yuxi-Know) 加入到 sys.path
+    sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+    asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())
 
 from src.services.run_worker import WorkerSettings
 
-__all__ = ["WorkerSettings"]
+__all__ = ["WorkerSettings"]

src/agents/common/base.py

@@ -12,6 +12,7 @@ from langgraph.graph.state import CompiledStateGraph
 
 from src import config as sys_config
 from src.agents.common.context import BaseContext
+from src.storage.postgres.manager import pg_manager
 from src.utils import logger
 
 
@@ -199,19 +200,9 @@ class BaseAgent:
             logger.warning(f"langgraph postgres checkpointer 不可用，回退 sqlite: {e}")
             return None
 
-        conn_str = postgres_url.replace("+asyncpg", "")
         try:
-            saver_factory = getattr(AsyncPostgresSaver, "from_conn_string", None)
+            saver = AsyncPostgresSaver(pg_manager.langgraph_pool)
-            if callable(saver_factory):
-                saver = saver_factory(conn_str)
-            else:
-                saver = AsyncPostgresSaver(conn_str)  # type: ignore[call-arg]
 
-            setup_fn = getattr(saver, "setup", None)
-            if callable(setup_fn):
-                result = setup_fn()
-                if hasattr(result, "__await__"):
-                    await result
             logger.info(f"{self.name} 使用 postgres checkpointer")
             return saver
         except Exception as e:

src/storage/postgres/manager.py

@@ -4,6 +4,7 @@ import json
 import os
 from contextlib import asynccontextmanager
 
+from psycopg_pool import AsyncConnectionPool
 from sqlalchemy import text
 from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker, create_async_engine
 from sqlalchemy.orm import declarative_base
@@ -33,6 +34,7 @@ class PostgresManager(metaclass=SingletonMeta):
     def __init__(self):
         self.async_engine = None
         self.AsyncSession = None
+        self.langgraph_pool = None
         self._initialized = False
 
     def initialize(self):
@@ -65,6 +67,21 @@ class PostgresManager(metaclass=SingletonMeta):
                 expire_on_commit=False,
             )
 
+            # ==========================================
+            # 2. 为 LangGraph 专门初始化一个原生 psycopg_pool
+            # ==========================================
+            # ⚠️ 注意：psycopg 不认识 "+asyncpg" 这样的 SQLAlchemy 方言标识。
+            # 如果你的 db_url 是 "postgresql+asyncpg://user:pwd@host/db"，
+            # 需要把它清洗成标准的 "postgresql://user:pwd@host/db"
+            langgraph_db_url = db_url.replace("+asyncpg", "").replace("+psycopg", "")
+
+            # 创建 LangGraph 专属连接池
+            self.langgraph_pool = AsyncConnectionPool(
+                conninfo=langgraph_db_url,
+                max_size=10, # 根据你的 Agent 并发情况设置，通常 5-10 足够了
+                kwargs={"autocommit": True} # LangGraph Checkpoint 强依赖 autocommit
+            )
+
             self._initialized = True
             logger.info(f"PostgreSQL manager initialized for knowledge base: {db_url.split('@')[0]}://***")
         except Exception as e:
@@ -229,6 +246,9 @@ class PostgresManager(metaclass=SingletonMeta):
         if self.async_engine:
             await self.async_engine.dispose()
 
+        if self.langgraph_pool:
+            await self.langgraph_pool.close()
+
     async def async_check_first_run(self):
         """检查是否首次运行（异步版本）- 检查用户表是否有数据"""
         from sqlalchemy import func, select