fix(runtime): 隔离线程任务上下文 (#6420)

* fix(runtime): isolate thread task contexts

* test: use canonical message schema import

app/adapters/network/browser.py

@@ -3,6 +3,7 @@ import threading
 import time
 import uuid
 from concurrent.futures import ThreadPoolExecutor
+from contextvars import Context, copy_context
 from dataclasses import dataclass, field
 from typing import Any, Callable, Optional, Protocol
 from urllib.parse import urlparse
@@ -288,7 +289,11 @@ class BrowserSessionHelper:
 
         executor = cls._get_existing_session_executor(session_key)
         if executor:
-            future = executor.submit(
+            context = copy_context()
+            # 会话线程保持空底层上下文，每次操作只使用当前调用快照。
+            future = Context().run(
+                executor.submit,
+                context.run,
                 cls._run_session_task,
                 session_key,
                 cls._close_session_in_thread,
@@ -387,7 +392,11 @@ class BrowserSessionHelper:
         for _ in range(2):
             executor = cls._get_session_executor(session_key)
             try:
-                future = executor.submit(
+                context = copy_context()
+                # 会话线程保持空底层上下文，每次操作只使用当前调用快照。
+                future = Context().run(
+                    executor.submit,
+                    context.run,
                     cls._run_session_task,
                     session_key,
                     callback,

app/agent/tools/base.py

@@ -4,6 +4,7 @@ import json
 import threading
 from abc import ABCMeta, abstractmethod
 from concurrent.futures import ThreadPoolExecutor
+from contextvars import Context, copy_context
 from functools import partial
 from pathlib import Path
 from typing import TYPE_CHECKING, Any, Callable, ClassVar, Optional, Protocol
@@ -225,7 +226,13 @@ async def run_agent_blocking(
 
     await semaphore.acquire()
     try:
-        future = _get_blocking_executor(bucket_name).submit(bound_call)
+        context = copy_context()
+        # 长期 worker 保持空底层上下文，每个任务只在自己的调用快照内运行。
+        future = Context().run(
+            _get_blocking_executor(bucket_name).submit,
+            context.run,
+            bound_call,
+        )
     except Exception:
         semaphore.release()
         raise

app/application/messaging/message.py

@@ -7,7 +7,9 @@ import queue
 import re
 import threading
 import time
+from contextvars import Context, copy_context
 from datetime import datetime
+from functools import partial
 from typing import Any, Literal, Optional, List, Dict, Protocol, Union
 from typing import Callable
 
@@ -959,8 +961,16 @@ class MessageQueueManager(metaclass=SingletonClass):
         if immediately or self._is_in_scheduled_time(datetime.now()):
             # _send 会执行具体渠道回调，可能包含网络 IO；放到 executor
             # 避免 async 调用方所在事件循环被同步发送阻塞。
+            context = copy_context()
+            call = partial(self._send, *args, **kwargs)
             loop = asyncio.get_running_loop()
-            await loop.run_in_executor(None, lambda: self._send(*args, **kwargs))
+            # 默认执行器保持空底层上下文，渠道调用只使用当前消息快照。
+            await Context().run(
+                loop.run_in_executor,
+                None,
+                context.run,
+                call,
+            )
             return
         self.queue.put({
             "args": args,

app/chain/workflow.py

@@ -6,7 +6,9 @@ import pickle
 import threading
 from collections import defaultdict, deque
 from concurrent.futures import ThreadPoolExecutor
+from contextvars import Context, copy_context
 from datetime import date, datetime
+from functools import partial
 from time import sleep
 from typing import Any, Callable, List, Optional, Tuple
 
@@ -371,14 +373,17 @@ class WorkflowExecutor:
                 if not node_id:
                     continue
 
-                # 提交任务到线程池，每个节点使用上下文快照，避免并行节点互相修改同一个对象。
-                future = self.executor.submit(
+                # 节点分别复制业务上下文和调用上下文，避免共享可变状态或丢失触发链路。
+                context = copy_context()
+                future = Context().run(
+                    self.executor.submit,
+                    context.run,
                     self.execute_node,
                     self.workflow.id,
                     node_id,
-                    copy.deepcopy(self.context)
+                    copy.deepcopy(self.context),
                 )
-                future.add_done_callback(self.on_node_complete)
+                future.add_done_callback(partial(context.run, self.on_node_complete))
         finally:
             self.executor.shutdown(wait=True, cancel_futures=True)