Merge pull request 'fix(avatar): 部署重启后自动恢复 BOXIM 接管' (#6) from codex/avatar-takeover-restart-safe-20260826 into main

Reviewed-on: #6

digital-avatar-app/backend/services/takeover_service.py

@@ -201,13 +201,20 @@ class TakeoverService:
     def _forget_boxim_session(self, user_id: str):
         self._sessions.pop(user_id, None)
 
-    def _disable_after_connection_failure(
+    def _record_connection_failure(
         self,
         db: Session,
         avatar: Avatar,
         cursor: TakeoverCursor,
         message: str,
+        *,
+        disable_takeover: bool,
     ):
+        cursor.last_error = message
+        cursor.last_polled_at = self.now()
+        if not disable_takeover:
+            return
+
         permissions = (avatar.config or {}).get("authorizationPermissions", [])
         avatar.config = {
             **(avatar.config or {}),
@@ -217,8 +224,6 @@ class TakeoverService:
                 if permission != TAKEOVER_PERMISSION
             ],
         }
-        cursor.last_error = message
-        cursor.last_polled_at = self.now()
         tasks = (
             db.query(TakeoverReplyTask)
             .filter(
@@ -245,11 +250,12 @@ class TakeoverService:
                 db.add(cursor)
                 db.flush()
             if not user or not user.huihui_token:
-                self._disable_after_connection_failure(
+                self._record_connection_failure(
                     db,
                     avatar,
                     cursor,
                     "请重新登录会会生产账号后再开启主动接管",
+                    disable_takeover=True,
                 )
                 db.commit()
                 return False
@@ -268,11 +274,24 @@ class TakeoverService:
                 if isinstance(exc, BoxIMError) and exc.auth_error:
                     self._forget_boxim_session(user.id)
                     message = "BOXIM 授权已失效，请重新登录会会生产账号"
+                    disable_takeover = True
                 else:
                     message = f"BOXIM 暂时连接失败：{str(exc)[:160]}"
-                self._disable_after_connection_failure(db, avatar, cursor, message)
+                    disable_takeover = False
+                self._record_connection_failure(
+                    db,
+                    avatar,
+                    cursor,
+                    message,
+                    disable_takeover=disable_takeover,
+                )
                 db.commit()
-                logger.warning("BOXIM sync failed for avatar %s: %s", avatar.id, exc)
+                logger.warning(
+                    "BOXIM sync failed for avatar %s (will_retry=%s): %s",
+                    avatar.id,
+                    not disable_takeover,
+                    exc,
+                )
                 return False
 
             messages.sort(key=lambda item: (_numeric_id(item.get("id")), item.get("sendTime") or 0))

digital-avatar-app/backend/tests/test_takeover_service.py

@@ -291,5 +291,41 @@ async def test_connection_failure_disables_takeover_and_stops_retrying(service_c
     boxim.exchange_access_token.assert_awaited_once_with("prod-huihui-token")
 
 
+@pytest.mark.asyncio
+async def test_transient_connection_failure_keeps_takeover_and_recovers(service_context):
+    session_factory, service, boxim, _ = service_context
+    boxim.exchange_access_token = AsyncMock(
+        side_effect=[
+            BoxIMError("连接超时"),
+            {"accessToken": "box-token", "accessTokenExpiresIn": 3600},
+        ]
+    )
+
+    await service.poll_messages()
+
+    db = session_factory()
+    try:
+        avatar = db.query(Avatar).one()
+        cursor = db.query(TakeoverCursor).one()
+        assert "takeover" in avatar.config["authorizationPermissions"]
+        assert cursor.initialized is False
+        assert "暂时连接失败" in cursor.last_error
+    finally:
+        db.close()
+
+    await service.poll_messages()
+
+    db = session_factory()
+    try:
+        avatar = db.query(Avatar).one()
+        cursor = db.query(TakeoverCursor).one()
+        assert "takeover" in avatar.config["authorizationPermissions"]
+        assert cursor.initialized is True
+        assert cursor.last_error == ""
+    finally:
+        db.close()
+    assert boxim.exchange_access_token.await_count == 2
+
+
 def test_plain_text_reply_removes_markdown_and_empty_lines():
     assert _plain_text_reply("## 建议\n\n**不能自行用药**\n`必要时就医`") == "建议\n不能自行用药\n必要时就医"