test: conn_limiter 差异覆盖与导入清理

Co-Authored-By: Claude Code <noreply@anthropic.com>

app/engine/routes/chat.py

@@ -6,7 +6,7 @@ from fastapi import APIRouter, WebSocket, WebSocketDisconnect
 from app.engine.models import ConversationSource, ConversationStatus
 from app.engine.routes.runtime import get_runtime
 from app.engine.services.conn_limiter import DEFAULT_WS_CONN_LIMIT
-from app.engine.services.conn_limiter import limiter as conn_limiter
+from app.engine.services.conn_limiter import limiter
 from app.engine.services.conversation import ConversationManager
 
 logger = logging.getLogger(__name__)
@@ -27,7 +27,7 @@ async def ws_chat(websocket: WebSocket):
 
     limit = runtime.config.data["engine"].get(
         "ws_conn_limit", DEFAULT_WS_CONN_LIMIT)
-    if not conn_limiter.try_acquire(runtime.config.slug, client_ip, limit=limit):
+    if not limiter.try_acquire(runtime.config.slug, client_ip, limit=limit):
         await websocket.close(code=1013, reason="too many connections")
         return
 
@@ -123,4 +123,4 @@ async def ws_chat(websocket: WebSocket):
     finally:
         # Task 8: 正常关闭与异常断开都归还名额——acquire 成功后所有路径
         # 必经此处,不泄漏(accept 失败/循环中抛错同样覆盖)
-        conn_limiter.release(runtime.config.slug, client_ip)
+        limiter.release(runtime.config.slug, client_ip)

app/platform/defaults.py

@@ -17,6 +17,7 @@ DEFAULT_CONFIG: dict = {
              "max_attempts": 3, "retry_backoff_base": 1.0},
     "ai": {"provider": "glm", "api_key": None, "base_url": None, "model": "glm-4-flash"},
     "engine": {"context_seed_keys": {}, "rate_limit_per_hour": 30,
+               # 双源防漂移: 默认值须与 conn_limiter.DEFAULT_WS_CONN_LIMIT 一致
                "ws_conn_limit": 10},
     "misc": {"name_to_region_extras": {}},
 }