refactor: improve PDF extraction, error handling, and proxy configuration

app/utils.py

@@ -93,13 +93,14 @@ def release_lock(db, lock) -> None:
 
 
 def make_http_client(
-    *, sync: bool = False, follow_redirects: bool = False, **kwargs
+    *, sync: bool = False, follow_redirects: bool = False, use_proxy: bool = True, **kwargs
 ) -> httpx.AsyncClient | httpx.Client:
     """创建带 proxy 和默认配置的 httpx 客户端。
 
     Args:
         sync: True 返回同步 Client，False 返回 AsyncClient
         follow_redirects: 是否跟随重定向
+        use_proxy: 是否使用 HF_PROXY 代理（默认 True；国内 API 如 embedder 应传 False）
         **kwargs: 覆盖默认参数
     """
     defaults: dict = {
@@ -107,7 +108,7 @@ def make_http_client(
         "headers": {"User-Agent": settings.HTTP_USER_AGENT},
         "follow_redirects": follow_redirects,
     }
-    if settings.http_proxy:
+    if use_proxy and settings.http_proxy:
         defaults["transport"] = (
             httpx.HTTPTransport(proxy=settings.http_proxy)
             if sync