fix(proxy): strip router_settings_override on the api_key/api_base routing branch

route_request only removed router_settings_override in a later elif, so requests
carrying api_key/api_base sent it to the provider (Anthropic 400) and dropped
the key/team router settings. Pop it up front and merge its settings for every branch.

Fixes #43582

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

litellm/proxy/route_llm_request.py

@@ -423,6 +423,29 @@ RouteType = Literal[
 ]
 
 
+# Settings that the Router accepts as per-request kwargs. These override the
+# global router settings for this specific request.
+_PER_REQUEST_ROUTER_SETTINGS: Final = [
+    "fallbacks",
+    "context_window_fallbacks",
+    "content_policy_fallbacks",
+    "num_retries",
+    "timeout",
+    "model_group_retry_policy",
+    "routing_strategy",
+    "enable_tag_filtering",
+]
+
+
+def _apply_router_settings_override(data: dict, override_settings: Any) -> None:
+    """Merge key/team router settings into ``data`` (request values win)."""
+    if not isinstance(override_settings, dict):
+        return
+    for key in _PER_REQUEST_ROUTER_SETTINGS:
+        if key in override_settings and key not in data:
+            data[key] = override_settings[key]
+
+
 async def route_request(
     data: dict,
     llm_router: LitellmRouter | None,
@@ -476,6 +499,12 @@ async def _route_request_single_attempt(  # noqa: ANN202  # returns unawaited pr
 
     data.pop("enable_tag_filtering", None)
 
+    # Always remove ``router_settings_override`` from the body so it can't leak
+    # to the provider, and apply its settings on every routing branch.
+    has_router_settings_override: Final = "router_settings_override" in data
+    if has_router_settings_override:
+        _apply_router_settings_override(data, data.pop("router_settings_override"))
+
     team_id: Final = get_team_id_from_data(data)
     router_model_names: Final = llm_router.model_names if llm_router is not None else []
     is_proxy_admin_without_team: Final = team_id is None and _is_proxy_admin_request(data)
@@ -508,31 +537,8 @@ async def _route_request_single_attempt(  # noqa: ANN202  # returns unawaited pr
     elif "user_config" in data:
         return _route_user_config_request(data, route_type)
 
-    elif "router_settings_override" in data:
-        # Apply per-request router settings overrides from key/team config
-        # Instead of creating a new Router (expensive), merge settings into kwargs
-        # The Router already supports per-request overrides for these settings
-        override_settings: Final = data.pop("router_settings_override")
-
-        # Settings that the Router accepts as per-request kwargs
-        # These override the global router settings for this specific request
-        per_request_settings: Final = [
-            "fallbacks",
-            "context_window_fallbacks",
-            "content_policy_fallbacks",
-            "num_retries",
-            "timeout",
-            "model_group_retry_policy",
-            "routing_strategy",
-            "enable_tag_filtering",
-        ]
-
-        # Merge override settings into data (only if not already set in request)
-        for key in per_request_settings:
-            if key in override_settings and key not in data:
-                data[key] = override_settings[key]
-
-        # Use main router with overridden kwargs
+    elif has_router_settings_override:
+        # Settings were already merged into ``data`` above; use the main router
         if llm_router is not None:
             return getattr(llm_router, f"{route_type}")(**data)
         else:

tests/test_litellm/proxy/test_route_llm_request.py

@@ -1325,3 +1325,28 @@ def test_proxy_model_not_found_error_keeps_the_raw_model_only_in_the_client_resp
     assert raw_model in error.detail["error"]
     assert raw_model not in error.spend_log_error_message
     assert error.spend_log_error_message.startswith("/chat/completions: Invalid model name passed in")
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize("extra", [{"api_key": "sk-test"}, {"api_base": "http://localhost:1234"}])
+async def test_route_request_router_settings_override_stripped_with_api_key_or_base(extra):
+    """
+    router_settings_override must not leak to the provider when the request
+    carries api_key/api_base, and its settings must still reach the Router.
+    """
+    data = {
+        "model": "gpt-3.5-turbo",
+        "messages": [{"role": "user", "content": "Hello"}],
+        "router_settings_override": {"num_retries": 2},
+        **extra,
+    }
+
+    llm_router = MagicMock()
+    llm_router.acompletion.return_value = "success"
+
+    response = await route_request(data, llm_router, None, "acompletion")
+
+    assert response == "success"
+    call_kwargs = llm_router.acompletion.call_args[1]
+    assert "router_settings_override" not in call_kwargs
+    assert call_kwargs["num_retries"] == 2