From 673b63e4dd85b214bdca801b56b9f878998521ea Mon Sep 17 00:00:00 2001 From: Deepanshu Pal <40927968+DeepanshuPal@users.noreply.github.com> Date: Mon, 14 Sep 2026 20:21:08 +0530 Subject: [PATCH] fix: forward custom_llm_provider in Responses WebSocket provider resolution _aresponses_websocket built GenericLiteLLMParams from the router's kwargs but dropped custom_llm_provider when calling get_llm_provider, so wildcard OpenAI-compatible deployments (custom_llm_provider: openai, unprefixed upstream model names) failed with 'LLM Provider NOT provided' over /v1/responses WebSocket while the equivalent HTTP request succeeded. Fixes #41081 --- litellm/responses/main.py | 1 + .../test_responses_api_request_body.py | 24 +++++++++++++++++++ 2 files changed, 25 insertions(+) diff --git a/litellm/responses/main.py b/litellm/responses/main.py index a68cd02e61b..d96f165f306 100644 --- a/litellm/responses/main.py +++ b/litellm/responses/main.py @@ -2229,6 +2229,7 @@ async def _aresponses_websocket( dynamic_api_base, ) = litellm.get_llm_provider( model=model, + custom_llm_provider=litellm_params.custom_llm_provider, api_base=api_base, api_key=api_key, ) diff --git a/tests/test_litellm/responses/test_responses_api_request_body.py b/tests/test_litellm/responses/test_responses_api_request_body.py index 5fced458208..bfba3b5bdbe 100644 --- a/tests/test_litellm/responses/test_responses_api_request_body.py +++ b/tests/test_litellm/responses/test_responses_api_request_body.py @@ -424,6 +424,30 @@ async def test_aresponses_websocket_strips_responses_routing_prefix_from_openai_ assert mock_ws.call_args.kwargs["custom_llm_provider"] == "openai" +@pytest.mark.asyncio +async def test_aresponses_websocket_uses_custom_llm_provider_from_kwargs(): + """Router-selected deployments pass custom_llm_provider via kwargs (#41081).""" + from unittest.mock import MagicMock + + from litellm.responses.main import _aresponses_websocket + + with patch.object( + import_module("litellm.responses.main").base_llm_http_handler, "async_responses_websocket", + new_callable=AsyncMock, + ) as mock_ws: + await _aresponses_websocket( + model="model-x", + websocket=MagicMock(), + api_key="sk-test", + custom_llm_provider="openai", + litellm_logging_obj=MagicMock(), + ) + + mock_ws.assert_awaited_once() + assert mock_ws.call_args.kwargs["custom_llm_provider"] == "openai" + assert mock_ws.call_args.kwargs["model"] == "model-x" + + _INJECTION_POINT_INPUT = [{"role": "system", "content": "You are terse."}, {"role": "user", "content": "hi"}] _SYSTEM_POINT = {"location": "message", "role": "system"} _USER_POINT = {"location": "message", "role": "user"}