diff --git a/litellm/responses/litellm_completion_transformation/handler.py b/litellm/responses/litellm_completion_transformation/handler.py index e8a2e8edaad..12a91647764 100644 --- a/litellm/responses/litellm_completion_transformation/handler.py +++ b/litellm/responses/litellm_completion_transformation/handler.py @@ -28,6 +28,14 @@ from litellm.types.utils import ModelResponse _REQUEST_KWARGS_ADAPTER: Final = TypeAdapter(dict[str, object]) +def _chat_verbosity(responses_api_request: Mapping[str, object]) -> Mapping[str, object]: + match responses_api_request.get("text"): + case {"verbosity": str() as verbosity}: + return {"verbosity": verbosity} + case _: + return {} + + class LiteLLMCompletionTransformationHandler: def response_api_handler( self, @@ -52,8 +60,6 @@ class LiteLLMCompletionTransformationHandler: bridged_request: Final[ResponsesAPIOptionalRequestParams] = ( {**responses_api_request, "tools": list(hoisted.tools)} if hoisted.hoisted else responses_api_request ) - text_config: Final = bridged_request.get("text") - verbosity: Final = text_config.get("verbosity") if isinstance(text_config, Mapping) else None litellm_completion_request: Final[dict] = { **LiteLLMCompletionResponsesConfig.transform_responses_api_request_to_chat_completion_request( model=model, @@ -64,7 +70,7 @@ class LiteLLMCompletionTransformationHandler: extra_headers=extra_headers, **kwargs, ), - **({"verbosity": verbosity} if verbosity is not None else {}), + **_chat_verbosity(bridged_request), } if _is_async: