diff --git a/litellm/responses/litellm_completion_transformation/transformation.py b/litellm/responses/litellm_completion_transformation/transformation.py index ce8841c9190..dc91445d6ac 100644 --- a/litellm/responses/litellm_completion_transformation/transformation.py +++ b/litellm/responses/litellm_completion_transformation/transformation.py @@ -1214,6 +1214,11 @@ class LiteLLMCompletionResponsesConfig: "metadata": kwargs.get("metadata"), "service_tier": responses_api_request.get("service_tier"), "safety_identifier": responses_api_request.get("safety_identifier"), + "prompt_cache_key": responses_api_request.get("prompt_cache_key"), + "prompt_cache_retention": responses_api_request.get("prompt_cache_retention"), + "verbosity": ( + text_param.get("verbosity") if isinstance(text_param, dict) else None + ), "web_search_options": web_search_options, "response_format": response_format, "reasoning_effort": reasoning.effort, diff --git a/tests/unit/responses/litellm_completion_transformation/test_litellm_completion_responses.py b/tests/unit/responses/litellm_completion_transformation/test_litellm_completion_responses.py index 4250c78c669..bef04ff1bb5 100644 --- a/tests/unit/responses/litellm_completion_transformation/test_litellm_completion_responses.py +++ b/tests/unit/responses/litellm_completion_transformation/test_litellm_completion_responses.py @@ -1288,6 +1288,27 @@ class TestFunctionCallTransformation: assert result["safety_identifier"] == "user-7f3a" + def test_text_verbosity_forwarded_as_top_level_verbosity(self) -> None: + result: Final = LiteLLMCompletionResponsesConfig.transform_responses_api_request_to_chat_completion_request( + model="openai/gpt-5.6", + input="Explain what an HTTP 400 is.", + responses_api_request={"text": {"verbosity": "high"}}, + custom_llm_provider="openai", + ) + + assert result["verbosity"] == "high" + + def test_prompt_cache_key_forwarded_to_chat_completion_request(self) -> None: + result: Final = LiteLLMCompletionResponsesConfig.transform_responses_api_request_to_chat_completion_request( + model="openai/gpt-5.6", + input="hi", + responses_api_request={"prompt_cache_key": "sess-abc", "prompt_cache_retention": "24h"}, + custom_llm_provider="openai", + ) + + assert result["prompt_cache_key"] == "sess-abc" + assert result["prompt_cache_retention"] == "24h" + def test_parallel_tool_calls_dropped_when_no_chat_tools_remain(self) -> None: transform: Final = LiteLLMCompletionResponsesConfig.transform_responses_api_request_to_chat_completion_request hosted_tool_search: Final = {"type": "tool_search", "execution": "server"}