fixes for deleting responses, response API

litellm/llms/custom_httpx/llm_http_handler.py

@@ -1259,10 +1259,10 @@ class BaseLLMHTTPHandler:
     def delete_response_api_handler(
         self,
         response_id: str,
-        custom_llm_provider: str,
         responses_api_provider_config: BaseResponsesAPIConfig,
         litellm_params: GenericLiteLLMParams,
         logging_obj: LiteLLMLoggingObj,
+        custom_llm_provider: Optional[str],
         extra_headers: Optional[Dict[str, Any]] = None,
         extra_body: Optional[Dict[str, Any]] = None,
         timeout: Optional[Union[float, httpx.Timeout]] = None,
@@ -1313,7 +1313,7 @@ class BaseLLMHTTPHandler:
 
         try:
             response = sync_httpx_client.delete(
-                url=api_base, headers=headers, data=json.dumps(data), timeout=timeout
+                url=url, headers=headers, data=json.dumps(data), timeout=timeout
             )
 
         except Exception as e:

litellm/llms/openai/responses/transformation.py

@@ -230,7 +230,7 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig):
         OpenAI API expects the following request
         - DELETE /v1/responses/{response_id}
         """
-        url = f"{api_base}/responses/{response_id}"
+        url = f"{api_base}/{response_id}"
         data = {}
         return url, data
 

litellm/responses/main.py

@@ -24,6 +24,7 @@ from litellm.types.llms.openai import (
     ToolChoice,
     ToolParam,
 )
+from litellm.types.responses.main import *
 from litellm.types.router import GenericLiteLLMParams
 from litellm.utils import ProviderConfigManager, client
 
@@ -122,6 +123,7 @@ async def aresponses(
             response = ResponsesAPIRequestUtils._update_responses_api_response_id_with_model_id(
                 responses_api_response=response,
                 kwargs=kwargs,
+                custom_llm_provider=custom_llm_provider,
             )
         return response
     except Exception as e:
@@ -260,6 +262,7 @@ def responses(
             response = ResponsesAPIRequestUtils._update_responses_api_response_id_with_model_id(
                 responses_api_response=response,
                 kwargs=kwargs,
+                custom_llm_provider=custom_llm_provider,
             )
 
         return response
@@ -271,3 +274,94 @@ def responses(
             completion_kwargs=local_vars,
             extra_kwargs=kwargs,
         )
+
+
+@client
+def delete_responses(
+    response_id: str,
+    # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
+    # The extra values given here take precedence over values defined on the client or passed to this method.
+    extra_headers: Optional[Dict[str, Any]] = None,
+    extra_query: Optional[Dict[str, Any]] = None,
+    extra_body: Optional[Dict[str, Any]] = None,
+    timeout: Optional[Union[float, httpx.Timeout]] = None,
+    # LiteLLM specific params,
+    custom_llm_provider: Optional[str] = None,
+    **kwargs,
+) -> DeleteResponseResult:
+    """
+    Synchronous version of the DELETE Responses API
+
+    DELETE /v1/responses/{response_id} endpoint in the responses API
+
+    """
+    local_vars = locals()
+    try:
+        litellm_logging_obj: LiteLLMLoggingObj = kwargs.get("litellm_logging_obj")  # type: ignore
+        litellm_call_id: Optional[str] = kwargs.get("litellm_call_id", None)
+        _is_async = kwargs.pop("aresponses", False) is True
+
+        # get llm provider logic
+        litellm_params = GenericLiteLLMParams(**kwargs)
+
+        # get custom llm provider from response_id
+        decoded_response_id: DecodedResponseId = (
+            ResponsesAPIRequestUtils._decode_responses_api_response_id(
+                response_id=response_id,
+            )
+        )
+        response_id = decoded_response_id.get("response_id") or response_id
+        custom_llm_provider = (
+            decoded_response_id.get("custom_llm_provider") or custom_llm_provider
+        )
+
+        # get provider config
+        responses_api_provider_config: Optional[BaseResponsesAPIConfig] = (
+            ProviderConfigManager.get_provider_responses_api_config(
+                model=None,
+                provider=litellm.LlmProviders(custom_llm_provider),
+            )
+        )
+
+        if responses_api_provider_config is None:
+            raise ValueError(
+                f"DELETE responses is not supported for {custom_llm_provider}"
+            )
+
+        local_vars.update(kwargs)
+
+        # Pre Call logging
+        litellm_logging_obj.update_environment_variables(
+            model=None,
+            optional_params={
+                "response_id": response_id,
+            },
+            litellm_params={
+                "litellm_call_id": litellm_call_id,
+            },
+            custom_llm_provider=custom_llm_provider,
+        )
+
+        # Call the handler with _is_async flag instead of directly calling the async handler
+        response = base_llm_http_handler.delete_response_api_handler(
+            response_id=response_id,
+            custom_llm_provider=custom_llm_provider,
+            responses_api_provider_config=responses_api_provider_config,
+            litellm_params=litellm_params,
+            logging_obj=litellm_logging_obj,
+            extra_headers=extra_headers,
+            extra_body=extra_body,
+            timeout=timeout or request_timeout,
+            _is_async=_is_async,
+            client=kwargs.get("client"),
+        )
+
+        return response
+    except Exception as e:
+        raise litellm.exception_type(
+            model=None,
+            custom_llm_provider=custom_llm_provider,
+            original_exception=e,
+            completion_kwargs=local_vars,
+            extra_kwargs=kwargs,
+        )

litellm/responses/utils.py

@@ -1,6 +1,8 @@
 import base64
 from typing import Any, Dict, Optional, Tuple, Union, cast, get_type_hints
 
+from typing_extensions import TypedDict
+
 import litellm
 from litellm._logging import verbose_logger
 from litellm.llms.base_llm.responses.transformation import BaseResponsesAPIConfig
@@ -9,6 +11,7 @@ from litellm.types.llms.openai import (
     ResponsesAPIOptionalRequestParams,
     ResponsesAPIResponse,
 )
+from litellm.types.responses.main import DecodedResponseId
 from litellm.types.utils import SpecialEnums, Usage
 
 
@@ -84,29 +87,35 @@ class ResponsesAPIRequestUtils:
     def _update_responses_api_response_id_with_model_id(
         responses_api_response: ResponsesAPIResponse,
         kwargs: Dict[str, Any],
+        custom_llm_provider: Optional[str],
     ) -> ResponsesAPIResponse:
-        """Update the responses_api_response_id with the model_id"""
+        """
+        Update the responses_api_response_id with model_id and custom_llm_provider
+
+        This builds a composite ID containing the custom LLM provider, model ID, and original response ID
+        """
         litellm_metadata: Dict[str, Any] = kwargs.get("litellm_metadata", {}) or {}
         model_info: Dict[str, Any] = litellm_metadata.get("model_info", {}) or {}
         model_id = model_info.get("id")
         updated_id = ResponsesAPIRequestUtils._build_responses_api_response_id(
             model_id=model_id,
+            custom_llm_provider=custom_llm_provider,
             response_id=responses_api_response.id,
         )
+
         responses_api_response.id = updated_id
         return responses_api_response
 
     @staticmethod
     def _build_responses_api_response_id(
+        custom_llm_provider: Optional[str],
         model_id: Optional[str],
         response_id: str,
     ) -> str:
         """Build the responses_api_response_id"""
-        if model_id is None:
-            return response_id
         assembled_id: str = str(
             SpecialEnums.LITELLM_MANAGED_RESPONSE_COMPLETE_STR.value
-        ).format(model_id, response_id)
+        ).format(custom_llm_provider, model_id, response_id)
         base64_encoded_id: str = base64.b64encode(assembled_id.encode("utf-8")).decode(
             "utf-8"
         )
@@ -115,12 +124,12 @@ class ResponsesAPIRequestUtils:
     @staticmethod
     def _decode_responses_api_response_id(
         response_id: str,
-    ) -> Tuple[Optional[str], str]:
+    ) -> DecodedResponseId:
         """
         Decode the responses_api_response_id
 
         Returns:
-            Tuple of model_id, response_id (from upstream provider)
+            DecodedResponseId: Structured tuple with custom_llm_provider, model_id, and response_id
         """
         try:
             # Remove prefix and decode
@@ -129,16 +138,45 @@ class ResponsesAPIRequestUtils:
 
             # Parse components using known prefixes
             if ";" not in decoded_id:
-                return None, response_id
+                return DecodedResponseId(
+                    custom_llm_provider=None,
+                    model_id=None,
+                    response_id=response_id,
+                )
 
-            model_part, response_part = decoded_id.split(";", 1)
+            parts = decoded_id.split(";")
-            model_id = model_part.replace("litellm:model_id:", "")
+
+            # Format: litellm:custom_llm_provider:{};model_id:{};response_id:{}
+            custom_llm_provider = None
+            model_id = None
+
+            if (
+                len(parts) >= 3
+            ):  # Full format with custom_llm_provider, model_id, and response_id
+                custom_llm_provider_part = parts[0]
+                model_id_part = parts[1]
+                response_part = parts[2]
+
+                custom_llm_provider = custom_llm_provider_part.replace(
+                    "litellm:custom_llm_provider:", ""
+                )
+                model_id = model_id_part.replace("model_id:", "")
                 decoded_response_id = response_part.replace("response_id:", "")
+            else:
+                decoded_response_id = response_id
 
-            return model_id, decoded_response_id
+            return DecodedResponseId(
+                custom_llm_provider=custom_llm_provider,
+                model_id=model_id,
+                response_id=decoded_response_id,
+            )
         except Exception as e:
             verbose_logger.debug(f"Error decoding response_id '{response_id}': {e}")
-            return None, response_id
+            return DecodedResponseId(
+                custom_llm_provider=None,
+                model_id=None,
+                response_id=response_id,
+            )
 
 
 class ResponseAPILoggingUtils:

litellm/router_utils/pre_call_checks/responses_api_deployment_check.py

@@ -31,11 +31,10 @@ class ResponsesApiDeploymentCheck(CustomLogger):
         if previous_response_id is None:
             return healthy_deployments
 
-        model_id, response_id = (
+        decoded_response = ResponsesAPIRequestUtils._decode_responses_api_response_id(
-            ResponsesAPIRequestUtils._decode_responses_api_response_id(
             response_id=previous_response_id,
         )
-        )
+        model_id = decoded_response.get("model_id")
         if model_id is None:
             return healthy_deployments
 

litellm/types/responses/main.py

@@ -1,5 +1,6 @@
 from typing import Literal
 
+from pydantic import PrivateAttr
 from typing_extensions import Any, List, Optional, TypedDict
 
 from litellm.types.llms.base import BaseLiteLLMOpenAIResponseObject
@@ -62,3 +63,14 @@ class DeleteResponseResult(BaseLiteLLMOpenAIResponseObject):
     id: Optional[str]
     object: Optional[str]
     deleted: Optional[bool]
+
+    # Define private attributes using PrivateAttr
+    _hidden_params: dict = PrivateAttr(default_factory=dict)
+
+
+class DecodedResponseId(TypedDict, total=False):
+    """Structure representing a decoded response ID"""
+
+    custom_llm_provider: Optional[str]
+    model_id: Optional[str]
+    response_id: str

litellm/types/utils.py

@@ -2254,7 +2254,9 @@ class SpecialEnums(Enum):
     LITELM_MANAGED_FILE_ID_PREFIX = "litellm_proxy"
     LITELLM_MANAGED_FILE_COMPLETE_STR = "litellm_proxy:{};unified_id,{}"
 
-    LITELLM_MANAGED_RESPONSE_COMPLETE_STR = "litellm:model_id:{};response_id:{}"
+    LITELLM_MANAGED_RESPONSE_COMPLETE_STR = (
+        "litellm:custom_llm_provider:{};model_id:{};response_id:{}"
+    )
 
 
 LLMResponseTypes = Union[

litellm/utils.py

@@ -516,9 +516,9 @@ def function_setup(  # noqa: PLR0915
         function_id: Optional[str] = kwargs["id"] if "id" in kwargs else None
 
         ## DYNAMIC CALLBACKS ##
-        dynamic_callbacks: Optional[
+        dynamic_callbacks: Optional[List[Union[str, Callable, CustomLogger]]] = (
-            List[Union[str, Callable, CustomLogger]]
+            kwargs.pop("callbacks", None)
-        ] = kwargs.pop("callbacks", None)
+        )
         all_callbacks = get_dynamic_callbacks(dynamic_callbacks=dynamic_callbacks)
 
         if len(all_callbacks) > 0:
@@ -1202,9 +1202,9 @@ def client(original_function):  # noqa: PLR0915
                         exception=e,
                         retry_policy=kwargs.get("retry_policy"),
                     )
-                    kwargs[
+                    kwargs["retry_policy"] = (
-                        "retry_policy"
+                        reset_retry_policy()
-                    ] = reset_retry_policy()  # prevent infinite loops
+                    )  # prevent infinite loops
                 litellm.num_retries = (
                     None  # set retries to None to prevent infinite loops
                 )
@@ -3028,16 +3028,16 @@ def get_optional_params(  # noqa: PLR0915
                     True  # so that main.py adds the function call to the prompt
                 )
                 if "tools" in non_default_params:
-                    optional_params[
+                    optional_params["functions_unsupported_model"] = (
-                        "functions_unsupported_model"
+                        non_default_params.pop("tools")
-                    ] = non_default_params.pop("tools")
+                    )
                     non_default_params.pop(
                         "tool_choice", None
                     )  # causes ollama requests to hang
                 elif "functions" in non_default_params:
-                    optional_params[
+                    optional_params["functions_unsupported_model"] = (
-                        "functions_unsupported_model"
+                        non_default_params.pop("functions")
-                    ] = non_default_params.pop("functions")
+                    )
             elif (
                 litellm.add_function_to_prompt
             ):  # if user opts to add it to prompt instead
@@ -3060,11 +3060,11 @@ def get_optional_params(  # noqa: PLR0915
 
     if "response_format" in non_default_params:
         if provider_config is not None:
-            non_default_params[
+            non_default_params["response_format"] = (
-                "response_format"
+                provider_config.get_json_schema_from_pydantic_object(
-            ] = provider_config.get_json_schema_from_pydantic_object(
                     response_format=non_default_params["response_format"]
                 )
+            )
         else:
             non_default_params["response_format"] = type_to_response_format_param(
                 response_format=non_default_params["response_format"]
@@ -4079,9 +4079,9 @@ def _count_characters(text: str) -> int:
 
 
 def get_response_string(response_obj: Union[ModelResponse, ModelResponseStream]) -> str:
-    _choices: Union[
+    _choices: Union[List[Union[Choices, StreamingChoices]], List[StreamingChoices]] = (
-        List[Union[Choices, StreamingChoices]], List[StreamingChoices]
+        response_obj.choices
-    ] = response_obj.choices
+    )
 
     response_str = ""
     for choice in _choices:
@@ -6625,8 +6625,8 @@ class ProviderConfigManager:
 
     @staticmethod
     def get_provider_responses_api_config(
-        model: str,
         provider: LlmProviders,
+        model: Optional[str] = None,
     ) -> Optional[BaseResponsesAPIConfig]:
         if litellm.LlmProviders.OPENAI == provider:
             return litellm.OpenAIResponsesAPIConfig()

tests/llm_responses_api_testing/base_responses_api.py

@@ -189,6 +189,75 @@ class BaseResponsesAPITest(ABC):
 
 
 
+    @pytest.mark.parametrize("sync_mode", [True, False])
+    @pytest.mark.asyncio
+    async def test_basic_openai_responses_delete_endpoint(self, sync_mode):
+        litellm._turn_on_debug()
+        litellm.set_verbose = True
+        base_completion_call_args = self.get_base_completion_call_args()
+        if sync_mode:
+            response = litellm.responses(
+                input="Basic ping", max_output_tokens=20,
+                **base_completion_call_args
+            )
+
+            # delete the response
+            if isinstance(response, ResponsesAPIResponse):
+                litellm.delete_responses(
+                    response_id=response.id,
+                )
+            else:
+                raise ValueError("response is not a ResponsesAPIResponse")
+        # else:
+        #     response = await litellm.aresponses(
+        #         input="Basic ping", max_output_tokens=20,
+        #         **base_completion_call_args
+        #     )
+
+        #     # async delete the response
+        #     await litellm.adelete_responses(
+        #         response_id=response.id,
+        #     )
+    
+
+    # @pytest.mark.parametrize("sync_mode", [True, False])
+    # @pytest.mark.asyncio
+    # async def test_basic_openai_responses_streaming_delete_endpoint(self, sync_mode):
+    #     litellm._turn_on_debug()
+    #     litellm.set_verbose = True
+    #     base_completion_call_args = self.get_base_completion_call_args()
+    #     if sync_mode:
+    #         response_id = None
+    #         response = litellm.responses(
+    #             input="Basic ping", max_output_tokens=20,
+    #             stream=True,
+    #             **base_completion_call_args
+    #         )
+    #         for event in response:
+    #             if event.type == "response.completed":
+    #                 response_id = event.response.id
+    #                 break
+
+    #         # delete the response
+    #         litellm.delete_responses(
+    #             response_id=response_id,
+    #         )
+    #     else:
+    #         response = await litellm.aresponses(
+    #             input="Basic ping", max_output_tokens=20,
+    #             stream=True,
+    #             **base_completion_call_args
+    #         )
+    #         async for event in response:
+    #             if event.type == "response.completed":
+    #                 response_id = event.response.id
+    #                 break
+
+    #         # async delete the response
+    #         await litellm.adelete_responses(
+    #             response_id=response_id,
+    #         )
+