feat(responses): add supports_native_file_search() capability flag to BaseResponsesAPIConfig

Add a new method supports_native_file_search() to BaseResponsesAPIConfig that lets
each provider declare whether it natively handles the file_search tool with
vector_store_ids in the Responses API.

- BaseResponsesAPIConfig: default returns False (RAG injection mode)
- OpenAIResponsesAPIConfig: returns True (native passthrough)
- AzureOpenAIResponsesAPIConfig inherits True from OpenAI via subclassing

This flag is used by VectorStorePreCallHook to decide between:
- Native passthrough (OpenAI/Azure): pass file_search tool through as-is
- RAG injection (all others): search vector stores, inject context, strip tool

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

litellm/llms/base_llm/responses/transformation.py

@@ -230,6 +230,21 @@ class BaseResponsesAPIConfig(ABC):
         """
         return False
 
+    def supports_native_file_search(self) -> bool:
+        """
+        Returns True if the provider natively handles the file_search tool with
+        vector_store_ids in the Responses API.
+
+        When True:  the file_search tool is passed through as-is to the provider,
+                    and LiteLLM maps any managed vector store IDs to provider-specific IDs.
+        When False: LiteLLM intercepts the file_search tool, performs RAG search against
+                    the referenced vector stores, injects context into the input, and
+                    strips the file_search tool before forwarding to the provider.
+
+        Default: False (RAG injection mode)
+        """
+        return False
+
     #########################################################
     ########## CANCEL RESPONSE API TRANSFORMATION ##########
     #########################################################

litellm/llms/openai/responses/transformation.py

@@ -51,6 +51,10 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig):
             )
         )
 
+    def supports_native_file_search(self) -> bool:
+        """OpenAI natively handles file_search with vector_store_ids."""
+        return True
+
     def map_openai_params(
         self,
         response_api_optional_params: ResponsesAPIOptionalRequestParams,