test

# What does this PR do?


## Test Plan

llama_stack/providers/inline/inference/meta_reference/inference.py

@@ -173,5 +173,6 @@ class MetaReferenceInferenceImpl(
         top_logprobs: int | None = None,
         top_p: float | None = None,
         user: str | None = None,
+        **kwargs: Any,
     ) -> OpenAIChatCompletion | AsyncIterator[OpenAIChatCompletionChunk]:
         raise NotImplementedError("OpenAI chat completion not supported by meta-reference inference provider")

llama_stack/providers/inline/inference/sentence_transformers/sentence_transformers.py

@@ -124,5 +124,6 @@ class SentenceTransformersInferenceImpl(
         top_logprobs: int | None = None,
         top_p: float | None = None,
         user: str | None = None,
+        **kwargs: Any,
     ) -> OpenAIChatCompletion | AsyncIterator[OpenAIChatCompletionChunk]:
         raise NotImplementedError("OpenAI chat completion not supported by sentence transformers provider")

llama_stack/providers/remote/inference/bedrock/bedrock.py

@@ -186,5 +186,6 @@ class BedrockInferenceAdapter(
         top_logprobs: int | None = None,
         top_p: float | None = None,
         user: str | None = None,
+        **kwargs: Any,
     ) -> OpenAIChatCompletion | AsyncIterator[OpenAIChatCompletionChunk]:
         raise NotImplementedError("OpenAI chat completion not supported by the Bedrock provider")

llama_stack/providers/remote/inference/passthrough/passthrough.py

@@ -153,6 +153,7 @@ class PassthroughInferenceAdapter(Inference):
         top_logprobs: int | None = None,
         top_p: float | None = None,
         user: str | None = None,
+        **kwargs: Any,
     ) -> OpenAIChatCompletion | AsyncIterator[OpenAIChatCompletionChunk]:
         client = self._get_client()
         model_obj = await self.model_store.get_model(model)
@@ -181,6 +182,7 @@ class PassthroughInferenceAdapter(Inference):
             top_logprobs=top_logprobs,
             top_p=top_p,
             user=user,
+            **kwargs,
         )
 
         return await client.inference.openai_chat_completion(**params)

llama_stack/providers/remote/inference/runpod/runpod.py

@@ -57,6 +57,7 @@ class RunpodInferenceAdapter(OpenAIMixin):
         top_logprobs: int | None = None,
         top_p: float | None = None,
         user: str | None = None,
+        **kwargs: Any,
     ):
         """Override to add RunPod-specific stream_options requirement."""
         if stream and not stream_options:
@@ -86,4 +87,5 @@ class RunpodInferenceAdapter(OpenAIMixin):
             top_logprobs=top_logprobs,
             top_p=top_p,
             user=user,
+            **kwargs,
         )

llama_stack/providers/remote/inference/vllm/vllm.py

@@ -102,6 +102,7 @@ class VLLMInferenceAdapter(OpenAIMixin):
         top_logprobs: int | None = None,
         top_p: float | None = None,
         user: str | None = None,
+        **kwargs: Any,
     ) -> OpenAIChatCompletion | AsyncIterator[OpenAIChatCompletionChunk]:
         max_tokens = max_tokens or self.config.max_tokens
 
@@ -136,4 +137,5 @@ class VLLMInferenceAdapter(OpenAIMixin):
             top_logprobs=top_logprobs,
             top_p=top_p,
             user=user,
+            **kwargs,
         )

llama_stack/providers/utils/inference/litellm_openai_mixin.py

@@ -299,6 +299,7 @@ class LiteLLMOpenAIMixin(
         top_logprobs: int | None = None,
         top_p: float | None = None,
         user: str | None = None,
+        **kwargs: Any,
     ) -> OpenAIChatCompletion | AsyncIterator[OpenAIChatCompletionChunk]:
         # Add usage tracking for streaming when telemetry is active
         from llama_stack.providers.utils.telemetry.tracing import get_current_span
@@ -335,6 +336,7 @@ class LiteLLMOpenAIMixin(
             user=user,
             api_key=self.get_api_key(),
             api_base=self.api_base,
+            **kwargs,
         )
         return await litellm.acompletion(**params)
 

llama_stack/providers/utils/inference/openai_mixin.py

@@ -313,6 +313,7 @@ class OpenAIMixin(NeedsRequestProviderData, ABC, BaseModel):
         top_logprobs: int | None = None,
         top_p: float | None = None,
         user: str | None = None,
+        **kwargs: Any,
     ) -> OpenAIChatCompletion | AsyncIterator[OpenAIChatCompletionChunk]:
         """
         Direct OpenAI chat completion API call.
@@ -361,7 +362,10 @@ class OpenAIMixin(NeedsRequestProviderData, ABC, BaseModel):
             user=user,
         )
 
-        resp = await self.client.chat.completions.create(**params)
+        # Pass any additional provider-specific parameters as extra_body
+        extra_body = kwargs if kwargs else {}
+
+        resp = await self.client.chat.completions.create(**params, extra_body=extra_body)
 
         return await self._maybe_overwrite_id(resp, stream)  # type: ignore[no-any-return]