featu: support passing "extra body" throught to providers

# What does this PR do?
Allows passing through extra_body parameters to inference providers.


closes #2720

## Test Plan
CI and added new test

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

@@ -12,8 +12,8 @@ from botocore.client import BaseClient
 from llama_stack.apis.inference import (
     ChatCompletionRequest,
     Inference,
-    OpenAIChatCompletionRequest,
-    OpenAICompletionRequest,
+    OpenAIChatCompletionRequestWithExtraBody,
+    OpenAICompletionRequestWithExtraBody,
     OpenAIEmbeddingsResponse,
 )
 from llama_stack.apis.inference.inference import (
@@ -134,12 +134,12 @@ class BedrockInferenceAdapter(
 
     async def openai_completion(
         self,
-        params: OpenAICompletionRequest,
+        params: OpenAICompletionRequestWithExtraBody,
     ) -> OpenAICompletion:
         raise NotImplementedError("OpenAI completion not supported by the Bedrock provider")
 
     async def openai_chat_completion(
         self,
-        params: OpenAIChatCompletionRequest,
+        params: OpenAIChatCompletionRequestWithExtraBody,
     ) -> OpenAIChatCompletion | AsyncIterator[OpenAIChatCompletionChunk]:
         raise NotImplementedError("OpenAI chat completion not supported by the Bedrock provider")

llama_stack/providers/remote/inference/databricks/databricks.py

@@ -8,7 +8,7 @@ from collections.abc import Iterable
 
 from databricks.sdk import WorkspaceClient
 
-from llama_stack.apis.inference import OpenAICompletion, OpenAICompletionRequest
+from llama_stack.apis.inference import OpenAICompletion, OpenAICompletionRequestWithExtraBody
 from llama_stack.log import get_logger
 from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
 
@@ -39,6 +39,6 @@ class DatabricksInferenceAdapter(OpenAIMixin):
 
     async def openai_completion(
         self,
-        params: OpenAICompletionRequest,
+        params: OpenAICompletionRequestWithExtraBody,
     ) -> OpenAICompletion:
         raise NotImplementedError()

llama_stack/providers/remote/inference/llama_openai_compat/llama.py

@@ -3,7 +3,12 @@
 #
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
-from llama_stack.apis.inference.inference import OpenAICompletion, OpenAICompletionRequest, OpenAIEmbeddingsResponse
+
+from llama_stack.apis.inference.inference import (
+    OpenAICompletion,
+    OpenAICompletionRequestWithExtraBody,
+    OpenAIEmbeddingsResponse,
+)
 from llama_stack.log import get_logger
 from llama_stack.providers.remote.inference.llama_openai_compat.config import LlamaCompatConfig
 from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
@@ -29,7 +34,7 @@ class LlamaCompatInferenceAdapter(OpenAIMixin):
 
     async def openai_completion(
         self,
-        params: OpenAICompletionRequest,
+        params: OpenAICompletionRequestWithExtraBody,
     ) -> OpenAICompletion:
         raise NotImplementedError()
 

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

@@ -13,9 +13,9 @@ from llama_stack.apis.inference import (
     Inference,
     OpenAIChatCompletion,
     OpenAIChatCompletionChunk,
-    OpenAIChatCompletionRequest,
+    OpenAIChatCompletionRequestWithExtraBody,
     OpenAICompletion,
-    OpenAICompletionRequest,
+    OpenAICompletionRequestWithExtraBody,
     OpenAIEmbeddingsResponse,
 )
 from llama_stack.apis.models import Model
@@ -79,7 +79,7 @@ class PassthroughInferenceAdapter(Inference):
 
     async def openai_completion(
         self,
-        params: OpenAICompletionRequest,
+        params: OpenAICompletionRequestWithExtraBody,
     ) -> OpenAICompletion:
         client = self._get_client()
         model_obj = await self.model_store.get_model(params.model)
@@ -93,7 +93,7 @@ class PassthroughInferenceAdapter(Inference):
 
     async def openai_chat_completion(
         self,
-        params: OpenAIChatCompletionRequest,
+        params: OpenAIChatCompletionRequestWithExtraBody,
     ) -> OpenAIChatCompletion | AsyncIterator[OpenAIChatCompletionChunk]:
         client = self._get_client()
         model_obj = await self.model_store.get_model(params.model)

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

@@ -9,7 +9,7 @@ from collections.abc import AsyncIterator
 from llama_stack.apis.inference import (
     OpenAIChatCompletion,
     OpenAIChatCompletionChunk,
-    OpenAIChatCompletionRequest,
+    OpenAIChatCompletionRequestWithExtraBody,
 )
 from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
 
@@ -31,7 +31,7 @@ class RunpodInferenceAdapter(OpenAIMixin):
 
     async def openai_chat_completion(
         self,
-        params: OpenAIChatCompletionRequest,
+        params: OpenAIChatCompletionRequestWithExtraBody,
     ) -> OpenAIChatCompletion | AsyncIterator[OpenAIChatCompletionChunk]:
         """Override to add RunPod-specific stream_options requirement."""
         params = params.model_copy()

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

@@ -14,7 +14,7 @@ from pydantic import ConfigDict
 
 from llama_stack.apis.inference import (
     OpenAIChatCompletion,
-    OpenAIChatCompletionRequest,
+    OpenAIChatCompletionRequestWithExtraBody,
     ToolChoice,
 )
 from llama_stack.log import get_logger
@@ -93,7 +93,7 @@ class VLLMInferenceAdapter(OpenAIMixin):
 
     async def openai_chat_completion(
         self,
-        params: OpenAIChatCompletionRequest,
+        params: OpenAIChatCompletionRequestWithExtraBody,
     ) -> OpenAIChatCompletion | AsyncIterator[OpenAIChatCompletionChunk]:
         params = params.model_copy()