feat(api)!: BREAKING CHANGE: support passing extra_body through to providers (#3777)

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

With this, we removed the 2 vllm-specific parameters from completions
API into `extra_body`.
Before/After
<img width="1883" height="324" alt="image"
src="https://github.com/user-attachments/assets/acb27c08-c748-46c9-b1da-0de64e9908a1"
/>



closes #2720

## Test Plan
CI and added new test
```
❯ uv run pytest -s -v tests/integration/ --stack-config=server:starter --inference-mode=record -k 'not( builtin_tool or safety_with_image or code_interpreter or test_rag ) and test_openai_completion_guided_choice' --setup=vllm --suite=base --color=yes
Uninstalled 3 packages in 125ms
Installed 3 packages in 19ms
INFO     2025-10-10 14:29:54,317 tests.integration.conftest:118 tests: Applying setup 'vllm' for suite base
INFO     2025-10-10 14:29:54,331 tests.integration.conftest:47 tests: Test stack config type: server
         (stack_config=server:starter)
============================================================================================================== test session starts ==============================================================================================================
platform darwin -- Python 3.12.11, pytest-8.4.2, pluggy-1.6.0 -- /Users/erichuang/projects/llama-stack-1/.venv/bin/python
cachedir: .pytest_cache
metadata: {'Python': '3.12.11', 'Platform': 'macOS-15.6.1-arm64-arm-64bit', 'Packages': {'pytest': '8.4.2', 'pluggy': '1.6.0'}, 'Plugins': {'anyio': '4.9.0', 'html': '4.1.1', 'socket': '0.7.0', 'asyncio': '1.1.0', 'json-report': '1.5.0', 'timeout': '2.4.0', 'metadata': '3.1.1', 'cov': '6.2.1', 'nbval': '0.11.0'}}
rootdir: /Users/erichuang/projects/llama-stack-1
configfile: pyproject.toml
plugins: anyio-4.9.0, html-4.1.1, socket-0.7.0, asyncio-1.1.0, json-report-1.5.0, timeout-2.4.0, metadata-3.1.1, cov-6.2.1, nbval-0.11.0
asyncio: mode=Mode.AUTO, asyncio_default_fixture_loop_scope=None, asyncio_default_test_loop_scope=function
collected 285 items / 284 deselected / 1 selected

tests/integration/inference/test_openai_completion.py::test_openai_completion_guided_choice[txt=vllm/Qwen/Qwen3-0.6B]
instantiating llama_stack_client
Starting llama stack server with config 'starter' on port 8321...
Waiting for server at http://localhost:8321... (0.0s elapsed)
Waiting for server at http://localhost:8321... (0.5s elapsed)
Waiting for server at http://localhost:8321... (5.1s elapsed)
Waiting for server at http://localhost:8321... (5.6s elapsed)
Waiting for server at http://localhost:8321... (10.1s elapsed)
Waiting for server at http://localhost:8321... (10.6s elapsed)
Server is ready at http://localhost:8321
llama_stack_client instantiated in 11.773s
PASSEDTerminating llama stack server process...
Terminating process 98444 and its group...
Server process and children terminated gracefully


============================================================================================================= slowest 10 durations ==============================================================================================================
11.88s setup    tests/integration/inference/test_openai_completion.py::test_openai_completion_guided_choice[txt=vllm/Qwen/Qwen3-0.6B]
3.02s call     tests/integration/inference/test_openai_completion.py::test_openai_completion_guided_choice[txt=vllm/Qwen/Qwen3-0.6B]
0.01s teardown tests/integration/inference/test_openai_completion.py::test_openai_completion_guided_choice[txt=vllm/Qwen/Qwen3-0.6B]
================================================================================================ 1 passed, 284 deselected, 3 warnings in 16.21s =================================================================================================
```

tests/unit/providers/inference/test_remote_vllm.py

@@ -13,11 +13,16 @@ import pytest
 from llama_stack.apis.inference import (
     OpenAIAssistantMessageParam,
     OpenAIChatCompletion,
-    OpenAIChatCompletionRequest,
+    OpenAIChatCompletionRequestWithExtraBody,
     OpenAIChoice,
+    OpenAICompletion,
+    OpenAICompletionChoice,
+    OpenAICompletionRequestWithExtraBody,
     ToolChoice,
 )
 from llama_stack.apis.models import Model
+from llama_stack.core.routers.inference import InferenceRouter
+from llama_stack.core.routing_tables.models import ModelsRoutingTable
 from llama_stack.providers.datatypes import HealthStatus
 from llama_stack.providers.remote.inference.vllm.config import VLLMInferenceAdapterConfig
 from llama_stack.providers.remote.inference.vllm.vllm import VLLMInferenceAdapter
@@ -57,7 +62,7 @@ async def test_old_vllm_tool_choice(vllm_inference_adapter):
         mock_client_property.return_value = mock_client
 
         # No tools but auto tool choice
-        params = OpenAIChatCompletionRequest(
+        params = OpenAIChatCompletionRequestWithExtraBody(
             model="mock-model",
             messages=[{"role": "user", "content": "test"}],
             stream=False,
@@ -173,7 +178,7 @@ async def test_openai_chat_completion_is_async(vllm_inference_adapter):
         )
 
     async def do_inference():
-        params = OpenAIChatCompletionRequest(
+        params = OpenAIChatCompletionRequestWithExtraBody(
             model="mock-model",
             messages=[{"role": "user", "content": "one fish two fish"}],
             stream=False,
@@ -191,3 +196,148 @@ async def test_openai_chat_completion_is_async(vllm_inference_adapter):
 
         assert mock_create_client.call_count == 4  # no cheating
         assert total_time < (sleep_time * 2), f"Total time taken: {total_time}s exceeded expected max"
+
+
+async def test_vllm_completion_extra_body():
+    """
+    Test that vLLM-specific guided_choice and prompt_logprobs parameters are correctly forwarded
+    via extra_body to the underlying OpenAI client through the InferenceRouter.
+    """
+    # Set up the vLLM adapter
+    config = VLLMInferenceAdapterConfig(url="http://mocked.localhost:12345")
+    vllm_adapter = VLLMInferenceAdapter(config=config)
+    vllm_adapter.__provider_id__ = "vllm"
+    await vllm_adapter.initialize()
+
+    # Create a mock model store
+    mock_model_store = AsyncMock()
+    mock_model = Model(identifier="mock-model", provider_resource_id="mock-model", provider_id="vllm")
+    mock_model_store.get_model.return_value = mock_model
+    mock_model_store.has_model.return_value = True
+
+    # Create a mock dist_registry
+    mock_dist_registry = MagicMock()
+    mock_dist_registry.get = AsyncMock(return_value=mock_model)
+    mock_dist_registry.set = AsyncMock()
+
+    # Set up the routing table
+    routing_table = ModelsRoutingTable(
+        impls_by_provider_id={"vllm": vllm_adapter},
+        dist_registry=mock_dist_registry,
+        policy=[],
+    )
+    # Inject the model store into the adapter
+    vllm_adapter.model_store = routing_table
+
+    # Create the InferenceRouter
+    router = InferenceRouter(routing_table=routing_table)
+
+    # Patch the OpenAI client
+    with patch.object(VLLMInferenceAdapter, "client", new_callable=PropertyMock) as mock_client_property:
+        mock_client = MagicMock()
+        mock_client.completions.create = AsyncMock(
+            return_value=OpenAICompletion(
+                id="cmpl-abc123",
+                created=1,
+                model="mock-model",
+                choices=[
+                    OpenAICompletionChoice(
+                        text="joy",
+                        finish_reason="stop",
+                        index=0,
+                    )
+                ],
+            )
+        )
+        mock_client_property.return_value = mock_client
+
+        # Test with guided_choice and prompt_logprobs as extra fields
+        params = OpenAICompletionRequestWithExtraBody(
+            model="mock-model",
+            prompt="I am feeling happy",
+            stream=False,
+            guided_choice=["joy", "sadness"],
+            prompt_logprobs=5,
+        )
+        await router.openai_completion(params)
+
+        # Verify that the client was called with extra_body containing both parameters
+        mock_client.completions.create.assert_called_once()
+        call_kwargs = mock_client.completions.create.call_args.kwargs
+        assert "extra_body" in call_kwargs
+        assert "guided_choice" in call_kwargs["extra_body"]
+        assert call_kwargs["extra_body"]["guided_choice"] == ["joy", "sadness"]
+        assert "prompt_logprobs" in call_kwargs["extra_body"]
+        assert call_kwargs["extra_body"]["prompt_logprobs"] == 5
+
+
+async def test_vllm_chat_completion_extra_body():
+    """
+    Test that vLLM-specific parameters (e.g., chat_template_kwargs) are correctly forwarded
+    via extra_body to the underlying OpenAI client through the InferenceRouter for chat completion.
+    """
+    # Set up the vLLM adapter
+    config = VLLMInferenceAdapterConfig(url="http://mocked.localhost:12345")
+    vllm_adapter = VLLMInferenceAdapter(config=config)
+    vllm_adapter.__provider_id__ = "vllm"
+    await vllm_adapter.initialize()
+
+    # Create a mock model store
+    mock_model_store = AsyncMock()
+    mock_model = Model(identifier="mock-model", provider_resource_id="mock-model", provider_id="vllm")
+    mock_model_store.get_model.return_value = mock_model
+    mock_model_store.has_model.return_value = True
+
+    # Create a mock dist_registry
+    mock_dist_registry = MagicMock()
+    mock_dist_registry.get = AsyncMock(return_value=mock_model)
+    mock_dist_registry.set = AsyncMock()
+
+    # Set up the routing table
+    routing_table = ModelsRoutingTable(
+        impls_by_provider_id={"vllm": vllm_adapter},
+        dist_registry=mock_dist_registry,
+        policy=[],
+    )
+    # Inject the model store into the adapter
+    vllm_adapter.model_store = routing_table
+
+    # Create the InferenceRouter
+    router = InferenceRouter(routing_table=routing_table)
+
+    # Patch the OpenAI client
+    with patch.object(VLLMInferenceAdapter, "client", new_callable=PropertyMock) as mock_client_property:
+        mock_client = MagicMock()
+        mock_client.chat.completions.create = AsyncMock(
+            return_value=OpenAIChatCompletion(
+                id="chatcmpl-abc123",
+                created=1,
+                model="mock-model",
+                choices=[
+                    OpenAIChoice(
+                        message=OpenAIAssistantMessageParam(
+                            content="test response",
+                        ),
+                        finish_reason="stop",
+                        index=0,
+                    )
+                ],
+            )
+        )
+        mock_client_property.return_value = mock_client
+
+        # Test with chat_template_kwargs as extra field
+        params = OpenAIChatCompletionRequestWithExtraBody(
+            model="mock-model",
+            messages=[{"role": "user", "content": "test"}],
+            stream=False,
+            chat_template_kwargs={"thinking": True},
+        )
+        await router.openai_chat_completion(params)
+
+        # Verify that the client was called with extra_body containing chat_template_kwargs
+        mock_client.chat.completions.create.assert_called_once()
+        call_kwargs = mock_client.chat.completions.create.call_args.kwargs
+        assert "extra_body" in call_kwargs
+        assert "chat_template_kwargs" in call_kwargs["extra_body"]
+        assert call_kwargs["extra_body"]["chat_template_kwargs"] == {"thinking": True}