Add reasoning effort parameter to OpenAI Spec ChatCompletionRequest

src/litserve/specs/openai.py

@@ -176,6 +176,7 @@ class ChatCompletionRequest(BaseModel):
     tools: Optional[List[Tool]] = None
     tool_choice: Optional[ToolChoice] = ToolChoice.auto
     response_format: Optional[ResponseFormat] = None
+    reasoning_effort: Optional[Literal["low", "medium", "high"]] = None
     metadata: Optional[Dict[str, str]] = None
 
 

src/litserve/test_examples/openai_spec_example.py

@@ -53,6 +53,14 @@ def encode_response(self, output):
         )
 
 
+class TestAPIWithReasoningEffort(TestAPI):
+    def encode_response(self, output, context):
+        yield ChatMessage(
+            role="assistant",
+            content=f"This is a generated output with reasoning effort: {context.get('reasoning_effort', None)}",
+        )
+
+
 class OpenAIBatchContext(LitAPI):
     def setup(self, device: str) -> None:
         self.model = None

tests/test_specs.py

@@ -24,6 +24,7 @@
     OpenAIWithUsageEncodeResponse,
     TestAPI,
     TestAPIWithCustomEncode,
+    TestAPIWithReasoningEffort,
     TestAPIWithStructuredOutput,
     TestAPIWithToolCalls,
 )
@@ -252,6 +253,26 @@ async def test_openai_spec_metadata_required_fail(openai_request_data):
             assert "Missing required metadata" in resp.text
 
 
+@pytest.mark.asyncio
+@pytest.mark.parametrize("reasoning_effort", ["low", "medium", "high", None, "random"])
+async def test_openai_spec_reasoning_effort(reasoning_effort, openai_request_data):
+    server = ls.LitServer(TestAPIWithReasoningEffort(), spec=OpenAISpec())
+    openai_request_data["reasoning_effort"] = reasoning_effort
+    with wrap_litserve_start(server) as server:
+        async with LifespanManager(server.app) as manager, AsyncClient(
+            transport=ASGITransport(app=manager.app), base_url="http://test"
+        ) as ac:
+            resp = await ac.post("/v1/chat/completions", json=openai_request_data)
+            if reasoning_effort == "random":
+                assert resp.status_code == 422  # as random is not a valid reasoning effort
+            else:
+                assert resp.status_code == 200
+                assert (
+                    resp.json()["choices"][0]["message"]["content"]
+                    == f"This is a generated output with reasoning effort: {reasoning_effort}"
+                )
+
+
 class IncorrectAPI1(ls.LitAPI):
     def setup(self, device):
         self.model = None