[responsesAPI] support input output messages for non harmony models (#29549)

Signed-off-by: Andrew Xia <axia@fb.com>
Co-authored-by: Andrew Xia <axia@fb.com>

tests/entrypoints/openai/test_response_api_simple.py

@@ -42,6 +42,24 @@ async def test_basic(client: OpenAI, model_name: str):
     assert response.status == "completed"
 
 
+@pytest.mark.asyncio
+@pytest.mark.parametrize("model_name", [MODEL_NAME])
+async def test_enable_response_messages(client: OpenAI, model_name: str):
+    response = await client.responses.create(
+        model=model_name,
+        input="Hello?",
+        extra_body={"enable_response_messages": True},
+    )
+    assert response.status == "completed"
+    assert response.input_messages[0]["type"] == "raw_message_tokens"
+    assert type(response.input_messages[0]["message"]) is str
+    assert len(response.input_messages[0]["message"]) > 10
+    assert type(response.input_messages[0]["tokens"][0]) is int
+    assert type(response.output_messages[0]["message"]) is str
+    assert len(response.output_messages[0]["message"]) > 10
+    assert type(response.output_messages[0]["tokens"][0]) is int
+
+
 @pytest.mark.asyncio
 @pytest.mark.parametrize("model_name", [MODEL_NAME])
 async def test_reasoning_item(client: OpenAI, model_name: str):