feat: The response.incomplete event is not handled when using stream_mode=['messages'] (#33871)

Author: riunyfir

libs/partners/openai/langchain_openai/chat_models/base.py

@@ -4351,7 +4351,7 @@ def _advance(output_idx: int, sub_idx: int | None = None) -> None:
     elif chunk.type == "response.created":
         id = chunk.response.id
         response_metadata["id"] = chunk.response.id  # Backwards compatibility
-    elif chunk.type == "response.completed":
+    elif chunk.type in ("response.completed", "response.incomplete"):
         msg = cast(
             AIMessage,
             (

libs/partners/openai/tests/cassettes/test_incomplete_response.yaml.gz

@@ -58,6 +58,26 @@ def _check_response(response: BaseMessage | None) -> None:
     assert response.response_metadata["service_tier"]  # type: ignore[typeddict-item]
 
 
+@pytest.mark.vcr
+def test_incomplete_response() -> None:
+    model = ChatOpenAI(
+        model=MODEL_NAME, use_responses_api=True, max_completion_tokens=16
+    )
+    response = model.invoke("Tell me a 100 word story about a bear.")
+    assert response.response_metadata["incomplete_details"]
+    assert response.response_metadata["incomplete_details"]["reason"]
+    assert response.response_metadata["status"] == "incomplete"
+
+    full: AIMessageChunk | None = None
+    for chunk in model.stream("Tell me a 100 word story about a bear."):
+        assert isinstance(chunk, AIMessageChunk)
+        full = chunk if full is None else full + chunk
+    assert isinstance(full, AIMessageChunk)
+    assert full.response_metadata["incomplete_details"]
+    assert full.response_metadata["incomplete_details"]["reason"]
+    assert full.response_metadata["status"] == "incomplete"
+
+
 @pytest.mark.default_cassette("test_web_search.yaml.gz")
 @pytest.mark.vcr
 @pytest.mark.parametrize("output_version", ["responses/v1", "v1"])