Skip to content

The agent engineering platform.

openai: ignore function_calls if tool_calls are present (#31198)

ccurme868cfc4a8f57
diff --git a/libs/partners/openai/langchain_openai/chat_models/base.py b/libs/partners/openai/langchain_openai/chat_models/base.py
index cb1b68903b..5ef15b8a60 100644
--- a/libs/partners/openai/langchain_openai/chat_models/base.py
+++ b/libs/partners/openai/langchain_openai/chat_models/base.py
@@ -251,8 +251,6 @@ def _convert_message_to_dict(message: BaseMessage) -> dict:
message_dict["role"] = "user"
elif isinstance(message, AIMessage):
message_dict["role"] = "assistant"
- if "function_call" in message.additional_kwargs:
- message_dict["function_call"] = message.additional_kwargs["function_call"]
if message.tool_calls or message.invalid_tool_calls:
message_dict["tool_calls"] = [
_lc_tool_call_to_openai_tool_call(tc) for tc in message.tool_calls
@@ -267,6 +265,10 @@ def _convert_message_to_dict(message: BaseMessage) -> dict:
{k: v for k, v in tool_call.items() if k in tool_call_supported_props}
for tool_call in message_dict["tool_calls"]
]
+ elif "function_call" in message.additional_kwargs:
+ # OpenAI raises 400 if both function_call and tool_calls are present in the
+ # same message.
+ message_dict["function_call"] = message.additional_kwargs["function_call"]
else:
pass
# If tool calls present, content null value should be None not empty string.
diff --git a/libs/partners/openai/tests/unit_tests/chat_models/test_base.py b/libs/partners/openai/tests/unit_tests/chat_models/test_base.py
index c762b3dab5..2331d465ef 100644
--- a/libs/partners/openai/tests/unit_tests/chat_models/test_base.py
+++ b/libs/partners/openai/tests/unit_tests/chat_models/test_base.py
@@ -614,6 +614,50 @@ def test_openai_invoke_name(mock_client: MagicMock) -> None:
assert res.name == "Erick"
+def test_function_calls_with_tool_calls(mock_client: MagicMock) -> None:
+ # Test that we ignore function calls if tool_calls are present
+ llm = ChatOpenAI(model="gpt-4.1-mini")
+ tool_call_message = AIMessage(
+ content="",
+ additional_kwargs={
+ "function_call": {
+ "name": "get_weather",
+ "arguments": '{"location": "Boston"}',
+ }
+ },
+ tool_calls=[
+ {
+ "name": "get_weather",
+ "args": {"location": "Boston"},
+ "id": "abc123",
+ "type": "tool_call",
+ }
+ ],
+ )
+ messages = [
+ HumanMessage("What's the weather in Boston?"),
+ tool_call_message,
+ ToolMessage(content="It's sunny.", name="get_weather", tool_call_id="abc123"),
+ ]
+ with patch.object(llm, "client", mock_client):
+ _ = llm.invoke(messages)
+ _, call_kwargs = mock_client.create.call_args
+ call_messages = call_kwargs["messages"]
+ tool_call_message_payload = call_messages[1]
+ assert "tool_calls" in tool_call_message_payload
+ assert "function_call" not in tool_call_message_payload
+
+ # Test we don't ignore function calls if tool_calls are not present
+ cast(AIMessage, messages[1]).tool_calls = []
+ with patch.object(llm, "client", mock_client):
+ _ = llm.invoke(messages)
+ _, call_kwargs = mock_client.create.call_args
+ call_messages = call_kwargs["messages"]
+ tool_call_message_payload = call_messages[1]
+ assert "function_call" in tool_call_message_payload
+ assert "tool_calls" not in tool_call_message_payload
+
+
def test_custom_token_counting() -> None:
def token_encoder(text: str) -> list[int]:
return [1, 2, 3]