Chat Completions: fails with function name for tool_choice parameter w/ streaming enabled #980
Description
Please read this first
- Have you read the docs? Yes
- Have you searched for related issues? Yes
Describe the bug
A clear and concise description of what the bug is.
This only happens when running with Runner.run_streamed
When a function name is provided to ModelSettings's tool_choice, the converter here will convert it to a dict structure:
@classmethod
def convert_tool_choice(
cls, tool_choice: Literal["auto", "required", "none"] | str | None
) -> ChatCompletionToolChoiceOptionParam | NotGiven:
if tool_choice is None:
return NOT_GIVEN
elif tool_choice == "auto":
return "auto"
elif tool_choice == "required":
return "required"
elif tool_choice == "none":
return "none"
else:
return {
"type": "function",
"function": {
"name": tool_choice,
},
}
However, at the below line, we restrict the input to be one of the three Literals.
Resulting in the below pydantic validation error:
validated_self = self.__pydantic_validator__.validate_python(data, self_instance=self)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
pydantic_core._pydantic_core.ValidationError: 5 validation errors for Response
tool_choice.literal['none','auto','required']
Input should be 'none', 'auto' or 'required' [type=literal_error, input_value={'type': 'function', 'fun...ognize_named_entities'}}, input_type=dict]
For further information visit https://errors.pydantic.dev/2.11/v/literal_error
tool_choice.ToolChoiceTypes.type
Input should be 'file_search', 'web_search_preview', 'computer_use_preview', 'web_search_preview_2025_03_11', 'image_generation' or 'code_interpreter' [type=literal_error, input_value='function', input_type=str]
For further information visit https://errors.pydantic.dev/2.11/v/literal_error
tool_choice.ToolChoiceFunction.name
Field required [type=missing, input_value={'type': 'function', 'fun...ognize_named_entities'}}, input_type=dict]
For further information visit https://errors.pydantic.dev/2.11/v/missing
tool_choice.ToolChoiceMcp.server_label
Field required [type=missing, input_value={'type': 'function', 'fun...ognize_named_entities'}}, input_type=dict]
For further information visit https://errors.pydantic.dev/2.11/v/missing
tool_choice.ToolChoiceMcp.type
Input should be 'mcp' [type=literal_error, input_value='function', input_type=str]
For further information visit https://errors.pydantic.dev/2.11/v/literal_error
Debug information
- Agents SDK version:
v0.1.0 - Python version: Python 3.12
Repro steps
Run with Runner.run_streamed and set the model settings to a tool function name instead of the default three Literals.
Expected behavior
Without any validation issues.