FunctionCallOutput is inconsistent with OpenAI API docs

[zhukoff-pavel]

Confirm this is an issue with the Python library and not an underlying OpenAI API

• This is an issue with the Python library

Describe the bug

Hi!

The model for https://github.com/openai/openai-python/blob/main/src/openai/types/responses/response_input_item.py#L99 seems to be inconsistent with what is listed in https://platform.openai.com/docs/api-reference/responses/list.

The documentation lists output value for FunctionCallOutput as a string or list of objects (see https://platform.openai.com/docs/api-reference/responses/list#responses/list-data-function-tool-call-output-output).

However, Pydantic model in Python library types that field as string only: https://github.com/openai/openai-python/blob/main/src/openai/types/responses/response_input_item.py#L103.

To Reproduce

N/A

Code snippets

OS

Any

Python version

Python v3.x

Library version

openai v1.109.1

[zhukoff-pavel]

The problem might be deeper and connected with OpenAI OpenAPI specification. I've created an issue there as well: openai/openai-openapi#488

[michael-dommett-deel]

most likely #2670 is linked as well to the openapi spec