Incomplete responses are not handled by `parse_response` when using Structured Output

[ejm-betterup]

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

I'm using Structured Outputs and have a max token length of 2000 for my finetuned model. When I generate a response, response.output has a length of 8 and the last output is incomplete. When response.output contains an incomplete output, parse_response doesn't handle this and instead raises a ValidationError.

It seems like there should either be a way to constrain the length of response.output or ignore incomplete outputs in parsing, for example:

if output.type == "message" and output.status != "incomplete":

instead of https://github.com/openai/openai-python/blob/main/src/openai/lib/_parsing/_responses.py#L63

To Reproduce

To repro, ensure that response.output contains an incomplete output and then call parse_response with a structured output.

raw_response.output[-1] = ResponseOutputMessage(id='xxx', content=[ResponseOutputText(annotations=[], text='{"a": {"b": 5.0, , "c": {"d": 4.0,', type='output_text', logprobs=[])], role='assistant', status='incomplete', type='message')

parse_response(
   text_format=ExpectedResponse,
   response=raw_response,
)

Code snippets

client = OpenAI() 
resp = client.responses.parse(
    model=model,
    instructions=prompt,
    input=input_text.to_list(),
    temperature=temperature,
    max_output_tokens=max_tokens
    text_format=ExpectedResponse,
)


yields


  File "/.venv/lib/python3.13/site-packages/openai/_base_client.py", line 1124, in _process_response
    return api_response.parse()
           ~~~~~~~~~~~~~~~~~~^^
  File "/venv/lib/python3.13/site-packages/openai/_response.py", line 325, in parse
    parsed = self._options.post_parser(parsed)
  File "/.venv/lib/python3.13/site-packages/openai/resources/responses/responses.py", line 936, in parser
    return parse_response(
        input_tools=tools,
        text_format=text_format,
        response=raw_response,
    )
  File "/.venv/lib/python3.13/site-packages/openai/lib/_parsing/_responses.py", line 75, in parse_response
    "parsed": parse_text(item.text, text_format=text_format),
              ~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/.venv/lib/python3.13/site-packages/openai/lib/_parsing/_responses.py", line 137, in parse_text
    return cast(TextFormatT, model_parse_json(text_format, text))
                             ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^
  File "/.venv/lib/python3.13/site-packages/openai/_compat.py", line 169, in model_parse_json
    return model.model_validate_json(data)
           ~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^
  File "/.venv/lib/python3.13/site-packages/pydantic/main.py", line 746, in model_validate_json
    return cls.__pydantic_validator__.validate_json(
           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^
        json_data, strict=strict, context=context, by_alias=by_alias, by_name=by_name
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    )
    ^
pydantic_core._pydantic_core.ValidationError: 1 validation error for ExpectedResponse
  Invalid JSON: EOF while parsing an object at line 1 column 234 [type=json_invalid, input_value='{"example_key":...n": {"clarity": 6', input_type=str]
    For further information visit https://errors.pydantic.dev/2.11/v/json_invalid

OS

macOS

Python version

python v3.13

Library version

openai v1.86.0