Using Responses API with Structured Output and Pydantic - Optional Values not working

API
1

Hey all,

I’m currently struggeling with above combination.

I do have the following Pydatic Base Model, where founder is optional:

class Capital(BaseModel):
    capital: str
    country: str
    population: float
    longitude: float
    latitude: float
    founder: Optional[str]

class Result(BaseModel):
    capitals: list[Capital]

When I send this via the responses.parse endpoint the generated request seems valid:

"founder": {
                    "title": "Founder",
                    "anyOf": [
                        {"type": "string"},
                        {"type": "null"}
                    ]
                }

and founder is a mandatory field.

Here is my request:

response = client.responses.parse(
    model="gpt-4.1-mini",
    input=[
        {"role": "system", "content": "Provide the relevant information based on the user's question. For each capital, if the founder is unknown or if the founder is not a single person, set the 'founder' field to '-'."},
        {
            "role": "user",
            "content": "Hello, which are the 5 biggest capitals in Europe?!"
        }
    ],
    text_format=Result
)

As long as all returned rows do contain a founder everything works like charm, but once at least one founder is null / not returned by the LLM the response parsing (within the OpenAI SDK fails).

  File "C:\Users\<homefolder>\source\sample\.venv\Lib\site-packages\openai\lib\_parsing\_responses.py", line 62, in parse_response      
    for output in response.output:
TypeError: 'NoneType' object is not iterable

Am I doing something wrong? Is this a limitation or can I overcome it somehow.
I came across some OpenAI documentation using Pydantic and Optional as well. But I also read you should not trust all the OpenAI documentaiton.

Thanks in advance