Instruct refusal when using Model response format

API Feedback
1

Hello,
This is a feature request. I’ve been playing with latest structured outputs feature particularly using python with pydantic. I love it.

Currently, a refusal from the model means there is unsafe input/output. In my use case, I would like the model to refuse to process the request for various reasons (e.g.: Using vison API, when there is nothing to see in the image).

For example:

completion = client.beta.chat.completions.parse(
        model="gpt-4o-2024-08-06",
        response_format=Ticket,
        messages=[
            {
                "role": "user",
                "content": [
                    {
                        "type": "text",
                        "text": "Extract the necessary information from the image. Format your response in JSON. If no information can be extracted, refuse to process the request.",
                    },
                    {
                        "type": "image_url",
                        "image_url": {"detail": "auto", "url": image.file},
                    },
                ],
            }
        ],
    )