Exclude model field from result

[iloveitaly]

What I would like to do is have a pydantic model where the AI model call should fill out all the fields but one or two.

The idea here is I'm looping through a list of elements that need to be processed by a particular agent. I have an ID or some sort of deterministic value that I want set on each result and I don't want the agent to touch that field. I could create two separate pydantic models to represent these two states, but I'd rather have a single pydantic model and somehow mark one of the fields as handled by me instead of handled by the agent.

Is this sort of thing possible?

Here's a quick example.

class LabelResult(BaseModel):
    labels: list[str]
    reasoning: str
    summary: str
    message_id: str # I want the model to avoid filling this field in

[HamzaFarhan]

Yes. Like this:

from pydantic import BaseModel
from pydantic.json_schema import SkipJsonSchema


class LabelResult(BaseModel):
    labels: list[str]
    reasoning: str
    summary: str
    message_id: SkipJsonSchema[str]


LabelResult.model_json_schema()

{'properties': {'labels': {'items': {'type': 'string'},
                           'title': 'Labels',
                           'type': 'array'},
                'reasoning': {'title': 'Reasoning', 'type': 'string'},
                'summary': {'title': 'Summary', 'type': 'string'}},
 'required': ['labels', 'reasoning', 'summary'],
 'title': 'LabelResult',
 'type': 'object'}

I have successfully done this for skipping id Fields.

[mikeedjones]

Think that would mean this field will be missing if the schema is reproduced for another reason - like the swagger docs of a fastapi deployment which uses this model.