Using pydantic validators with structured outputs and AzureAssistantAgent

[smokhasi]

With model gpt-4o-2024-08-06 and newer, and AzureAssistantAgent, is it possible to use pydantic validators with structured outputs?

For example, using a pydantic class as follows

class ResponseModel(KernelBaseModel):
    county: str = Field(default=None, description='County or city')

    @field_validator('county')
    def county_must_contain_spaces(cls, v):
        if ' ' not in v:
            raise ValueError('must contain a space')
        return v

We are generating a json_schema from the above class using Pydantic's model_json_schema() method which results in

{'type': 'json_schema', 'json_schema': {'name': 'ResponseFormat', 'schema': {'properties': {'county': {'default': None, 'description': 'County or city.', 'title': 'County', 'type': 'string'}}, 'title': 'ResponseModel', 'type': 'object'}}}

However, the above schema does not contain any validation information.

Can you provide any examples of using validators with structured outputs and AzureAssistantAgent?
If this is not possible with structured outputs, what is the recommended approach?

[moonbox3]

Hi @smokhasi, please have a look at this AzureAssistantAgent structured outputs sample.