Downsides to passing dataclasses.dataclass to Agent.output_type?

[patrickkidd]

Question

I need a common data model defined in a way that can be shared between two apps: a web app that uses pydantic_ai, and a desktop app that cannot include heavy non-pure python dependencies like pydantic/pydantic_ai in its exe. All I want is to ensure that pydantic_ai enforces the output JSON to be valid against my schema, which includes an object hierarchy including lists and some enums (enums can be worked around if necessary). What are the downsides to just passing a dataclasses.dataclass to Agent? Is it going to be a fragile approach?

Additional Context

python 11.6, latest pydantic_ai

[DouweM]

@patrickkidd No downsides, that's fully supported as mentioned under https://ai.pydantic.dev/output/#structured-output.

I'll close the issue assuming that answers the question, but feel free to ask followup questions.

[patrickkidd]

It sure does! ThanksOn Oct 23, 2025, at 4:05 PM, Douwe Maan ***@***.***> wrote:DouweM left a comment (pydantic/pydantic-ai#3234) @patrickkidd No downsides, that's fully supported as mentioned under https://ai.pydantic.dev/output/#structured-output. I'll close the issue assuming that answers the question, but feel free to ask followup questions. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: ***@***.***>