Samuel Colvin

you can pass this data as JSON, is it really worth changing the input type and serializing to JSON withint the library?

I'm not sure I really understand either either case. Taking Case 1 as an example, internally pydantic-ai would still be serializing structured as JSON and sending it to the model...

thanks so much.

LGTM, please can you add an screenshot or similar example to PR showing how this improves the display?

thanks so much.

We discussed this at length on the [OT Python SOG call](https://docs.google.com/document/d/1CIMGoIOZ-c3-igzbd6_Pnxx1SjAkjwqoYSUWxPY8XIs/edit?tab=t.0) - I think it's pretty clear that in python at least, this change represents a profound regression in functionality...

> If you use it for many dialects in a single application (as I do in my use case) then your suggestion will lead to an instantiation of the entire...

Creating issues like this with no description or suggestions is not particularly friendly. Especially when you're trying to build with Python 3.14!

Also, perhaps related, the docs [refer](https://www.python-httpx.org/advanced/transports/#custom-transports) to `httpx.Mounts`, but that doesn't seem to exist.

I'd love to see this package upgrade to pydantic v2. If we can help at all, let us know. I don't think 40% of downloads means 40% of AI/ML use...