Marvin has quite a few multi-modal examples:

https://news.ycombinator.com/item?id=39036670

PedanticAI can do multi-modal, too:

https://dev.to/stephenc222/how-to-use-pydanticai-for-structu...