David
David
Hey @DouweM , I went back to basics with this one. Can't say that I simplified the parts manager but I hope I at least made it easier to understand....
hey @DouweM , I finally found out why the final flush wasn't covered in `__init__.py` the current coverage failure is on [messages.py](https://github.com/pydantic/pydantic-ai/blob/359c6d2d7b58f2b4921d927346bbb4d724f73c42/pydantic_ai_slim/pydantic_ai/messages.py#L1035-L1037) (I'm currently confused why that line has a...
keeping openai as an example, the problem is that that timestamp on the `UserPromptPart` is set to the current local utc (which includes microseconds), but the openai response's `created` field...
excellent, I'll do that then. quick comment on #3422 , should we call that field explicitly "developer_metadata"? could surely reduce much need for clarification by both humans and their llms....
there's a flaky outlines test breaking CI :(
there's a missing for loop in the method inplementstion, will push a fix tomorrow!
Hey @lars20070 , I appreciate the change. Nonetheless, if in the end the host will be the one running ollama I expect the container only needs to reach the endpoint,...
I just pushed an update to this removing the AST aspect and (hopefully) fixing the tests so they pass in CI haven't addressed the comments yet so it isn't reviewable...
been having some fun testing the combinations, for anyone who wants to try them: combinations to try out the agent UI ### 1. Generic agent with explicit model source .env...
working on removing mcp support and the remaining 4 comments