pydantic-ai
https://github.com/pydantic/pydantic-ai
Python
GenAI Agent Framework, the Pydantic way
Triage Issues!
When you volunteer to triage issues, you'll receive an email each day with a link to an open issue that needs help in this project. You'll also receive instructions on how to triage issues.
Triage Docs!
Receive a documented method or class from your favorite GitHub repos in your inbox every day. If you're really pro, receive undocumented methods or classes and supercharge your commit history.
Python not yet supported6 Subscribers
Add a CodeTriage badge to pydantic-ai
Help out
- Issues
- Prompt-cache prefix stability as a default-on, monitored framework guarantee
- NativeOutput silently prevents function-tool calls with vLLM structured decoding
- Emit OTel `gen_ai.usage.reasoning.output_tokens` and `gen_ai.request.reasoning.level`
- pydantic-evals: round out the built-in evaluator pack (safety metrics, similarity heuristics, named-preset policy)
- Verify UI event-path cache-prefix fidelity with golden fixtures generated by the real Vercel AI / AG-UI client reducers
- Inter-agent communication / async subagent result injection lacks a dedicated message type
- Add OTel observability for silent message-history mutations (repair, sanitize, prepare_messages)
- Apply tool timeouts in `ToolManager` so all toolsets honor them
- Optional provenance for human-in-the-loop tool approvals over UI adapters (verify an approval maps to a real server-issued pause)
- Cancellation semantics: a level-triggered contract for agent runs, streams, tools, and durable execution
- Docs
- Python not yet supported