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
- Ollama's context-overflow error isn't classified as `ModelContextWindowExceededError`
- Hybrid-thinking models: make the thinking on/off switch a per-gateway setting, and keep family profiles to "hybrid" or "always on"
- Gemma 4: unified `thinking` is dropped on most routes, although Gemma 4 takes `reasoning_effort`
- `CohereModel`: support `NativeOutput` through Cohere's `response_format` JSON Schema mode
- `XaiModel`: map `service_tier` to xAI priority processing instead of rejecting it as unsupported
- `GroqModel`: map `service_tier` to Groq's service tiers instead of stripping it
- `OpenAICompatibleProvider` / `LiteLLMProvider`: vendor-prefixed profile lookup is case-sensitive and misses Hugging Face org names
- `VercelProvider` profile map uses prefixes Vercel doesn't use, so most non-OpenAI models get no family profile
- `LiteLLMProvider` resolves profiles with OpenRouter-style prefixes, so LiteLLM's own `gemini/`, `xai/` and `vertex_ai/` ids get the OpenAI profile
- `AlibabaProvider`: structured-output flags don't match DashScope's documented support
- Docs
- Python not yet supported