3.10.0 (2026-11-10)
Features
- decision-models: support OpenAI Decisions API in Python classification evaluators (#17514) (f6dc888)
- evals: support TypeSafe Jev decision models in Python classification evaluators (#16953) (9089efc)
- evals: use Anthropic native structured output in AnthropicAdapter (#17476) (2fffecb)
Bug Fixes
- evals: ask for the Anthropic tool call instead of forcing tool_choice (#17366) (be808f6)
- evals: drop hardcoded decision instructions (#17372) (a600c71)
- evals: keep metadata from dictionary results in create_evaluator (#17484) (52a0b8c)
- evals: raise a clear error when Google or LiteLLM tool calling returns no tool call (#17313) (e886020)