Description
Stochastic code and LLM-backed behavior can be evaluated with repeatable test workflows. AI developers use pydantic-evals for model regression checks, agent scoring, datasets, and quality experiments. Evaluation inputs and outputs may include sensitive prompts or generated content.