Turn specs into evals for any agent with ASSERT
Adaptive Spec-driven Scoring for Evaluation and Regression Testing (ASSERT) is an open-source framework for converting natural language behavior requirements into executable evaluations of AI models and agents.