Evaluate changes
Run example conversations and compare their execution and answer quality.
Overview
Test real application conversations, inspect their answers, compare saved results, and configure meaningful evaluation checks in Studio.
Run your first evaluation
Define one conversation, expose your application endpoint, connect it to Cloud or self-hosted Studio, and inspect the first saved result.
Configure evaluation judging
Choose an application or Studio judge, define evidence-based checks, and distinguish answer quality failures from judge execution errors.
Read and compare evaluation results
Inspect execution outcomes and quality checks, compare saved evaluations, and understand workflow, judge, and partial model costs.