Evals
Evaluate agents and teams for accuracy, model-judged quality, performance, reliability, and reusable suites.
| Example | Description |
|---|---|
| Accuracy | Accuracy examples evaluate how well responses match expected outputs. |
| Agent As Judge | Agent-as-judge examples evaluate output quality with model-based scoring. |
| Performance | Performance examples benchmark runtime and memory impact for agents and teams. |
| Reliability | Reliability examples validate whether expected expected tool executions are recorded. |
| Suite | Declare reusable eval cases and run them together with the built-in suite CLI. |