Testing

Testing & Validation

2Signal provides multiple layers of testing for AI agents — from unit-testing individual evaluators to running full regression suites against datasets. Combine these approaches to catch issues at every stage of development and deployment.

Testing Approaches

ApproachWhat It TestsWhen to UseTools
Unit TestingIndividual evaluators in isolationDuring evaluator developmentVitest, Pytest
Dataset EvaluationAgent behavior across many inputsBefore deploys, regression testingDashboard, CLI, CI/CD
Production MonitoringLive agent performanceAlways-onEvaluators, Alerts
Trace ReplayModel/prompt changes against real dataA/B testing, optimizationDashboard

Have questions? Join our community!

Connect with other developers and the 2Signal team.

Join Discord