evaluate

Overview ¶

Package main demonstrates standalone eval execution with sdk.Evaluate().

This example shows:

• Running evals from a PromptPack against a conversation snapshot
• No live provider or agent connection needed — just messages in, results out
• Filtering evals by trigger (every_turn vs on_session_complete)
• Running evals from inline definitions (no pack file needed)
• Receiving eval results via the EventBus for reactive workflows
• Recording eval results as Prometheus metrics via MetricsCollector
• Validating that all eval types are registered before execution
• Extending the eval registry with custom exec handlers via RuntimeConfig

Run with:

go run .