aiconfig-ai-metrics

To instrument AI Config calls for metrics tracking, wrap your provider calls using the trackMetricsOf function or implement a manual Tier-4 path. This automatically captures duration, token usage, and success or error data for observability.

The four-tier instrumentation ladder for AI Config metrics provides escalating abstraction levels: managed runner, provider package, custom extractor with trackMetricsOf, and raw manual tracking. It ensures monitoring data is captured for any provider type in Python and Node environments.

Yes, you can track AI Config duration and tokens in both Python and Node environments. The instrumentation ladder applies across multiple provider types in both languages, allowing you to wrap calls and flush metrics to ensure monitoring visibility.

To ensure metrics are visible when config.enabled is checked, respect the config.enabled checks during your wrapped calls and explicitly flush metrics after execution. This guarantees that monitoring data is successfully captured and sent to the Monitoring tab.

The best way to capture success and error metrics for AI Config pipelines is using the trackMetricsOf wrapper or the manual Tier-4 pattern. This approach validates deployments, troubleshoots failures, and improves tooling with minimal toil.