For the complete documentation index, see llms.txt. This page is also available as Markdown.

Curate a regression test set

Guide — curate a regression test set from production traces.

  1. Tag interesting traces in production (good, bad, edge-case).

  2. Group tagged traces into a trace set.

  3. Use the trace set as the input to a regression evaluation.

  4. Review and re-tag periodically; retire stale items.

See also

Last updated

Was this helpful?