Skip to main content
An evaluation is the record of one judgment. The list and the detail both require judgments:read. GET /api/evaluations returns up to 50 rows and a nextCursor. Pass nextCursor back as cursor. Each summary row is the evaluation id, pack id, pack version id, source, status, model id, latency, and created time. It does not include the ticket. GET /api/evaluations/{id} returns the redacted ticket, the full result, usage, and labels. Labels include agreement for that question. See Concepts. A row is not overwritten. Judge the ticket again and you get a new id. The pack version id on the row is the version that ran, including after you publish a newer one.