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

Evaluations

REST — Evaluations endpoints.

Method
Path
Description

POST

/api/v1/evaluations

Create

GET

/api/v1/evaluations/{id}

Get

GET

/api/v1/evaluations

List

GET

/api/v1/evaluations/{id}/rows

Per-row results

POST

/api/v1/evaluations/{id}/cancel

Cancel running

Create body

{
 "model_id": "...",
 "benchmark_id": "...",
 "scorers": [...],
 "judges": [...]
}

Status

queued, running, completed, failed, cancelled.

Idempotency

POST /api/v1/evaluations accepts an Idempotency-Key header. Repeat requests with the same key within 24 hours return the original result rather than creating a duplicate evaluation.

Use a UUID per logical request (e.g., one per CI run). Keys are scoped per-org and per-endpoint.

See also

Last updated

Was this helpful?