v3

latestOpenAPI 3.1.02026-07-311,4541,5202.3 MB
clickhouse

Workflows Eval Runs List

Run history for one evaluator pipeline — one row per graded log.

Newest-first list of the logs this pipeline graded, each with every grader's score (grader_scores) plus the original log's input/output/model/status. Backs the evaluator metrics tab's lower table (replaces the per-version rollup): each row is an actual pipeline run, not a version. Keyed by the workflow family id; the family→versions fan-out matches the run-metrics views. With scope=automation the family id is an automation's and both the output rows and the grader sub-scores match on the stamped caller automation_id — the automation's run history across every evaluator pipeline it runs.

get/clickhouse/workflows/{workflow_id}/eval-runs

Path parameters

workflow_idstring required

Query parameters

limitinteger

Max runs to return (default 100, capped at 500).

offsetinteger

Number of runs to skip (pagination).

scope'automation'

Set to 'automation' when the URL family id is an automation's — scores are then matched by the caller automation_id stamped on each row instead of the evaluator pipeline's version ids.

Headers

Authorizationstring required

JWT access token or Respan API key

Response

countinteger required