Evals
List Eval Runs
List runs, newest first, paged like every other list here: a year of CI runs is the volume this feature is for.
GET
List Eval Runs
Headers
Query Parameters
Where a run is, and how it ended.
errored is deliberately not a kind of failed: it means the harness did
not complete (a connect failure, a judge outage), which is neither a pass
nor a fail and never counts toward a rate. A judge outage reading as an
agent regression is how a suite loses its audience.
Available options:
queued, running, passed, failed, errored, cancelled 1-based page number
Required range:
x >= 1Required range:
1 <= x <= 100Response
Successful Response
The response is of type Response List Eval Runs V1 Eval Runs Get · object.