Overview
Test sets
Section titled “Test sets”Test set catalog (list, get, versions, publish, unregister) — part of the admin control plane. Full versioned CRUD, mirrors policies 1:1. Registry-only: test-run execution + per-kind grader dispatch live on the test-runs tag. evalKind selects the shape of spec — closed enum, extended additively (accuracy, pairwise, regression, human-review, benchmark, custom, corrections). The cases of a corrections test set are copies of corrected runs, built before an agent’s corrections were its test set (/v1/agents/{agentId}/test-set); they’re read at GET .../versions/{version}/cases (FrozenCaseStoreBinding), and GET /v1/test-sets doesn’t list them. Runs against these test sets are started through the test-runs tag. Caller-plugged via TestSetRegistryBinding.