internal

Compare Regression Row

One baseline risk beside what the new path produced for it.

Both sides are resolved live from document_risks rather than read from the snapshot on the regression row. The snapshot is what the sweep saw; a reader opening this panel wants the risk as it stands, including any review that has happened since. The stored verdict rides along unchanged so the two can be compared rather than silently reconciled.

get/internal/evals/rollout-consistency/sweeps/{sweep_id}/regressions/{regression_id}

Request

  • Base URL
    https://api.kobaltlabs.com (the document declares no server; this is the origin it was published from)
  • URL
    https://api.kobaltlabs.com/internal/evals/rollout-consistency/sweeps/{sweep_id}/regressions/{regression_id}
  • Auth
    none declared

Path parameters

sweep_idstring uuid required
regression_idstring uuid required

Response

Successful Response

regression_idstring uuid required
outcome'flagged_carried' | 'flagged_lost' | 'dismissed_dropped' | 'dismissed_reintroduced' | 'unreviewed' | 'suppressed' required

What became of one baseline risk under the new path.

The reviewer's verdict on the baseline risk decides which pair applies, and the two pairs are different questions. Losing a flagged risk is lost recall — a reviewer said it was real. Reproducing a dismissed one is a false positive the old path's validation pass caught and dedupe_findings does not, which is the number the rollout is expected to move.

baseline_verdictstring nullable required
best_match_scorenumber nullable required
near_missboolean required
is_net_newboolean required
after_requirementsstring[] required

Changes

No recorded changes to this operation across all 8 revisions of this API.