Run a frozen comparison

/api/v1/owner/evaluationsPOST

Runs the default frozen guardian matrix without domain operations. Replay is deterministic. The HTTP endpoint accepts only executionMode; use the evaluation CLI for case/provider/condition filters and budgets. The default live matrix exceeds the 100-call ceiling and is rejected before model calls.

Examples are synthetic and illustrate response shape; IDs, signatures, and credentials are not usable authority.

Authorization

ownerSession *cookie halfspan_session

Authenticated HttpOnly owner session. Not an executor credential.

Request Bodyapplication/jsonrequired

executionMode *"replay" | "live"

Response

200 · Successful response. A paused run is an application outcome, not an HTTP error.

id *string

version *string

suite *string

executionMode *"replay" | "live"

createdAt *string

caseCount *number

casesDigest *string

promptDigest *string

schemaDigest *string

repeat *number

maxCalls *number

callsUsed *number

captureErrors *object[]

captureSamples *object[]

samples *object[]

summaries *object[]

totalTimeMs *number

operationExecution *string

note *string

taskSuccess *null

recoverySuccess *null

unauthorizedExecutions *null

metricsNote *string

400 · Request validation failed.

error *string

reasonCode?string

401 · Sign in to the Halfspan workspace.

error *string

reasonCode?string

403 · Console requests must come from the local owner workspace.

error *string

reasonCode?string

500 · The local service could not complete this request.

error *string

reasonCode?string

Request example

curl -X POST "https://service.example.com/api/v1/owner/evaluations" \
  -H "x-halfspan-owner: 1" \
  -b "halfspan_session=<token>" \
  -H "Content-Type: application/json" \
  -d '{
  "executionMode": "replay"
}'

Response example

{
  "id": "11111111-1111-4111-8111-111111111111",
  "version": "cross-frozen-evaluation/2.0.0",
  "suite": "guardian",
  "executionMode": "replay",
  "createdAt": "2026-10-26T12:00:00.000Z",
  "caseCount": 17,
  "casesDigest": "illustrative-case-digest",
  "promptDigest": "illustrative-prompt-digest",
  "schemaDigest": "illustrative-schema-digest",
  "repeat": 1,
  "maxCalls": 100,
  "callsUsed": 0,
  "captureErrors": [],
  "captureSamples": [],
  "samples": [],
  "summaries": [],
  "totalTimeMs": 0,
  "operationExecution": "not performed; semantic evaluation only",
  "note": "Illustrative report; empty sample arrays are abbreviated. Use the CLI report for measured results.",
  "taskSuccess": null,
  "recoverySuccess": null,
  "unauthorizedExecutions": null,
  "metricsNote": "Frozen judgments do not measure task completion."
}