Administration

Run a company-locked full-agent Tally sandbox replay

Runs a human-approved redacted benchmark only inside a selected disposable demo company. The source firm and company are forbidden, firm-level tools are denied, raw response and tool payloads are not persisted, and the advisory grader cannot override deterministic or human gates. Requires admin:tally-quality:write.

post/admin/tally/flight-recorder/sandbox-replays

Request

  • Base URL: https://ledgerhq.pro/api/v1
  • URL: https://ledgerhq.pro/api/v1/admin/tally/flight-recorder/sandbox-replays
  • Auth: one of:
    • HTTP bearer and API key in header x-organization-id
    • HTTP bearer and API key in header x-organization-id

Headers

Idempotency-Keystring required

Stable retry key for this posting request within the LedgerHQ organization.

Request body

benchmarkIdstring uuid required
targetCompanyOrganizationIdstring uuid required
provider'anthropic' | 'openai_codex' | 'openrouter' required
modelstring required
graderProvider'anthropic' | 'openai_codex' | 'openrouter' required
graderModelstring required
strategy'single_agent' | 'specialist_guided' required
baselineReplayIdstring uuid nullable
confirmedtrue required

Response

Idempotent existing replay

Changes

No recorded changes to this operation across all 5 revisions of this API.