generated: '2026-07-20'
method: derived
source: openapi/*, https://studio.precog.cloud/docs/
authentication:
  style: HTTP Bearer token (opaque access token), applied globally.
  mcp: OAuth 2.0 (PKCE S256) on the MCP surface — see authentication/precog-authentication.yml
  cross_ref: authentication/precog-authentication.yml
idempotency:
  supported: false
  note: >-
    No Idempotency-Key header or idempotency contract is declared in the
    published specs. Mutating operations are not documented as idempotent.
pagination:
  style: none-detected
  note: >-
    Collection endpoints (GET /sources, /destinations, /pipelines, dataset
    lists) return full arrays with no cursor/offset/limit parameters or Link
    headers in the published specs.
versioning:
  scheme: spec-version
  current: '1.0'
  note: >-
    Specs carry info.version 1.0; no version prefix appears in the URL paths and
    no dated/header version scheme is documented.
  cross_ref: lifecycle/precog-lifecycle.yml
error_envelope:
  media_type: application/json
  shape: '{ "error": { "type": string, "details": string } }'
  rfc9457: false
  cross_ref: errors/precog-problem-types.yml
rate_limit_signaling:
  api_level: none-detected
  note: >-
    No 429 responses or X-RateLimit/Retry-After headers in the API. A
    source-side `ratelimitReset` (epoch millis) appears only inside
    dataset-telemetry, describing upstream connector throttling, not API limits.
async_jobs:
  pattern: >-
    Long-running work (semantic-model generation, pipeline loads) uses a
    job/resource model — POST returns 202 and a job id; poll GET
    /semantic-model/generation-jobs/{job-id} for status.