Skip to documentation content
Dataset APIDocs

Production checklist

Verify security, reproducibility, resilience, and observability before launch.

In this guide
  1. 1Secure credentials
  2. 2Pin releases
  3. 3Test failure paths
  4. 4Monitor quality
In this guide
  1. 1Secure credentials
  2. 2Pin releases
  3. 3Test failure paths
  4. 4Monitor quality

Security

  1. 1
    Use a production-only bearer key from a managed secret store
  2. 2
    Redact Authorization from logs and traces
  3. 3
    Use HTTPS and verify certificates
  4. 4
    Define credential rotation and emergency revocation owners

Data contract

  1. 1
    Validate required OpenAPI fields and enums
  2. 2
    Persist dataset_id, release_id, dataset_version, and quality_status
  3. 3
    Define whether partial releases are accepted
  4. 4
    Preserve identity_fields during reconciliation
  5. 5
    Archive required release provenance

Resilience

  1. 1
    Set connection and total deadlines
  2. 2
    Respect Retry-After on 429
  3. 3
    Use exponential backoff with jitter for safe retries
  4. 4
    Reconcile ambiguous build creation before resubmitting
  5. 5
    Stream downloads to temporary storage

Observability

  1. 1
    Track success, latency, 4xx, 429, and 5xx separately
  2. 2
    Log request_id and stable resource IDs
  3. 3
    Alert on partial releases and broken lineage
  4. 4
    Monitor build age by status
  5. 5
    Exercise sandbox and failure tests before each integration release
Was this page helpful?