Developer docs

Integration examples

Production integration patterns for the redaction API: queue-based workers, idempotent job creation, multi-tenant API keys and incident-safe state tracking.

Integration Examples

This page collects practical integration patterns for production systems.

Queue-based architecture

Recommended for high-volume ingestion:

  1. Application uploads source PDFs to your backend.
  2. Backend enqueues a redaction task.
  3. Worker creates Redact PDF AI jobs with idempotency keys.
  4. Worker polls status and downloads outputs.
  5. Worker stores redacted outputs in your system of record.

Multi-tenant key management

  • Keep one API key per environment at minimum.
  • For enterprise isolation, map internal tenants to independent service credentials and processing queues.

Incident-safe processing

  • Persist job_id, document_id, and request_id on every state change.
  • Use idempotency keys on all job creations.
  • Add dead-letter handling for repeated 5xx failures.

Compliance-aligned processing

  • Default to retention=ephemeral.
  • Use retention=studio only when human review is required.
  • Delete jobs after export with DELETE /v1/jobs/{job_id}.

Minimal production checklist

  • server-side key storage
  • idempotency on create job
  • retry policy for 429/5xx
  • alerting on error rates and quota events
  • data retention and deletion runbooks