Skip to main content

Define AI transformer pipelines

Use MailSlurp's AI Transformers to turn emails, attachments, and SMS into structured data.

  • Define the data you want to extract using prompts or JSON schemas.
  • Connect inboxes and phone numbers
  • View results in tables or connect webhooks, databases, or external apps directly

ai pipeline flow

Transformer design workflow​

  1. Define target fields in output schemas.
  2. Write a transformer prompt that reflects source message structure.
  3. Connect input sources (inboxes, attachments, SMS feeds).
  4. Run controlled test samples and inspect extraction output.
  5. Promote to production integrations once quality is stable.

Prompt and schema tips​

  • Use explicit field descriptions (for example, expected format and units).
  • Keep required fields minimal at first to reduce false negatives.
  • Add examples for ambiguous values (invoice numbers, order IDs, phone formats).
  • Prefer stable vocabulary for enum-like fields.

Operating guidance​

  • Maintain a versioned change log for transformer updates.
  • Re-test with representative historical messages after each prompt/schema change.
  • Link each transformer to an owner and downstream consumer for faster incident triage.

Continue with​