Skip to main content

​
Why this demo exists

PDF products often break at delivery time, not extraction time: variant sprawl, unclear cache keys, and brittle invalidation. This pattern uses OkraPDF URL grammar and deterministic profile tokens so every page/image artifact has a stable path.

​
What it proves

  1. Define named image/page profiles (thumb, card, hero, print).
  2. Generate canonical transform tokens from profile config.
  3. Serve immutable URLs through edge/CDN with predictable cache behavior.
  4. Promote hot variants to materialized paths when traffic justifies it.

​
Cloudinary technique transfer for PDFs

You keep the same ergonomics teams love in image pipelines:
  • URL-driven transforms
  • deterministic outputs
  • strong CDN cacheability
  • explicit profile contracts across frontend + backend
But the outputs are PDF-native assets and structured document artifacts.

​
Demo Source

runtime-demo/demos/04-cloudinary-technique-transfer.ts
runtime-demo/demos/05-pipeline-sweet-spot.ts
runtime-demo/apps/vercel-pdf-pipeline-starter

​
Delivery architecture

URL Builder

Build document/page/entity URLs with zero API calls.

Upload → Extract → Publish

Reference deploy flow for production document delivery.