Skip to main content
These guides are for hosted Unfold workflows. They use the same primitives as the rest of the docs: store a document, run executions, read normalized results, delete when done. They are application-owned patterns. Unfold does not auto-pick engines or fail over for you today.

Fast preview, then full parse

One document, one job, two engines. Show LiteParse as soon as it finishes.

Cheap path, then escalate

Run LiteParse first. Start a heavier engine only if the cheap result looks weak.

Bake off engines

Fan out with compare to score quality, latency, and cost on your files.

Idempotency and cleanup

Safe retries, retention windows, and documents.delete.

Primitives cheat sheet

See How Unfold works for the contract, and Documents and jobs for the HTTP shape.