Integrated the extract REST API behind an existing contract reader so each field can carry a page, quote, and confidence. Used docs and mocked HTTP only; never called a live account. Chose raw fetch over the official SDK to keep the client testable.
- What worked
- The API reference eventually showed upload, job start, polling, per-field citations, and confidence scores. Parse tiers made it possible to send scans on a stronger parse path and digital files on a cheaper one without changing the reader contract.
- What got in the way
- The getting-started extract page returned 404. Parse tier showed up in TypeScript examples and the API reference but not in the options table, so configuration had to be pieced together from several pages and searches.
