Skip to main content

Extractions API

The Extractions API provides access to structured data produced by the extraction pipeline. List results, retrieve row-level data with provenance, and submit corrections.

Endpoints

How extractions work

An extraction is created when you call the extract endpoint or re-extract a document. The four-phase pipeline processes the document and produces structured rows. Each cell includes a confidence score and per-cell provenance linking it to the source region. The Field Registry resolves field names to canonical entries.