Skip to main content
A schema is an output contract. You name the fields you care about; Abacus Docs fills those fields when it processes a document or a batch. The sidebar label is Schemas (/structs/). The list page title is Output Schemas.

​
What a schema contains

Each field has:
  • A name and a stable field id (used in the grid, CSV, and API).
  • A type — text, number, date, boolean, enum, or a table with columns.
  • An optional description that tells the extractor what the value looks like.
  • Optional location hints (page, zone, nearby labels) so the extractor looks in the right place.
A completed extraction shows each value with a confidence and a source page. You can edit a cell later; the edit is stored as an override and appears in the grid and exports.
Transcription (the full text of the document) is separate from schema fields. Every processed document gets a transcription. Schema fields appear only after you apply a published schema.

​
Lifecycle

Publishing creates an immutable version. Later edits stay on the draft until you publish again. See Publish, versions, and the test set.

​
Two processing modes

Every schema is either Document (variable layouts such as contracts) or Form (fixed layouts such as IDs and licenses). The mode chooses the extraction engine and the field-count cap. See Form vs Document mode.

​
Where schemas are used

  • My Documents — apply a schema on the document viewer, then reprocess.
  • Batch Processing — pick a default schema when you create a job; every file in the batch uses that version (you can reassign a document later).
  • Black Hole (selected organizations) — a participating schema receives auto-filed documents into its default batch.

Open the builder

Create a schema, upload a sample, and add fields in the three-column builder.