Schema Configuration for Document Extraction

Define custom extraction schemas for any document type — field types, validation rules, and line item extraction.

Field types

Supported field types: text, number, currency, date, boolean, enum, and line_item. Each field has a name, type, description, and an extraction prompt that the model uses to find the value.

Validation rules

Add per-field validation: regex patterns, min/max bounds, required vs optional, and custom evaluators. Failed validations surface as warnings on the Review & Edit screen.

Line item extraction

For tabular data (invoice line items, PO lines), define a line_item field with a sub-schema. The model extracts the table and returns one row per line item with the sub-fields populated.