Document extraction

Invoices, delivery notes, purchase orders and receipts. Everything runs locally.

…

Drop an invoice or delivery note here

PDF or image, up to 25 MB. Scanned documents are fine.

Read, then checked

The parser extracts text and tables, a model reads the fields, and then the arithmetic is recomputed independently. A document that does not add up is flagged, not corrected.

Every value traceable

Each field records what the parser saw, on which page, and how firmly the value was found there. A value that appears nowhere in the document is reported as such.

Nothing invented

Absent information stays absent. A missing total is null, never a calculated guess.

Recent documents

Held in memory for this demo, so restarting the API clears the list.

Loading…