How it works
1. We read the text layer — in memory
Your PDF is checked (file signature and 10 MB limit) and parsed in memory. The original file is not written to disk or stored.
2. Lines are rebuilt from positions
Each text fragment carries coordinates. We group fragments by vertical position, order them left to right, and rebuild the rows a human sees on the page.
3. Fields and line items are extracted
Label-aware matching finds vendor, invoice number, dates and currency. Labelled amounts give subtotal, tax and total. Rows ending in an amount become line items, parsed into description, quantity, unit price and line total where the layout allows.
4. Totals validate each other
We compare the sum of line items with the subtotal, and subtotal plus tax with the total. You see pass / warning / fail plus the exact difference, and a confidence level per field and row.
5. You review, edit and export
Everything is editable. Export produces a real .xlsx (or .csv) in Standard, QuickBooks Bill or Xero Bill columns.