Check the schema on a real sample.
Review fields and run a sample test before finalizing. Changes to fields, instructions, or the test sample require another test.
Schema Extraction
Your samples. Your structure.
Start with your documents, not a blank schema. AI proposes the structure; you review it and test it. Then a vision-language model (VLM) reads every page as an image — tables, stamps, handwriting and layout included — and fills your schema across the batch.
AI identifies a proposed structure from sample invoices. Proposed fields: Supplier, text; Invoice date, date; Total, number; Line items, table. Review and test this structure before use.
A few samples. A head start on every field.
AI reads your samples and proposes a schema; the VLM then extracts against it page by page. Explore the steps below, then switch between the prepared fields, JSON, and table views. You review and test the proposal before using it for the batch.
Sample data — not a live extraction. This demo does not upload or process documents.
Extractions · Restaurant bills
Schema ready · 6 fields
Name the workflow
Choose the recurring document task.
Restaurant bills
Use a name you will recognize when the next batch arrives.
JSON export
Excel
Restaurant namestring
Copper Kettle Kitchen
bill-0417.pdf · p.1confirmed
Bill numberstring
CK/2026/0417
bill-0417.pdf · p.1confirmed
Bill datedate
18-08-2026
bill-0417.pdf · p.1confirmed
Line itemsarray · 6
6 rows captured
bill-0417.pdf · p.1confirmed
GST numberstring
29ABCDE1234F1Z5
bill-0417.pdf · p.2needs review
Final amountnumber
2,148.00
bill-0417.pdf · p.2confirmed
Every field keeps the page it was read from, so a value can be checked against the source before it leaves the workspace.
Built for the details your work needs.
Upload the batch, not one file at a time.
Select multiple documents or drag them into a finalized workflow. Extraction starts automatically for the uploaded files, with progress and failures visible per document.
Simple values, grouped details, repeating tables.
Define typed fields for text, numbers, dates, choices, and yes/no values. Group related details and capture repeating table rows such as invoice line items.
Keep, correct, and check the source.
Inspect extracted values beside the document. Review flagged fields, keep correct values, and fix missing or uncertain results before using them.
From documents to downloads.
JSON for structured data.
Keep field names, typed values, grouped objects, and repeating arrays in a machine-readable format. Export ready workflow results or selected ready files.
Excel for spreadsheet work.
Download one workbook: a summary sheet with a row per document, plus a sheet for each repeating table such as line items. Amounts are real numbers, so columns sum.
Sample data — not a live extraction. Download an example invoice record.
{
"document": "invoice-1042.pdf",
"invoice_number": "INV-2026-1042",
"supplier": "HussleBook Demo Studio",
"total": 6000.0,
"currency": "USD"
}Bring the batch. Take the results.
Upload your documents, review the extracted fields, then export results as JSON or Excel.
Check feature availability