Chapter 1 of 7
Drop it in the bucket. The rest is magic.
MagicBucket.pro turns business documents into a ready table. You drop a file; you fish out the original, a CSV row, a checked summary and a receipt for every value.
What you drop
- Invoices and receipts, including structured XML e-invoices
- Payroll lists and payslips
- Contracts
- Financial and non-financial statements
As a PDF, a scan, a phone photo, or a Word, Excel or XML file. No renaming, sorting or tagging: the bucket recognizes the file type by its content, not its name, and works out the document type by itself.
What you fish out
| File | What's in it |
|---|---|
| original | your file, exactly as you dropped it |
.csv | one row with the fields you chose, ready to open in your spreadsheet |
.md | a short summary in which every sentence was checked against the document |
.log.json | the receipt: where each value came from, who voted for it, which checks passed |
.zip | all of the above, together |
You always get the full set, even when something goes wrong. The status then says so plainly:
- done
- done, 2 disputed
- unreadable
- no template
- page limit
Your bucket, your fields
We don't train models on your documents. Your organization's "model" is simply an approved list of fields for each document type: a template. Your administrator builds it in the browser from a handful of samples.
- Samples. Drop 3–10 examples of one document type: different issuers, scans and PDFs, better and worse quality.
- Proposal. The strongest model studies them and proposes the fields: column name, type (text, number, amount, date, yes/no, list), where to find it on the page, and the check it must pass, such as an IBAN check digit or net + VAT = gross.
- Edit. Switch fields off, add your own, change names, types and order.
- "Can it be done?" The full reading pipeline runs on every sample. A field gets a "yes" when it was confirmed in at least 90% of the samples. A field that appears in no sample gets a "no": agreeing that something is missing is not a success.
- Accept. Every acceptance creates a new template version, and every document remembers the version it was read with.
Any country, any paperwork
MagicBucket.pro isn't tied to one country's forms. You show it your documents and it learns their fields, so an invoice from Lisbon, a payslip from Toronto and a balance sheet from Singapore go into the same bucket. The models read documents in all major languages, and you name your fields in the language your team works in. Amounts, dates and numbers are normalized whatever the local notation, and identifiers that carry check digits get checked.