Supported Input Formats
Input is divided into two groups: files from which data is read by the model, and files that already contain structured data.
Read by the Model
| Format | Note |
|---|---|
| Digital and scanned, up to 20 pages | |
| JPG, PNG | Photos from a mobile phone, output from a scanner |
| GIF, WebP, BMP | Automatically re-encoded |
| Word, Excel, PowerPoint | Text from the document as well as embedded photos |
E-mail .eml | Envelope: a document is created from each attachment, up to ten from one message |
Recognition and the model are applied to these formats. The result is an estimate, albeit a good one, and therefore a check is performed. TIFF is not supported; convert it to PDF or PNG.
Structured
| Format | Note |
|---|---|
| ISDOC, ISDOCX | Czech electronic invoicing standard, version 5.x and 6.x |
| UBL 2.1 | Peppol BIS, XRechnung |
| CII D16B | ZUGFeRD, Factur-X, including PDF with embedded XML |
| Pohoda XML | Batch from Pohoda, up to 50 documents in one file |
Here, nothing is debated; the data is loaded directly and the document costs no credit. In the listing, it has an electronic invoice icon. Checks run the same way, because even a structured document can have mismatched totals or invalid DIČ; the finding indicates that the error is with the issuer.
In the case of ISDOCX, an archive is opened, the invoice is read, and the attached PDF is saved as the original, so the preview shows the invoice and not the XML. A Pohoda batch with multiple documents will create as many documents as there are in it; in the listing, they are distinguished by the order in the name.
What is Not Supported
ZIP archives (except ISDOCX), TIFF, HTML pages.