Supported file types
All document formats you can upload to the platform.
Fully supported
| Format | Extensions | Notes |
|---|---|---|
| Native (digital) PDFs give best results. Scanned pages are transcribed by a vision model — see below. | ||
| Microsoft Word | .doc, .docx | Both old (.doc) and new (.docx) formats. |
| Microsoft Excel | .xls, .xlsx, .xlsm, .xlsb | Good for pricing tables, technical matrices, evaluation grids. |
| Microsoft PowerPoint | .ppt, .pptx | For presentation-format proposals or briefings. |
| Other documents | .rtf, .odt, .epub, .txt, .md, .html, .xml | Rich Text, OpenDocument, EPUB, plain text, Markdown, HTML, and XML are all read as text. |
| Spreadsheets & data | .csv, .tsv, .ods | Comma/tab-separated and OpenDocument spreadsheets. |
| EDOC archives | .edoc, .zip (ASiC-E) | Latvia-specific digitally signed archive (ASiC-E container). All contained files are extracted and parsed automatically — and nested EDOCs inside are unpacked recursively, each inner file parsed on its own. |
| 7-Zip archives | .7z | Extracted automatically; every file inside is parsed (up to 100 files per archive). |
Best practices by format
PDFs
Native PDFs (exported from Word, Excel, or other software) parse with near-perfect accuracy. The text is already digital.
Scanned or image-only PDFs contain no digital text — they're pictures of pages. During analysis, each such page is sent to a vision model that transcribes what it can see, so a scanned annex still reaches the review on its actual content. Two things to keep in mind:
- A blurry, faded, or badly skewed scan may yield little or no readable text. It is recorded that way rather than guessed at.
- If you have the original digital document, upload that instead. It parses faster and more accurately than any scan of it.
Excel files
Excel files are parsed sheet by sheet. The system handles:
- Multiple sheets
- Tables and data ranges
- Headers and formulas (values are extracted, not formulas)
- Basic formatting
Complex pivot tables, charts, and conditional formatting may not translate fully. If a requirement is hidden in a complex Excel formula, consider extracting that data to a simpler format.
EDOC and 7z archives
EDOC is a digitally signed document archive (ASiC-E container) used in Latvia. Upload the .edoc file directly — the system extracts all documents inside and parses each one individually. ASiC-E containers shared as .zip are accepted too. If an EDOC contains another EDOC inside it, the nested archive is unpacked recursively, and every inner file is parsed and classified on its own. You'll see each extracted file listed separately with its own parsing status.
.7z archives work the same way: upload the .7z file and the system extracts and parses every file inside (up to 100 files per archive).
File size limits
- Maximum per file: 50 MB
- No limit on number of files per procurement or bid
If a file exceeds 50 MB, try:
- Splitting multi-part documents into separate files
- Removing large embedded images that aren't relevant
- Compressing the PDF (many PDF tools can reduce file size)
Images, audio, and video
Image files (.jpg, .jpeg, .png, .gif, .bmp, .tif, .tiff, .webp) are read. During analysis a vision model transcribes each one and adds the text to the parsed document, so a photographed contract page, a stamped certificate, or a diagram with figures in it is analyzed on what it actually shows rather than on its filename.
Two limits apply: very small images — logos, bullets, layout artefacts — are skipped deliberately, and there is a ceiling on how many images and scanned pages a single run reads, so an unusually picture-heavy document set may not be read in full.
SVG files and audio/video files upload fine and are kept with your documents, but they are not transcribed. Use those as attachments or evidence rather than as a source the AI reads.
Genuinely unsupported
- AutoCAD (.dwg, .dxf) — convert drawings to PDF first
- Compressed archives other than .edoc, ASiC-E .zip, and .7z (for example, .rar)