docuflow-ai accepts more than 75 file extensions. Each accepted extension is a content type with an OCR engine that decides how its text is read. Platform administrators can switch content types on or off under Admin → System administration → Content Types; only active ones can be uploaded.
How files are read#
| Engine | Used for | How it works |
|---|---|---|
| Auto | Digital text is read directly; scanned pages are recognized with OCR. PDFs also go through AI splitting. | |
| Paddle | Images | Deep-learning OCR for photos and scans. |
| Kreuzberg | Office, OpenDocument, e-mail, e-books, markup, data and citation files | The document structure is parsed, so no OCR is needed and the text is exact. |
| Do not OCR | Plain text | The file content is used as-is. |
Accepted extensions#
| Category | Extensions | Engine |
|---|---|---|
.pdf | Auto | |
| Word | .doc .docx .docm .dot .dotx .dotm .odt .rtf | Kreuzberg |
| Spreadsheets | .xls .xlsx .xlsm .xlsb .xla .xlam .xlt .xltx .ods .csv .tsv | Kreuzberg |
| Presentations | .ppt .pptx .pptm .pps .ppsx .pot .potx .potm | Kreuzberg |
| Images | .jpg .jpeg .png .gif .tif .tiff .webp .bmp .jp2 .jpx .jpm .mj2 .jbig2 .jb2 .pnm .pbm .pgm .ppm | Paddle |
.eml .msg | Kreuzberg | |
| Web & markup | .html .htm .xml .svg .md .markdown .mdx .rst .org .djot .opml .dbk .docbook | Kreuzberg |
| Publishing | .epub .fb2 .tex .latex .typ .typst | Kreuzberg |
| Data & configuration | .yaml .yml .toml .ipynb | Kreuzberg |
| Citations | .bib .ris .enw .nbib .csl | Kreuzberg |
| Plain text | .txt | Do not OCR |
Developers can fetch the active list with
GET /api/content-types and pass the matching ContentTypeId
when uploading. The list is shared by every organization and is maintained by platform administrators under
Admin → System administration → Content Types.
In-app preview#
The View → Original tab renders a PDF preview. Non-PDF files are converted on the fly:
| Format | Preview |
|---|---|
| Shown as is | |
.docx .doc .rtf | Converted to PDF |
.xlsx .xls .xlsb .csv | Converted to PDF |
.png .jpg .jpeg .gif .bmp .tif .tiff .webp | Converted to PDF |
.html .htm .md .markdown .txt .log | Rendered and converted to PDF |
.pptx .ppt .pptm .ppsx .pps | Converted with LibreOffice when it is installed on the server |
| Everything else | No preview; the Text and Json tabs and Download still work |
Limits#
- Browser uploads: up to 200 MB per file.
- A file whose text cannot be read (for example a corrupted or password-protected file) is retried and then marked Permanently failed.
- Only PDFs are split into separate business documents.