Documents Analyzer

The Documents Analyzer reads the documents you upload to a project, splits them into pages and analyses them with the AI settings of the project. You open it from the Documents Analyzer tile in the AI Agent group.

This page lists the file formats you can upload, how each format is read and what limits apply.

Supported file formats

Format	File extensions	How the pages are made
PDF	.pdf	One page for each PDF page, read with OCR
Images	.png, .jpg, .jpeg	One page for each image, read with OCR
Text and data files	.txt, .md, .csv, .json, .xml	The text is split into pages
Web pages	.html, .htm	The text is split into pages
Word documents	.docx	The text is split into pages
Excel workbooks	.xlsx, .xls	The text is split into pages
PowerPoint presentations	.pptx, .ppt	The text is split into pages
Rich Text documents	.rtf	The text is split into pages
E-books	.epub	The text is split into pages
Emails	.eml, .msg	The text is split into pages

PDF files and images are read with OCR, using the OCR AI Configuration of the project.

The other formats are read as text, without OCR. Their text is split into pages of up to about 2,500 characters, about one A4 page of text. A table that does not fit on a page continues on the next page, with its header repeated. A CSV file becomes a table, and JSON and XML files are shown indented.

These pages are not the pages that Word, PowerPoint or another program shows. They are the pages that the Documents Analyzer counts and analyses: in the Pages column of the documents, when you page through a document, in the Block Size (pages) of the project, and in the page ranges of the analysis, such as Pages 1 - 5.

Pictures in Word, PowerPoint, Excel, Rich Text and e-book files are not read, so text that is only in a picture, such as a scanned page in a Word document, is missing. If you need that text, or pages that match the original document, such as one page for each slide, save the document as PDF and upload the PDF. Its pages are then read with OCR.

Limits

• A file can be up to 200 MB. A larger file is rejected at upload, with a message that names the file, its size and the limit.
• A file in a format that is not in the list above is rejected at upload, with a message that lists the allowed formats.
• Text and data files can be in UTF-8, in UTF-16 with a byte order mark, or in Windows-1251. In other encodings, letters such as ä or é can come out wrong: save the file in UTF-8 and upload it again.

Excel workbooks

• All sheets of the workbook are read, one after another. Each sheet starts with its name as a heading.
• A sheet does not start a new page: the pages are cut by length, as for other text. One page can hold several short sheets, and a long sheet can take several pages.
• A cell with a formula shows the value that was saved with the file. If the formulas of a file were never calculated and saved in Excel, as in some files created by other programs or downloaded from the internet, these cells show NaN. Open the file in Excel, save it, and upload it again.

Emails

• The text of the email is read, followed by each attachment under its file name.
• Attachments must be in a supported format. If an email has an attachment that cannot be read, such as a ZIP archive, the email gets the status Error, and its Error Text names the attachment. Upload the email without that attachment, and upload the files you need from it separately.

File formats that are not supported

• Word 97–2003 documents (.doc): these older Word files cannot be read reliably; tables, text encodings and embedded objects often come out wrong. Save the file as a Word document (.docx) and upload that.
• Macro-enabled Office files, such as .xlsm and .xltm: save the workbook as an Excel workbook (.xlsx) and upload that.
• Archives, such as .zip: unpack the archive and upload the files it contains.