Yenisei CLIFeatures
Working with files and documents
How the assistant reads code and documents, which of it costs credits and what the size limits are.
The assistant reads files in the working folder on its own: just name the file in the task, for example "Read contract.pdf and list all the deadlines". It reads text and code on your computer, and no credits are charged for that; only recognizing scans can cost money.
Text files and code
A file up to 60 KB is read in full. A larger file is read in 60 KB chunks, and only the chunks needed for the task. If the assistant found a place by searching, it reads the section around the matching line.
Documents
The assistant reads PDF, Word (.docx), Excel (.xlsx) and PowerPoint
(.pptx) files.
- Text. It is taken from the file itself, right on your computer; this is free.
- Tables and slides. Tables are read row by row, with numbers, dates and sheet names. Slides are read in presentation order.
- Old formats.
.doc,.xlsand.pptfiles are not read: the program suggests saving them again as.docx,.xlsx,.pptxor PDF.
Scans
A PDF without a text layer, that is, one saved as images, is recognized by the Yenisei Docs model. This is paid, and the program warns about it with a separate line: "scan without text, recognizing with the yenisei-doc model (paid, up to 3 min)…". Recognition can be stopped with Ctrl+C.
Images
The assistant does not pass images to the model.
Limits
| Limit | Value |
|---|---|
| Document size | Up to 25 MB. |
| Scan for recognition | Up to 10 MB. |
| Recognition time | Up to 3 minutes. |
| Document text per read | About 60 KB, that is, roughly 30 thousand characters of Russian text; a longer document is read in parts. |