> For the complete documentation index, see [llms.txt](https://help.steeple.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.steeple.com/en/assistant-ia/assistant-ia-documents.md).

# Attach a document to the assistant

The AI assistant can make use of the content of a document or image that you send it. It is useful for summarizing text, extracting information, translating, or asking questions about a specific file.

## Attach a file

* Click the icon **paperclip** next to the input field
* Select the file on your device
* The file appears above the input field once ready
* Send the message with the file attached

To remove an attached file before sending, click the cross on the thumbnail.

## Accepted formats

### Documents

* PDF
* Word (`.doc`, `.docx`)
* Excel (`.xls`, `.xlsx`)
* PowerPoint (`.ppt`, `.pptx`)
* OpenDocument (`.odt`, `.ods`, `.odp`)
* CSV, TXT, RTF
* HTML, Markdown
* EPUB

### Images

* JPEG, PNG, GIF, WebP and any format recognized by the browser

## Limits

* **One document and one image** maximum per message
* **Image size** : 10 MB maximum
* **Document size** : the content is analyzed then sent to the assistant. Very large documents may be truncated beyond a certain threshold. For a file that is too long, prefer a targeted excerpt.

## What the assistant does with your file

The document content is **extracted and sent to the assistant to process your current message**. It is not permanently stored on the assistant side and is not reused for subsequent conversations.

Images are also analyzed in the context of the sent message.

## Use cases

* Summarize a report or internal note
* Extract the key points from a report
* Translate a document
* Ask questions about a received file
* Ask the assistant to draft a post from a brief

## Limitations

* The assistant cannot modify or save the attached file
* Password-protected files cannot be analyzed
* Very large files or files with complex layouts (nested tables, non-OCR scans) may be only partially understood