Body
Table of Contents
CometAI allows you to upload files directly into a chat and ask questions about their contents. You can upload various document types, have various Language Models analyze or summarize them, and revisit them later without needing to re-upload. Firstly, Open the CometAI platform in your browser and log in using your UTD NetID and password.
Start in a New or Existing Chat
- From the main chat window, locate the + icon at the bottom left of the message bar.
- Click the + icon. Your file selection window will open.
Select and Upload Your Document
- In the file selection window, choose the file you want to upload.
- Click Open to attach the document.
View the Uploaded Document in Chat
- Once uploaded, your document will appear as an attachment above the text input bar.
- If you start typing and send a message, the attached document will automatically be included in the conversation.
Ask Questions About the Document
- Type your question into the chat and press Enter.
- The Model will process the document and respond based on its content.
Tip: For details on where documents are stored and how to access them in different storage modes, read the Conversation Storage Options article in the Help Center.
Access and Manage All Uploaded Documents
- To find previously uploaded documents, click the pages icon at the bottom left of the chat window.
- A Files popup window will appear, showing all available files.
- From this window, you can:
- Download a file
- Delete a file
- Sort by file type, date created, or name
- Click a file’s name to add it to the current chat.
|
File Category
|
File Extension(s)
|
Usage Notes / Limitations
|
|
Presentation
|
.ppt
|
- Convert to .pptx if possible, for better compatibility
|
|
.pptx
|
|
|
Web
|
.html
|
|
|
Data
|
.csv
|
- RAG for .csv and .xlsx files is currently unreliable. Large or complex files may return a result once but fail on subsequent requests. We’re working on a fix. Thank you for your understanding.
|
|
.xlsx
|
|
|
.json
|
|
|
Document
|
.docx
|
|
|
.pdf
|
- RAG on PDFs can be unreliable. Large or complex files may succeed once but fail on subsequent requests, and we’re working on a fix.
- Born-digital documents work best, while scanned images often need OCR and may yield imperfect results.
|
|
.md
|
|
|
.txt
|
|
|
Image
|
.jpg
|
|
|
.png
|
|
|
.gif
|
|
|
.heic
|
CometAI currently does not support HEIC image files.
Recommended workaround:
- Open the image on your device
- Save or export it as JPG or PNG format
- Upload the converted file
|
|
Script
|
.py
|
|
|
.sh
|
|
|
Archive
|
.zip
|
- Supported only for flat ZIPs - files must be at the root of the archive (no nested folders). If folders are present, unzip and re-zip with files at the top level.
|
- Limitations for file uploads are still being tested.
- Uploaded file performance depends on the model’s context size and file complexity.
- Large or complex files may take longer to process.
- For best results with spreadsheets, keep data sets small and use a higher-capability model.
- In the File Manager pop-up, lists with 5 or more files may not fully scroll. The last file can appear clipped (only the top half is visible). You can still select it, but details like name and size are hidden.
- Workaround: click the Created column to sort in ascending order so the last file becomes fully visible. We are working on a fix for it. Thank you for your understanding.
Best Practices
- Use a higher-context model such as o4 mini or Claude 3.7 Sonnet v2.
- For spreadsheets, limit the file to ~100 rows or fewer or best performance.