search

gptscript-ai / desktop

MIT License
19 stars 13 forks source link

Vision tool is not able to work with files in default workspace. #138

Open sangee2004 opened 1 month ago

sangee2004 commented 1 month ago

Version - 0.10.0-rc2 (0.10.0-rc2)

Steps to reproduce the problem:

  1. Create an assistant with vision and workspace tools.
  2. Start a chat.
  3. Upload an image file to workspace.
  4. Ask for describing the image in file from step 3. Notice that vision tool is not able to work with files in default workspace
Screenshot 2024-08-13 at 4 32 06 AM Screenshot 2024-08-13 at 4 32 39 AM

It falls back to using sys.read and fails with following errrors

Screenshot 2024-08-13 at 4 39 33 AM
g-linville commented 1 week ago

The fix for this has been merged

sangee2004 commented 1 week ago

Tested with build - 18eaca7d8

  1. Create an assistant with Vision and workspace tool.
  2. Chat with this tool
  3. Add grocerylist.png file to workspace.
  4. Ask to describe grocerylist.png file

Vision tool is not able to process the file and I get the following error

It seems there is a persistent issue with accessing the image file. The path appears to be correct, but the tool is unable to open it.

Would you like to try uploading the image again or provide any other instructions?

https://github.com/user-attachments/assets/148b5242-c4b7-473b-8c69-c17068e1ab68