Window Recall for Linux

To be renamed soon
Currently in proof of concept stage

Introduction

This is a simple program that is kind of like the new Windows recall feature.

Disclaimer/Notes

• I am not a AI expert or a machine learning expert. I am just a student who is trying to build some stuff by FAFO.
• This is a proof of concept and is not meant to be used in production.
• This hasn't been tested on any other system other than my own. So, it might not work on your system.
  • This will work on KDE Wayland provided you install the necessary dependencies.
  • Currently its Linux only but may soon add support for Windows and MacOS.
• This will probably drain your battery. So, use it at your own risk.
• This program doesn't provide any sort of encryption or security(yet). So, use it at your own risk.

Dependencies/Pre-requisites

Since it's just a proof of concept, I haven't added any checks for the dependencies. So, you need to install them manually.

• Have ollama installed and running with the following models downloaded
  • phi3
  • moondream
  • mxbai-embed-large
• Docker (Will be used for Qdrant database)
• Have Qdrant installed and running
• kdotool (Might already be in your package manager)
• tesseract

How to run?

• Make sure ollama & Qdrant are running.

• Install requirements using pip install -r requirements.txt

• Run the service.py file in the background so that it automatically takes screenshots. (Preferrably set it up as a service on ur system or have it on auto start)

• Run client.py to use the GUI. (Currently its very barebone, I will improve it overtime)

To Do

• Add support for other Linux environments
• Add support for Windows and MacOS
• Add more features to the client
• Improve the prompts
• Save the files opened by a program as well.
• Add option to reopen the files in the same program
• Add option to reopen the files in a different program
• Experiment with different models and see which one works best
• Improve the UI of the client
• Add encryption and security

Liked my efforts? Consider sponsoring me on GitHub or buying me a coffee.

Also, please star the repo, it motivates me to work on it more.