Llama-Zenity: A Minimal GUI to Launch Llama.cpp

Overview

Llama-Zenity is a minimal graphical interface to launch Llama.cpp . It works on Linux and MacOS, and maybe Windows.

Linux Installation

Dependencies

• Zenity: sudo apt update && sudo apt install zenity
• Llama.cpp: You should follow installation instructions on their website. I recommend installing into a directory called src under your home directory (i.e. $HOME/src/llama.cpp/).

Install

1. Edit Llama-Zenity.desktop in the line Exec. Make sure your home directory is correct (probably not just User). Also make sure the paths to this repository and to the llama.cpp repository are correct.

   editor Llama-Zenity.desktop

2. I recommend storing your GGUF model files in directory called models under your home directory (i.e. $HOME/models/). If you keep them somewhere else, you can modify MODEL_DIR in llama-zenity.sh .

   editor llama-zenity.sh

3. Copy Llama-Zenity.desktop to your desktop.

   cp Llama-Zenity.desktop $HOME/Desktop/

That's it! Now you can click on the desktop icon and start chatting away.

MacOS Installation

Dependencies

• Zenity: brew install zenity
• Llama.cpp: You should follow installation instructions on their website. I recommend installing into a directory called src under your home directory (i.e. $HOME/src/llama.cpp/).

Install

1. Edit the file Llama-Zenity.command . Make sure your home directory is correct (probably not just user). Also make sure the paths to this repository and to the llama.cpp repository are correct.

   open -t Llama-Zenity.command

2. I recommend storing your GGUF model files in directory called models under your home directory (i.e. $HOME/models/). If you keep them somewhere else, you can modify MODEL_DIR in llama-zenity.sh .

   open -t llama-zenity.sh

3. Copy Llama-Zenity.command to your desktop.

   cp Llama-Zenity.command $HOME/Desktop/

That's it! Now you can click on the desktop icon and start chatting away. You can also drag the icon to the dock, right of the separator.

Contributions

Contributions are welcome, via pull requests.