PromptGenie is a Command Line Interface (CLI) assistant designed to provide quick responses to a wide range of queries.
PromptGenie utilizes llama.cpp's main binary and a quantized language model to generate responses. The assistant can be invoked with the command pg followed by a prompt, like pg are pisachios healthy? or pg how to show free space in GB?.
PromptGenie consumes minimal system resources and can drastically reduce the time between thought and response, avoiding the need to move away from the command-line. The current configuration only uses the CPU, but you can easily edit the pg.sh file to use your GPU.
PromptGenie will optionally use glow to render markdown. The only downside of this is responses aren't be streamed to the console any more. In practice this isn't really an issue; there's a spinner to watch while you wait, and you can set a timeout value to prevent things getting out of hand.
- GNU/Linux - Windows compatibility may come soon.
- The
mainbinary from llama.cpp. - A GGML format language model.
- Charmbracelet's glow package for markdown rendering at the CLI.
- Find the installation instructions here: https://github.com/charmbracelet/glow
- Clone the repository:
-
git clone https://github.com/morpheus2448/promptgenie.git -
cd promptgenie
- Run the install.sh script to add an alias to your .bashrc file:
-
cd ~/pg -
./install.sh
- Edit the settings.txt file located in the pg folder within your home directory.
- Update the paths for llama.cpp and your GGML models.
- Decide if you want to use glow and set the variable accordingly.
- Change inference settings as you see fit.
`pg are pistachios healthy?`