• Arch: x86_64
• Size: 7.4M
• Vers: v01.01.01
• MusicGen: https://facebook-musicgen.hf.space/ (broken)
• OpenAI: https://openai.com
• Stable Diffusion: https://stablediffusionapi.com
Gh0sty is a CLI chatbot that uses OpenAI, Stable Diffusion and MusicGen to generate responses. Shes designed with effortlessness in mind and can be put to extensive use, with the sacrifice of significantly longer API call times than the web interfaces for the services mentioned (at standard connection speeds). The quality of the response content itself isnt affected by the fact that the call is being made locally, which is great. Text to speech (gTTS) is enabled by default and responses are spoken in a tolerable voice. A keyword system is used for basic controls and navigation between ChatGPT, Stable Diffusion, MusicGen, exiting and reloading etc.
• mpv: Free and open source media player
• python39: Major release 3.9 of the Python high-level programming language
• python-colorama: Cross platform colored terminal text
• python-gradio-client: UI generation library for machine learning models
• python-gtts: Google text to speech API wrapper
• python-mpv: Python bindings for the mpv media player
• python-openai: Python library for interfacing with OpenAI language models
• python-pydub: Audio processing library for Python
• python-pygame: Python library for game development
• python-termcolor: Cross platform colored terminal text
• qview: Qt image viewer
Open a terminal and clone this repo:
git clone 'https://github.com/c0rNCh1p/gh0sty.git' ||
git clone 'https://gitlab.com/c0rNCh1p/gh0sty.git'Go grab git bash if not installed 'https://git-scm.com/download/win.' Open Git Bash and follow steps for MacOS and Linux.
To put this chatbot to full use, valid API key for OpenAI and Stable Diffusion are needed. These keys should be saved in files named 'stable_diffusion_api_key' and 'openai_api_key' in the program directory. When the install.sh script is run it will copy the keys over to the install directory with the code:
echo 'sk_abcdefg123456789' >'gh0sty/keys/openai_api_key'
echo 'abcdefg123456789' >'gh0sty/keys/stable_diffusion_api_key'The 'install.sh' script will take care of the rest of the installation:
cd 'gh0sty'
chmod 764 'install.sh'
./'install.sh'These are the defined keywords:
<exit-|ex-|e-> leave the chat
<keywords-|kw-|k-> show this message
<reload-|rl-|r-> kill response and reset
<image-|ig-|i-> text to image (Stable Diffusion)
<music-|mc-|m-> text to audio (MusicGen)
<cancel-|rc-|c-> return to text chatThe main source file 'gh0sty.py' contains configuration settings at the beginning of the file for Chat GPT and Stable Diffusion response generation. Fine tuning and training can be implemented via the 'presets.py' file for further customization.
Support for English text to speech functionality is provided. This can be changed via the lang variable in the source code, but other languages have not been tested.
Text navigation actions like the left and right keys or the home and end keys to move the cursor are not supported (backspace works fine). However support for multiline input is provided, allowing users to enter more complex queries or code snippets as prompts. While the absence of traditional text editing shortcuts might limit some abilities, its easier to write complex prompts in a text editor and paste them into the input field anyway.
The prompt size includes the the entire session history and the preset data, so the limit can be reached fairly quickly after a few large prompts. The current current prompt size limit is approximately 120 lines if its the first prompt and no history has been built up. Habitually restarting the session after responses to large inputs are generated is the best workaround. Remembered context is nice, but also unnecessary if prompts are structured correctly to begin with.