nGPT

A simple, powerful, local, and open source AI.

PLEASE USE ONLY THE LATEST VERSION. OLDER VERSIONS CANNOT HAVE ANY BUGS FIXED OR CHANGES.

About

nGPT was created for a few reasons:

1. Companies like OpenAI are trying to make it harder to run AI locally.
2. There wasn't really an easy way to have an AI do more things besides a few simple ones
3. I want more people to have access to a simple, open-source, local AI that can do many things.

Features

As of now, the latest publicly-available model of nGPT (nGPT 2M) has the current features (without checkmarks = buggy or in progress):

• A nice, web hosted Flask Python UI.
• The ability to generate songs (via DeepAI and Selenium + Undetected Chromedriver).
• The ability to turn on/off your WIZ smart lights with no additional setup required.
• List devices on Wi-Fi and Bluetooth
• The ability to search the web.

Ideas from other sources in progress:

• List devices on Wi-Fi and Bluetooth
• Generate pixel art textures/skins

Installation

Important

IT IS STRONGLY AND HIGHLY RECOMMENDED TO HAVE A DEDICATED GPU. OTHERWISE, IT IS NOT ADVISED TO USE nGPT.

Method 1: LM Studio (WINDOWS)

• Download and install LM Studio.
• In LM Studio, in the search icon section at the bottom left, type in "Meta-Llama-3-8B-Instruct" and hit CTRL+ENTER.

Note

While nGPT can theoretically work with any model, Meta Llama 3 8B Instruct seems to work the best.

• Click the first result by LM Studio Community and click download.
• Go into LM Studio and click the console icon.
• Click "Load a Model" at the top.
• Click the model you downloaded (Meta Llama 3 8B Instruct).
• Use default settings and load the model.
• Press CTRL+R to start the server.
• Go back to the GitHub repo, go to Releases, then click on the version of nGPT you wish to install.
• Download the .py file for Windows.
• In the folder that you installed nGPT, open a Command Prompt, CD into that folder, and type python app.py.
• In the Command Prompt window, CTRL+CLICK the first link to be brought to nGPT WebUI.
• Have fun using Experimental nGPT!

Tip

Make sure to click the save button in the top-right corner to save your current chat. These will be saved locally so they will always be there. nGPT does not autosave.

Whenever you want to use nGPT, you will need to load the model and start the server again:

• Go into LM Studio and click the console icon.
• Click "Load a Model" at the top.
• Click the model you downloaded (Meta Llama 3 8B Instruct).
• Use default settings and load the model.
• Press CTRL+R to start the server.
• Go into the folder you have nGPT in, open a Command Prompt, CD into that folder, and type python app.py.
• In the Command Prompt window, CTRL+CLICK the first link to be brought to nGPT WebUI.

Note

LINUX is currently not supported, however it will be in the future.

I do NOT and will NOT support MacOS. This is mostly due to Apple's prices, ecosystem, and locking down devices with "Activation Lock", as well as doing many other money-generating practices which other companies unfortunately follow. In order to combat this, I do not want to make anything available for their product(s) as a sort of 'strike.' If you would like to test nGPT on MacOS, please run Windows or Linux with Parallels or another VM program.