Overview

This is a demo social-media platform that uses AI to assess toxicity in posts and comments. Users can create accounts, submit posts and comments, and receive a toxicity score generated by an ONNX model. The core features include:

• User registration and authentication
• Creating posts and comments
• Real-time toxicity scoring using a quantised ONNX model

Prerequisites

Ensure you have the following installed:

• Python 3.11+
• pip to install Python dependencies
• Node.js 22+ and pnpm for the Next.js frontend/backend
• A UNIX-style shell (macOS, Linux or WSL)

Setup & Installation

1. Clone repository

git clone https://github.com/newbee1905/aix.git
cd aix

2. Python Environment

cd scripts
python -m venv venv
source venv/bin/activate

pip install --upgrade pip
pip install -r requirements.txt

3. Install Node.js dependencies

pnpm i

Extracting & Quantising the Toxicity Model

cd scripts
source venv/bin/activate

python export_and_quantize.py                

cp -r ../public/models ../models

If you want to change the model, please use the username/model-name format on hugging face, and change the variable TOX_MODEL_ID.

Please also change the variable TOX_MODEL_ID inside src/lib/onnx.ts.

Environment Variables

Please create a .env with your database information. You can configure the JWT exiration and Refresh Token Expiration to your liking.

POSTGRES_PRISMA_URL="postgres://username@localhost:5432/database_name"
JWT_SECRET=your-secret
JWT_EXPIRES_IN=86400 # 1 day 
REFRESH_EXPIRES_IN=2592000 # 30 days
REFRESH_ROTATE_BEFORE=1296000 # 15 days

Running the Application

pnpm dev