Aura

A full-stack anonymous social platform with AI-powered content moderation. Built on the MERN stack with a multi-service Docker Compose architecture — the ML classifier runs as its own independent service, not embedded in the main backend.

---

Architecture

Three independently deployable services:

┌─────────────────┐     ┌─────────────────┐     ┌──────────────────────┐
│  React client   │────▶│  Express/Node   │────▶│  Python classifier   │
│  (Vite + Redux) │     │  + Socket.io    │     │  (Flask + sklearn)   │
└─────────────────┘     └────────┬────────┘     └──────────────────────┘
                                 │
                         ┌───────▼───────┐
                         │   MongoDB     │
                         └───────────────┘

Why a separate classifier service? Keeps ML dependencies (Python, scikit-learn, Perspective API) isolated from the Node.js runtime. The classifier can be retrained and redeployed independently without touching the backend. It's also how you'd run this in production — not as a subprocess.

Context-aware auth — login verification considers IP, location, and device metadata, not just credentials. Suspicious context (new device + new location) triggers additional verification.

GitHub Actions CI/CD — automated build and test pipeline on every push.

---

Features

• Threaded discussions with infinite recursive comment nesting
• Upvote/downvote with hot, new, top, and rising feed sorting
• Real-time interactions via Socket.io
• AI content moderation — posts are screened by the Python classifier before going live
• Dark mode with system preference detection
• Full GitHub Flavored Markdown rendering in posts and comments
• Community flairs, NSFW/Pinned/Locked post indicators

---

Stack

Layer	Technology
Frontend	React 18, Vite, Redux Toolkit, Tailwind CSS, Framer Motion
Backend	Node.js, Express, MongoDB (Mongoose), Socket.io, Passport.js
ML Moderation	Python, Flask, scikit-learn, Perspective API
Infrastructure	Docker Compose, GitHub Actions

---

Getting started

One-command setup (Docker):

git clone https://github.com/aarogyaojha/Aura.git
cd Aura
docker-compose up --build

This starts all three services — client, server, and classifier — plus MongoDB.

Manual setup:

# Clone
git clone https://github.com/aarogyaojha/Aura.git
cd Aura

# Configure environment
cp server/.env.example server/.env
cp client/.env.example client/.env

# Start services (each in its own terminal)
cd server && npm start
cd client && npm run dev
cd classifier_server && python classifier_api.py

---

License

MIT © Aarogya Ojha