Code Connect

Code Connect is a realtime collaborative coding workspace with authentication, shared rooms, live chat, file sync, drawing, code execution, and AI-assisted analysis.

Project Layout

docker-compose.yml        Local MongoDB / Kafka / Zookeeper / Redis stack
start-dev.ps1             Windows helper to start the full stack
client/                   Vite + React frontend
server/                   Express + Socket.IO backend
judge-worker/             Kafka consumer and Docker-based code execution worker

What You Need

• Node.js 18+ and npm
• Docker Desktop or Docker Engine for the local MongoDB, Kafka, Zookeeper, and Redis stack
• A backend JWT secret
• A reachable ML agent service for AI generation and analysis
• Optional: MongoDB Atlas if you do not want to use the local MongoDB container

URLs And Ports

Service	URL	Purpose
Frontend	http://localhost:5173	Vite dev server
Backend	http://localhost:3000	Express and Socket.IO API
MongoDB	mongodb://127.0.0.1:27017	Local MongoDB port
Kafka	localhost:9092	Broker port
Zookeeper	localhost:2181	Kafka coordination
Redis	localhost:6379	Pub/Sub relay for execution updates
ML agent	https://code-plag-fastapi.onrender.com	Default AI endpoint used by the backend

Environment Files

Copy the example files below into local .env files before starting the app.

client/.env.example

VITE_BACKEND_URL=http://localhost:3000

server/.env.example

PORT=3000
MONGO_URI=mongodb://127.0.0.1:27017/codeconnect
JWT_SECRET=replace-with-a-long-random-secret
KAFKA_BROKERS=localhost:9092
REDIS_HOST=localhost
REDIS_PORT=6379
ML_AGENT_URL=https://code-plag-fastapi.onrender.com

judge-worker/.env.example

KAFKA_BROKERS=localhost:9092
KAFKA_GROUP_ID=judge-workers
REDIS_HOST=localhost
REDIS_PORT=6379

Notes:

• If you use MongoDB Atlas, replace MONGO_URI with your Atlas connection string.
• The server also falls back to mongodb://127.0.0.1:27017/codeconnect if MONGO_URI is not set.
• There is no GEMINI_API_KEY in this repository. The backend calls the external ML agent instead.
• judge-worker uses the local Kafka and Redis defaults above. If you move those services elsewhere, set the matching environment variables before starting it.

Fresh Clone Setup

1. Clone the repository.
2. Copy the example environment files into place:

   Copy-Item client\.env.example client\.env
   Copy-Item server\.env.example server\.env
   Copy-Item judge-worker\.env.example judge-worker\.env

3. Update server/.env with a strong JWT_SECRET. Keep the local MONGO_URI for Docker Compose, or replace it with your MongoDB Atlas connection string.
4. Start the full stack from the repository root:

   .\start-dev.ps1

   The helper script starts Docker Compose and opens backend, worker, and frontend windows. It also installs dependencies the first time if a folder does not yet have node_modules.
5. If you prefer manual startup, run the services in this order: Infrastructure:

   docker compose up -d

   Backend:

   cd server
   npm install
   npm run dev

   Judge worker:

   cd judge-worker
   npm install
   npm run dev

   Frontend:

   cd client
   npm install
   npm run dev

   Docker Compose starts MongoDB, Kafka, Zookeeper, and Redis. The three npm commands start the backend, worker, and frontend.
6. Open http://localhost:5173 in your browser.

Where The Backend And AI Calls Live

• server/server.js connects to MongoDB, verifies JWT auth, serves room data, calls the ML agent at /generate and /analyze, and publishes execution events through Socket.IO.
• server/migrations/migrate_generated_code.js connects to MongoDB for the generated-code migration.
• judge-worker/worker.js consumes Kafka jobs, runs code inside Docker containers, and publishes results to Redis.
• client/src/context/AppContext.jsx verifies the stored login token and checks ML agent health.
• client/src/context/SocketContext.jsx connects the frontend to the backend socket.
• client/src/components/forms/AuthComponent.jsx sends the login and signup requests.
• client/src/components/editor/AdminPanel.jsx fetches room data from the backend.
• client/src/components/tabs/AnalysisTab.jsx renders the AI analysis returned by the backend.

Feature Notes

• The Gemini, ChatGPT, and Claude labels in the UI are model output tabs, not direct API keys in this repository.
• If the ML agent is unreachable, the core editor can still load, but AI generation and analysis will fail.
• If Kafka, Redis, or the judge worker are not running, code execution will not work.