Real-time Distributed Chat

This repository includes the frontend (React.js), backend (Go-Fiber), and code necessary to create a scalable, real-time messaging platform. Developers can use this project to learn about system design and gain insights into building end-to-end applications.

* * * * * *

Architecture Diagram 🌠

About the Project 🌌

• Multiple Go-Fiber servers provide API endpoints (JWT authentication) and WebSocket connections for full-duplex communication. These Go instances are configured under Nginx as a layer 7 load balancer.
• To distribute messages efficiently among users connected to different instances but in the same room, Redis is utilized using a Pub/Sub model. Each instance subscribes to specific channels in Redis and receives notifications upon receiving messages. All messages are persisted in PostgreSQL.
• The database handles a heavy write load, capable of processing up to 100 messages per second, by using Apache Kafka, a high-throughput, low-latency message streaming platform. A Kafka consumer (implemented in Go) fetches messages in batches and writes them to PostgreSQL.
• The frontend of the application is built with React.js and served via an Nginx container. All nodes are containerized using Docker and orchestrated with Docker Compose. Only the Reverse-Proxy (Nginx) is exposed to external requests, ensuring security and centralized request handling.

Setting Up 🔧

• Create a .env file based on env.sample.
• Populate the .env file with your configuration values.
• Ensure the .env file is in the same directory as docker-compose.yml.

# Redis Config
REDIS_PORT=6379
REDIS_HOST=redis

# Database Config
POSTGRES_DATABASE=chat_db
POSTGRES_USER=postgres
POSTGRES_PASSWORD=postgres
POSTGRES_HOST=postgres
POSTGRES_PORT=5432

# Kafka config
KAFKA_HOST=kafka
KAFKA_PORT=9092
KAFKA_TOPIC=chat_messages
KAFKA_GROUP_ID=chat_group
ZOOKEEPER_PORT=2181

# Authentication
JWT_SECRET=secret

# Reverse proxy config
NGINX_ENV=local
NGINX_PORT=8080
NGINX_HOST=localhost

# LLM config
LLM_PORT=11434

# Backend Servers
SERVER_PORT=8080

Running the Application

# Uncomment the relevant section in docker-compose.yml if you wish to run the LLM model (requires approximately 3.6GB in size).

To build the application containers, execute the following command:

$ docker-compose up --build

If the application starts successfully, navigate to http://localhost:8080/ in your web browser.

To use the LLM model, begin your message with @superchat followed by your text. Enjoy enhanced chat capabilities with natural language processing!