NestJS Chat Application

A real-time group chat application built with NestJS, MongoDB, Socket.IO, and AWS S3. This application provides secure group chat functionality with features like user authentication, file sharing, and real-time typing indicators.

Features

• 🔐 JWT-based Authentication
• 👥 Group Chat Management
• 📁 File Sharing (Images)
• ⚡ Real-time Communication
• ✍️ Typing Indicators
• 📝 Message Management
• 🖼️ Profile & Group Images
• 📚 Swagger API Documentation
• 🧠 Embeddings for RAG (Retrieval-Augmented Generation)
• 💬 AI-Powered Smart Replies
• 📊 OpenAI API Usage Tracking

Tech Stack

• Framework: NestJS
• Database: MongoDB
• Real-time Communication: Socket.IO
• File Storage: AWS S3
• Authentication: JWT, Passport
• API Documentation: Swagger
• Process Manager: PM2
• Containerization: Docker
• AI Integration:
  • LangChain
  • OpenAI
    • GPT-3.5-Turbo for completions
    • text-embedding-3-large for embeddings (256 dimensions)

Prerequisites

• Node.js (v18 or later)
• MongoDB
• AWS Account (for S3)
• OpenAI API Key and Organization ID
• Docker (optional)
• PM2 (optional)

Environment Setup

Create a .env file in the root directory:

# MongoDB Configuration
MONGODB_URI=mongodb://localhost:27017/chat_app
MONGODB_USER=your_mongodb_user
MONGODB_PASSWORD=your_mongodb_password

# JWT Configuration
JWT_SECRET=your_jwt_secret_key
JWT_EXPIRATION=24h

# AWS Configuration
AWS_ACCESS_KEY_ID=your_aws_access_key
AWS_SECRET_ACCESS_KEY=your_aws_secret_key
AWS_REGION=us-east-1
AWS_BUCKET_NAME=your-bucket-name

# OpenAI Configuration
OPENAI_API_KEY=your_openai_api_key
OPENAI_ORG_ID=your_openai_org_id

# Application Configuration
NODE_ENV=development
APP_VERSION=1.0.0

# Embedding Configuration
# Options: 'cosine', 'euclidean', 'dot-product'
SIMILARITY_ALGORITHM=dot-product

Installation

# Install dependencies
npm install

# Build the application
npm run build

# Start the application
npm run start:dev

RAG (Retrieval-Augmented Generation) Features

This application includes vector embeddings generation for chat messages to enable AI-powered features:

• All text messages are automatically embedded using OpenAI's text-embedding-3-large model
• Embeddings are stored with the messages in MongoDB
• This enables future features like semantic search, smart replies, and summarization
• Image captions can also be embedded for searchability

The embedding generation happens automatically:

• When sending a message via REST API
• When sending a message via WebSocket
• For both text messages and image captions

These embeddings enable future AI features without requiring access to the entire message history.

Smart Reply Feature

The application offers AI-generated smart reply suggestions for group chats:

• Get contextually relevant reply suggestions with a single API call
• Suggestions are generated based on recent message history
• Uses OpenAI's GPT-3.5-Turbo for efficient and cost-effective generation
• Each request returns 3 suggested replies tailored to the conversation
• Client applications can display these as quick-reply options

Smart Reply Implementation Details

1. Message Selection:

   • Uses embeddings to find the 5 most contextually relevant messages
   • Supports multiple similarity algorithms (configurable via SIMILARITY_ALGORITHM)
   • Falls back to recent messages if embeddings are unavailable

2. OpenAI Integration:

   • Uses GPT-3.5-Turbo for cost-effective completions
   • Includes user context and conversation history
   • Generates natural, contextually appropriate responses
   • Limited to 150 characters per suggestion

3. Request Tracking:

   • All OpenAI API calls are tracked in the OpenAI dashboard
   • Requests are tagged with:
     • User identifier (username_feature_environment)
     • Application name and version
     • Environment (development/production)
   • Enables monitoring of API usage and costs

To use this feature, make a POST request to /api/messages/smart-replies with the following body:

{
  "group_id": "your_group_id"
}

The response will contain an array of suggested replies:

{
  "suggestions": [
    "That's a great point!",
    "I agree with your approach.",
    "Let's discuss this further."
  ]
}

API Documentation

API documentation is available via Swagger UI at /api/docs when running in development mode.