News Intelligence Platform

A full-stack news intelligence application that fetches news articles from RSS feeds, stores them in a vector database, and provides an AI-powered chat interface to query and retrieve relevant news information using semantic search and Google Gemini AI.

🏗️ Architecture

This project consists of two main components:

• chat-bot - Backend API server built with Bun, Express, and TypeScript
• chat-client - Frontend web application built with Next.js and React

How It Works

1. News Ingestion: RSS feeds are periodically fetched (every 55 minutes) using Inngest scheduled jobs
2. Vector Storage: News articles are processed, chunked, and stored as embeddings in Qdrant vector database
3. Semantic Search: User queries are converted to embeddings and used to find relevant articles
4. AI Responses: Google Gemini AI generates contextual answers based on retrieved articles

🚀 Tech Stack

Backend (chat-bot)

• Runtime: Bun - Fast JavaScript runtime
• Framework: Express.js
• Database: PostgreSQL with Prisma ORM
• Vector Database: Qdrant
• AI: Google Gemini 2.0 Flash
• Task Scheduling: Inngest
• Other: RSS Parser, Axios, CORS, Helmet, Morgan

Frontend (chat-client)

• Framework: Next.js 16
• UI: React 19, Tailwind CSS
• Language: TypeScript

📁 Project Structure

.
├── chat-bot/              # Backend API server
│   ├── src/
│   │   ├── config/        # Configuration files (Qdrant, news URLs)
│   │   ├── controllers/   # Route controllers
│   │   ├── services/      # Business logic (chat, embedding, retrieval, etc.)
│   │   ├── route/         # API routes
│   │   ├── types/         # TypeScript types
│   │   ├── utils/         # Utility functions
│   │   ├── inngest/       # Inngest scheduled functions
│   │   ├── app.ts         # Express app setup
│   │   └── server.ts      # Server entry point
│   ├── prisma/            # Database schema and migrations
│   ├── docker-compose.yml # Docker services configuration
│   └── package.json
│
└── chat-client/           # Frontend web application
    ├── app/               # Next.js app directory
    ├── public/            # Static assets
    └── package.json

🛠️ Setup Instructions

Prerequisites

• Bun installed (v1.3.1+)
• Node.js installed (for frontend)
• PostgreSQL database
• Qdrant vector database (can be run via Docker)
• Google Gemini API key

Backend Setup (chat-bot)

1. Navigate to the backend directory:

cd chat-bot

2. Install dependencies:

bun install

3. Set up environment variables:

cp .env.example .env

Edit .env and configure:

# Database
DATABASE_URL="postgresql://user:password@localhost:5432/news_intelligence"

# Qdrant
QDRANT_URL="http://localhost:6333"
QDRANT_COLLECTION="news_vectors"

# Google Gemini
GEMINI_API_KEY="your-gemini-api-key"

# Server
PORT=5000

# Inngest (if using cloud)
INNGEST_EVENT_KEY="your-inngest-event-key"
INNGEST_SIGNING_KEY="your-inngest-signing-key"

4. Set up the database:

bunx prisma migrate dev
bunx prisma generate

5. Start Qdrant (if using Docker):

docker-compose -f qdrant.docker-compose.yml up -d

6. Run the backend server:

# Development mode (with watch)
bun run dev

# Production mode
bun run start

The API server will run on http://localhost:5000

Frontend Setup (chat-client)

1. Navigate to the frontend directory:

cd chat-client

2. Install dependencies:

npm install
# or
bun install

3. Run the development server:

npm run dev
# or
bun dev

The frontend will run on http://localhost:3000

Using Docker

You can also run the entire stack using Docker:

cd chat-bot
docker-compose up -d

📡 API Endpoints

POST /api/chat

Send a query to get AI-generated answers based on news articles.

Request:

{
  "query": "What are the latest developments in technology?"
}

Response:

{
  "answer": "Based on the latest news articles...",
  "responseTimeMs": 1234
}

🔄 Scheduled Jobs

The application uses Inngest to run scheduled tasks:

• News Fetching: Runs every 55 minutes
  • Fetches news from RSS feeds (currently BBC News)
  • Processes and chunks articles
  • Generates embeddings
  • Stores in Qdrant vector database

🔑 Environment Variables

Backend (.env)

Variable	Description	Required
DATABASE_URL	PostgreSQL connection string	Yes
QDRANT_URL	Qdrant server URL	Yes
QDRANT_COLLECTION	Vector collection name	Yes
GEMINI_API_KEY	Google Gemini API key	Yes
PORT	Server port	No (default: 5000)
INNGEST_EVENT_KEY	Inngest event key	Optional
INNGEST_SIGNING_KEY	Inngest signing key	Optional

📊 Database Schema

Interaction Model

Stores chat interactions and responses:

• id: Unique identifier
• sessionId: User session identifier
• userQuery: User's question
• llmResponse: AI-generated response
• responseTimeMs: Response time in milliseconds
• createdAt: Timestamp

🔍 Features

• ✅ RSS feed parsing and ingestion
• ✅ Semantic search using vector embeddings
• ✅ AI-powered question answering with Google Gemini
• ✅ Scheduled news updates via Inngest
• ✅ RESTful API for chat interactions
• ✅ Modern Next.js frontend
• ✅ PostgreSQL for data persistence
• ✅ Qdrant for vector similarity search

🧪 Development

Running in Development Mode

Backend:

cd chat-bot
bun run dev  # Runs with watch mode

Frontend:

cd chat-client
npm run dev  # Runs Next.js dev server

Database Migrations

cd chat-bot
bunx prisma migrate dev    # Create a new migration
bunx prisma generate       # Generate Prisma Client
bunx prisma studio         # Open Prisma Studio GUI

📝 License

This project is private.

🤝 Contributing

This is a private project. For contributions, please contact the repository owner.

---

Built with ❤️ using Bun, Next.js, and Google Gemini AI