RAG Q&A Chatbot

An intelligent Retrieval-Augmented Generation (RAG) chatbot that answers questions based on your uploaded documents. Features advanced document processing, vector similarity search, and multiple AI language model integrations.

Features

Core Capabilities

Document Upload & Processing: PDF, Word, Text, CSV, HTML, JSON support
Intelligent Search: Vector-based semantic similarity search
AI-Powered Responses: Multiple LLM providers (OpenAI, Claude, Gemini, Cohere, HuggingFace)
Real-time Chat: WebSocket streaming responses
Source Attribution: See which documents were used for answers
Beautiful Interface: Modern, responsive web UI

Document Processing

Format Support: PDF, DOCX, TXT, MD, CSV, HTML, JSON
Text Extraction: Advanced parsing for all supported formats
Chunking: Intelligent text segmentation with overlap
Embeddings: Vector representations for semantic search

AI Integration

Multiple Providers: OpenAI GPT, Anthropic Claude, Google Gemini, Cohere, HuggingFace
Fallback System: Works without API keys using free models
Streaming: Real-time response generation
Context-Aware: Uses retrieved document chunks for accurate answers

User Experience

Drag & Drop Upload: Easy document management
Live Chat: Real-time conversation with typing indicators
Source Preview: Click sources to see relevant content
Settings Panel: Customize search parameters
Responsive Design: Works on desktop and mobile

Quick Start

1. Installation

# Clone or download the project
git clone <repository-url>
cd rag-qa-chatbot

# Install dependencies
npm install

# Run setup (creates .env file and sample documents)
npm run setup

2. Configuration (Optional)

Edit .env file to add your API keys for enhanced AI capabilities:

# OpenAI (GPT models) - Get free credits at platform.openai.com
OPENAI_API_KEY=your_openai_api_key_here

# Anthropic Claude - Get free credits at console.anthropic.com
ANTHROPIC_API_KEY=your_anthropic_api_key_here

# Google Gemini - Free tier available at makersuite.google.com
GOOGLE_API_KEY=your_google_api_key_here

# Cohere - Free tier available at cohere.ai
COHERE_API_KEY=your_cohere_api_key_here

# Hugging Face (for free models) - Sign up at huggingface.co
HUGGINGFACE_API_KEY=your_huggingface_api_key_here

Note: The system works out-of-the-box with free HuggingFace models, no API keys required!

3. Start the Server

# Production mode
npm start

# Development mode (auto-reload)
npm run dev

4. Open in Browser

Navigate to: http://localhost:3000

How to Use

1. Upload Documents

Drag and drop files or click the upload area
Supported formats: PDF, Word, Text, CSV, HTML, JSON
Files are automatically processed and indexed

2. Ask Questions

Type questions about your uploaded documents
Use natural language queries
Get AI-powered answers with source citations

3. Example Queries

"What are the main points in this document?"
"Summarize the key findings"
"What recommendations are mentioned?"
"Compare different approaches discussed"
"Find information about [specific topic]"

4. Quick Actions

Use predefined quick action buttons
Summarize documents
Find key insights
Compare approaches

🛠️ Technical Architecture

Backend Components

Express.js Server: RESTful API and WebSocket support
Document Processor: Multi-format text extraction
Vector Store: In-memory similarity search
Embedding Service: Text vectorization
LLM Service: Multiple AI provider integration
Retrieval Service: Context-aware document search

Frontend Components

Modern Web UI: HTML5, CSS3, JavaScript
Real-time Chat: WebSocket communication
File Upload: Drag & drop with progress
Responsive Design: Mobile-friendly interface

AI Pipeline

Document Upload → Text extraction and cleaning
Text Chunking → Intelligent segmentation with overlap
Embedding Generation → Vector representations
Vector Storage → In-memory indexing
Query Processing → Similarity search
Response Generation → AI-powered answers

🔧 Configuration Options

Environment Variables

# Server Configuration
PORT=3000
NODE_ENV=development

# LLM Provider (choose your preferred)
DEFAULT_LLM_PROVIDER=huggingface  # openai, anthropic, google, cohere

# Document Processing
CHUNK_SIZE=500                    # Text chunk size
CHUNK_OVERLAP=50                  # Overlap between chunks
MAX_DOCUMENTS=1000                # Maximum documents to store

# Retrieval Settings
TOP_K_RESULTS=5                   # Number of results to retrieve
SIMILARITY_THRESHOLD=0.7          # Minimum similarity score
MAX_CONTEXT_LENGTH=4000          # Maximum context for LLM

# Rate Limiting
RATE_LIMIT_REQUESTS=100          # Requests per window
RATE_LIMIT_WINDOW=900000         # Window in milliseconds (15 min)

Supported File Types

Format	Extension	Description
PDF	`.pdf`	Portable Document Format
Word	`.doc`, `.docx`	Microsoft Word documents
Text	`.txt`	Plain text files
Markdown	`.md`	Markdown formatted text
CSV	`.csv`	Comma-separated values
HTML	`.html`	Web pages
JSON	`.json`	JavaScript Object Notation

API Endpoints

Chat API

POST /api/chat
Content-Type: application/json

{
  "message": "Your question here",
  "sessionId": "optional_session_id",
  "options": {
    "topK": 5,
    "threshold": 0.7
  }
}

Document Management

# Upload document
POST /api/documents/upload
Content-Type: multipart/form-data

# Get all documents
GET /api/documents/

# Delete document
DELETE /api/documents/{documentId}

# Search within document
POST /api/documents/{documentId}/search

Admin Panel

# System status
GET /api/admin/status

# Change LLM provider
POST /api/admin/llm/provider

WebSocket Events

Client to Server

// Send chat message
{
  "type": "chat",
  "data": {
    "message": "Your question",
    "sessionId": "session_id",
    "options": { "topK": 5 }
  }
}

// Stream chat (real-time response)
{
  "type": "stream_chat",
  "data": { /* same as above */ }
}

Server to Client

// Chat response
{
  "type": "chat_response",
  "data": {
    "message": "AI response",
    "sources": [...],
    "metadata": {...}
  }
}

// Streaming tokens
{
  "type": "stream_token",
  "token": "word "
}

🎨 Customization

UI Styling

Edit public/index.html for structure
Modify styles in the <style> section
Uses Tailwind CSS for responsive design

AI Providers

Add new LLM providers in src/services/llmService.js:

async yourProviderGenerate(prompt, options) {
  // Implement your provider logic
  return {
    text: response,
    provider: 'your-provider',
    model: 'your-model'
  };
}

Document Processors

Add new file format support in src/services/documentProcessor.js:

async extractYourFormat(buffer) {
  // Implement format-specific extraction
  return extractedText;
}

Troubleshooting

Common Issues

Files not uploading
- Check file size (max 50MB)
- Verify file format is supported
- Check network connection
AI not responding
- Verify API keys in .env
- Check provider status
- Try different LLM provider
Search not finding content
- Adjust similarity threshold
- Check document was processed
- Try different search terms
WebSocket connection issues
- Check browser console for errors
- Verify server is running
- Check firewall settings

Debug Mode

NODE_ENV=development npm run dev

Performance

Optimization Tips

Document Chunking: Adjust CHUNK_SIZE for your content
Search Results: Tune TOP_K_RESULTS and SIMILARITY_THRESHOLD
Memory Usage: Monitor with MAX_DOCUMENTS setting
API Costs: Use free providers for development

Scaling Considerations

Vector Database: Consider Pinecone, Weaviate for production
File Storage: Add cloud storage for document persistence
Caching: Implement Redis for session management
Load Balancing: Use multiple server instances

Contributing

Development Setup

# Fork the repository
git clone <your-fork>
cd rag-qa-chatbot

# Install dependencies
npm install

# Start development server
npm run dev

# Make your changes and test
# Submit a pull request

Areas for Contribution

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

OpenAI for GPT models
Anthropic for Claude
Google for Gemini
Cohere for language models
HuggingFace for free AI models
Tailwind CSS for styling
Express.js for backend framework

Built with ❤️ for the AI community

Need help? Open an issue or join our discussions!

FilesExpand file tree

Overview.md

Latest commit

History