RefNet - Research Paper Search & Citation Network Visualization

🏆 2nd Place Overall Winner at HackGT 12 🏆

Team Members:

• Nikhilesh
• Dhruva
• Shreyas
• Krishna

RefNet is a comprehensive tool for searching research papers and visualizing their citation networks. It combines a powerful search interface with an interactive graph visualization and AI-powered analysis to help researchers explore academic literature and understand citation relationships.

🤖 AI-Powered Research Analysis

• Custom Mastra Backend: Express.js + OpenAI GPT-4o for intelligent research analysis
• Smart Context: AI understands your selected papers and graph relationships
• Research Insights: Compare papers, identify patterns, and discover research gaps
• Real-time Analysis: Ask questions about your selected papers and get instant insights
• Review Paper Generation: AI-powered literature review creation with PDF export

Features

🔍 Advanced Search

• Search research papers by title, authors, topics, and keywords
• Filter results by publication date, citation count, and relevance
• Sort by most cited, relevance score, or publication date
• Paginated results with customizable page sizes
• Multiselect functionality - Select multiple papers to build combined citation networks

📊 Interactive Graph Visualization

• Build citation networks from any research paper or multiple papers
• Interactive D3.js-powered graph with zoom, pan, and drag functionality
• Node selection and highlighting
• Timeline-based color coding
• Light grey/white lines for clean, academic appearance
• Export selected papers and graph data
• Graph Rebuild Fallback: Automatic restoration of previous graph on rebuild failure

📄 AI-Powered Review Paper Generation

• Generate comprehensive literature reviews from selected papers
• AI-generated sections: Abstract, Introduction, Fundamentals, Types & Categories, State-of-the-Art
• Intelligent title generation based on paper analysis
• PDF export with academic formatting
• Text file fallback for compatibility

🚀 Modern Web Interface

• Responsive design that works on desktop and mobile
• Fast, modern React frontend
• Real-time API integration
• Intuitive user experience
• PWA Support: Manifest.json for progressive web app capabilities

Quick Start

Prerequisites

• Python 3.8+ (for search API)
• Node.js 18+ (for Mastra AI backend)
• npm or yarn
• OpenAI API key

Setup

1. Set your OpenAI API key:

   export OPENAI_API_KEY='your-openai-api-key-here'

2. Start all services:

   ./start_cedar_mastra.sh

   This will start:

   • Flask search API on http://localhost:8000
   • Mastra AI backend on http://localhost:4111
   • React frontend on http://localhost:3000

Manual Setup (Alternative)

If you prefer to start services manually:

1. Start Flask search API:

   python app.py

2. Start Mastra AI backend:

   cd mastra-backend
   npm start

3. Start React frontend:

   cd refnet/frontend
   npm start

Docker Setup (Production)

# Set your OpenAI API key
export OPENAI_API_KEY='your-openai-api-key-here'

# Start with Docker Compose
docker-compose up

Production Build

To build the frontend for production:

cd refnet/frontend
npm run build

The built files will be in refnet/frontend/build/ and will be automatically served by the Flask backend.

Usage

1. Search Papers: Use the landing page to search for research papers by entering keywords, author names, or topics.

2. Select Papers:

   • Use checkboxes to select multiple papers from search results
   • Click "Build Graph" to create a combined citation network from all selected papers
   • Or click "View Graph" on individual papers for single-paper networks

3. Explore Network:

   • Click and drag nodes to rearrange the graph
   • Click nodes to select/deselect them
   • Use the controls to adjust graph parameters (iterations, limits)
   • Export selected papers as JSON

4. AI Analysis:

   • Click the chat button to open the AI research assistant
   • Ask questions about your selected papers
   • Get intelligent insights, comparisons, and research recommendations
   • The AI understands your paper context and graph relationships

5. Generate Review Papers:

   • Select papers from your graph
   • Click "Generate Survey Paper" to create AI-powered literature review
   • Export as PDF or text file

6. Navigate: Use the back button to return to search results or start a new search.

API Endpoints

Search

• GET /api/search?q=query&page=1&per_page=25&sort=cited_by_count - Search papers

Papers

• GET /api/paper/{paper_id} - Get paper details
• GET /api/paper/{paper_id}/citations - Get paper citations
• GET /api/paper/{paper_id}/references - Get paper references

Graph

• GET /api/graph/{paper_id}?iterations=3&cited_limit=5&ref_limit=5 - Build citation graph from single paper
• POST /api/graph/multiple - Build graph from multiple papers (multiselect)
• GET /api/graph/data - Get current graph data
• POST /api/graph/clear - Clear current graph

AI Backend (Mastra)

• POST /chat - AI chat and research analysis
• GET /health - Health check

Project Structure

RefNet/
├── app.py                    # Flask search API entry point
├── config.py                # Configuration settings
├── requirements.txt         # Python dependencies
├── mastra-backend/          # Mastra AI backend (Node.js + Express)
│   ├── server.js           # AI agent server
│   ├── package.json        # Backend dependencies
│   └── README.md           # Backend documentation
├── refnet/
│   ├── api/                # API route blueprints
│   │   ├── chat_routes.py  # Chat API routes
│   │   ├── graph_routes.py # Graph API routes
│   │   ├── paper_routes.py # Paper API routes
│   │   └── search_routes.py # Search API routes
│   ├── models/             # Data models
│   │   ├── graph.py        # Graph data models
│   │   └── paper.py        # Paper data models
│   ├── services/           # Business logic services
│   │   ├── graph_service.py # Graph processing
│   │   └── openalex_service.py # OpenAlex API integration
│   ├── utils/              # Utility functions
│   │   ├── rate_limiter.py # API rate limiting
│   │   └── validators.py   # Data validation
│   ├── tests/              # Test files
│   └── frontend/           # React frontend
│       ├── src/
│       │   ├── components/
│       │   │   ├── GraphViewerClean.js    # Main graph visualization
│       │   │   ├── FloatingCedarChat.js   # AI chat interface
│       │   │   ├── LandingPage.js         # Search interface
│       │   │   └── ChatTracker.js         # Chat management
│       │   ├── services/
│       │   │   ├── api.js                 # API client
│       │   │   └── cedarAgent.js          # AI agent service
│       │   ├── cedar/                     # Cedar OS configuration
│       │   └── ...
│       ├── public/
│       │   ├── index.html
│       │   ├── favicon.ico
│       │   ├── logo.svg
│       │   ├── logo192.png
│       │   └── manifest.json              # PWA manifest
│       └── package.json
├── docker-compose.yml       # Docker Compose configuration
├── Dockerfile.flask         # Flask API Docker image
├── Dockerfile.mastra        # Mastra AI Docker image
├── start_cedar_mastra.sh    # Development startup script
└── README.md

Technologies Used

Backend

• Flask: Search API framework
• Express.js: Mastra AI backend framework
• OpenAI GPT-4o: Research analysis AI
• OpenAlex API: Research paper data source
• NetworkX: Graph analysis and processing
• Flask-CORS: Cross-origin resource sharing

Frontend

• React 18: UI framework
• React Router: Client-side routing
• D3.js: Graph visualization
• Axios: HTTP client
• Tailwind CSS: Styling
• PWA: Progressive Web App capabilities

Deployment

• Docker: Containerization
• AWS EC2: Cloud hosting
• Docker Compose: Multi-service orchestration
• Production URLs:
  • Frontend: https://refnet.wiki
  • API: https://api.refnet.wiki
  • Mastra AI: https://api.refnet.wiki/mastra

Review Paper Generation

The system generates comprehensive literature reviews using AI-powered content creation:

1. Paper Selection: Users select papers from the citation graph
2. AI Analysis: Each paper gets an AI-generated summary via GPT-4o
3. Content Generation: Creates 5 sections (Abstract, Introduction, Fundamentals, Types & Categories, State-of-the-Art)
4. PDF Export: Uses browser print functionality for professional PDF output
5. Fallback: Text file export if PDF generation fails

Tech Stack for Review Generation:

• AI Backend: Mastra (Express.js + OpenAI GPT-4o)
• PDF Generation: Browser native print functionality
• Content Processing: Custom algorithms for domain detection and title generation
• Formatting: HTML-to-PDF with academic styling

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

License

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

Acknowledgments

• OpenAlex for providing research paper data
• D3.js for graph visualization capabilities
• OpenAI for AI-powered research analysis
• The academic research community for inspiration and use cases
• HackGT 12 organizers and judges for the recognition