Skip to content
BranchesTags

Repository files navigation

Research Assistant AI

A full-stack research assistant tool for computer science grad students that uses real-time AI-powered search to find and visualize research papers related to your thesis topic using an interactive knowledge graph.

Features

  • 🔍 Real-time Paper Search: Uses Tavily AI to find actual research papers from academic sources
  • 🗺️ Enhanced Discovery Mode: Tavily Map traverses academic websites to discover interconnected papers through citation networks, author collaborations, and research clusters
  • 🧠 LLM-Powered Analysis: Gemini 1.5 Pro analyzes papers to identify semantic relationships and connections
  • 🎯 Smart Relevance Filtering: Papers are filtered and scored based on relevance to your topic
  • 📊 Interactive Knowledge Graph: D3.js-powered visualization with force-directed layout
  • 🔗 Intelligent Connections: Papers connect based on LLM-analyzed relationships (builds upon, validates, contradicts, etc.)
  • 📈 Live Statistics: Real-time stats including paper count, credibility scores, and top venues
  • 🎮 Interactive Controls: Drag nodes, zoom, pan, pause/resume simulation
  • 🎨 Beautiful Design: Modern gradient UI with glassmorphism effects
  • 🔄 Fallback System: Gracefully falls back to curated examples if API is unavailable

Color-coded Credibility:

  • 🔴 Red: High credibility (8-10) - Top venues like Nature, Science, IEEE
  • 🟡 Yellow: Medium credibility (5-7) - Good conferences and journals
  • 🟢 Green: Lower credibility (1-4) - Workshops, preprints
  • 🔵 Blue: Your thesis topic (center node)

Tech Stack

Frontend

  • Next.js 15 with App Router
  • TypeScript for type safety
  • Tailwind CSS for styling
  • D3.js for interactive graph visualization

Backend

  • FastAPI for high-performance API
  • Tavily AI for intelligent research paper search
  • Python 3.8+ with async/await support
  • Pydantic for data validation

Getting Started

Prerequisites

  • Node.js 18+ and npm
  • Python 3.8+ and pip
  • Tavily API Key (free tier available)
  • Google API Key (optional, for Gemini LLM analysis)

Installation

  1. Clone and setup the project:
git clone <repository-url>
cd research-assistant-ai
  1. Install frontend dependencies:
npm install
  1. Install backend dependencies:
pip install -r requirements.txt
  1. Configure API Keys:

Set your API keys as environment variables:

export TAVILY_API_KEY="your-actual-tavily-api-key"
export GOOGLE_API_KEY="your-google-api-key"  # Optional for Gemini LLM analysis

Or create a .env file in the project root:

TAVILY_API_KEY=your-actual-tavily-api-key
GOOGLE_API_KEY=your-google-api-key

Getting API Keys:

  • Tavily API Key: Sign up at Tavily for research paper search
  • Google API Key: Get one at Google AI Studio for Gemini LLM analysis

Note: The Google API key is optional. If not provided, the system will fall back to keyword-based similarity connections.

Running the Application

Option 1: Run both servers manually

Terminal 1 - Backend API:

cd backend
python start.py

The API will start at http://localhost:8000

Terminal 2 - Frontend:

npm run dev

The frontend will start at http://localhost:3000

Option 2: Development with API docs

  1. Start the backend first to see API documentation:
cd backend && python start.py

  1. Visit http://localhost:8000/docs to explore the API

  2. Start the frontend:

npm run dev
  1. Visit http://localhost:3000 to use the application

Usage

  1. Enter your research topic (e.g., "Transformer architectures for NLP")
  2. Toggle Enhanced Discovery Mode (🗺️) for comprehensive paper discovery through academic web mapping
  3. Click "Generate Knowledge Graph" to search for real papers
  4. Explore the visualization:
    • Hover over nodes to see paper details
    • Drag nodes to reposition them
    • Use controls to pause/resume, reset layout, or zoom to fit
  5. View statistics about your search results and mapping discoveries
  6. Click paper nodes to potentially visit their source URLs

Enhanced Discovery Mode

When enabled, the system uses Tavily Map to intelligently traverse academic websites and discover interconnected papers through:

  • 🔗 Citation Networks: Follows reference chains and related work sections
  • 👥 Author Collaborations: Discovers papers by the same research groups
  • 📚 Conference Proceedings: Explores papers from the same venues and conferences
  • 🌐 Academic Databases: Maps through ArXiv, Google Scholar, IEEE, ACM, and more
  • 🧠 Semantic Clustering: Groups papers by research themes and methodologies

This mode typically discovers 2-3x more relevant papers than standard search alone!

Paper Connection Intelligence

The enhanced system now generates intelligent connections between papers using:

  • 📊 Similarity Analysis: Compares keywords, authors, venues, and publication years
  • 🔗 Citation Networks: Links papers through Tavily Map traversal
  • 👥 Author Collaborations: Connects papers with shared authors
  • 🏛️ Venue Clustering: Groups papers from the same conferences/journals
  • 📈 Relationship Scoring: Quantifies connection strength (0-1 scale)

Minimum 15 Papers Guaranteed: The system ensures you always get at least 15 relevant papers through fallback mechanisms and enhanced search strategies.

API Endpoints

POST /search-papers

Search for research papers using Tavily AI.

Request:

{
  "topic": "machine learning interpretability",
  "max_results": 10
}

Response:

{
  "papers": [...],
  "total_found": 8,
  "query": "machine learning interpretability"
}

POST /discover-papers-with-mapping

🆕 Enhanced Discovery Mode - Combines Tavily Search, Map, and LLM analysis for comprehensive paper discovery through academic website traversal.

Request:

{
  "topic": "transformer architectures",
  "max_results": 10
}

Response:

{
  "papers": [...],
  "mapped_papers": [...],
  "total_found": 25,
  "mapping_stats": {
    "domains_explored": ["arxiv.org", "scholar.google.com"],
    "total_urls_discovered": 150,
    "papers_extracted": 12
  },
  "query": "transformer architectures",
  "graph_data": {
    "nodes": [...],
    "edges": [...]
  }
}

POST /search-papers-with-analysis

Search for research papers with LLM-powered relationship analysis.

Request:

{
  "topic": "transformer architectures",
  "max_results": 10
}

Response:

{
  "papers": [...],
  "total_found": 8,
  "query": "transformer architectures",
  "graph_data": {
    "nodes": [...],
    "edges": [
      {
        "source": "1",
        "target": "2",
        "relationship_type": "Builds_Upon",
        "strength": 4,
        "description": "Paper 2 extends the transformer architecture introduced in Paper 1"
      }
    ],
    "entity_clusters": [...]
  }
}

GET /search-papers-simple?topic=your-topic

Simple GET endpoint for searching papers.

GET /health

Check API health and configuration status.

GET /docs

Interactive API documentation (Swagger UI).

How It Works

Backend Intelligence

  • Tavily AI Integration: Searches academic domains for relevant papers
  • Smart Data Extraction: Automatically extracts authors, venues, years from content
  • Credibility Scoring: Calculates scores based on venue reputation and content quality
  • Citation Estimation: Estimates citation counts based on age and credibility

Frontend Visualization

  • Relevance Calculation: Filters papers based on keyword and abstract matching
  • Network Generation: Creates connections between papers with similar keywords
  • Force Simulation: Uses D3.js physics for natural node positioning
  • Interactive Features: Real-time manipulation and exploration

Customization

Adding More Academic Sources

Edit academic_domains in backend/main.py:

academic_domains = [
    "arxiv.org", "semanticscholar.org", "ieee.org",
    "your-institution.edu"  # Add your sources
]

Adjusting Relevance Thresholds

Modify the filtering thresholds in app/page.tsx:

const relevantPapers = paperNodes.filter((paper) => paper.relevance! > 0.2);

Customizing Credibility Scoring

Update the calculate_credibility() function in backend/main.py to match your field's venues.

Example Topics

Try these research topics to see the system in action:

  • "Transformer architectures for natural language processing"
  • "Graph neural networks for drug discovery"
  • "Explainable AI in healthcare"
  • "Computer vision for autonomous vehicles"
  • "Reinforcement learning for robotics"

Troubleshooting

API Connection Issues

  • Ensure the backend is running on port 8000
  • Check that CORS is properly configured
  • Verify your Tavily API key is valid

No Papers Found

  • Try broader or more specific search terms
  • Check if Tavily service is available
  • The system will fall back to curated examples

Graph Not Rendering

  • Ensure D3.js is properly loaded
  • Check browser console for JavaScript errors
  • Try refreshing the page

Future Enhancements

  • 🔐 User Authentication: Save and share research graphs
  • 📥 Export Features: Save graphs as images or datasets
  • 🔄 Citation Networks: Show how papers cite each other
  • 🤖 AI Recommendations: Suggest related papers and research directions
  • 👥 Collaboration: Share graphs with research teams
  • 📱 Mobile Support: Responsive design for tablets and phones

Contributing

Feel free to submit issues and enhancement requests!

License

This project is for educational and research purposes.

About

No description, website, or topics provided.

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages