🍋 Yuzu - Fresh Research Discovery

Squeeze knowledge from every paper. Discover academic research through an intuitive swipe interface with AI-powered summaries.

Features

• 🔍 Smart Search: Find papers from ArXiv by keyword
• ✨ AI Summaries: Three levels of depth powered by Gemini
• 💾 Easy Export: One-click BibTeX export for citations
• 🎉 Delightful UX: Confetti animations and toast notifications
• ⌨️ Keyboard Controls: Fast navigation with arrow keys
• 📱 Mobile Friendly: Touch gestures on mobile devices
• 🎨 Beautiful Design: Warm citrus-inspired interface

Quick Start

Prerequisites

• Docker and Docker Compose
• Google API key (for Gemini AI)

Development

# Clone repository
git clone https://github.com/yourusername/yuzu.git
cd yuzu

# Add your Google API key
echo "GOOGLE_API_KEY=your-key-here" > .env

# Start services
docker-compose up --build

# Visit http://localhost:3000

Production Deployment

# Set your Google API key
export GOOGLE_API_KEY=your-key-here

# Build and deploy production images
docker-compose -f docker-compose.prod.yml build
docker-compose -f docker-compose.prod.yml up -d

Tech Stack

• Frontend: Next.js 14, TypeScript, Tailwind CSS, Framer Motion
• Backend: FastAPI, Python 3.11
• APIs: ArXiv, Google Gemini (via OpenAI API)
• Deployment: Docker, Docker Compose

Usage

Search Papers

1. Enter a research topic on the landing page
2. Wait for Yuzu to fetch relevant papers from ArXiv

Navigate Papers

• ← Left Arrow or Swipe Left: Pass on current paper
• → Right Arrow or Swipe Right: See more details / Next paper
• Space Bar: Superlike and save paper to favorites

Detail Levels

Papers have three AI-generated summary levels:

1. Level 1: Quick overview (one paragraph)
2. Level 2: Detailed summary (multiple sections)
3. Level 3: Comprehensive analysis (full breakdown)

Press → or swipe right to dive deeper into each level.

Save & Export

• Press Space to superlike papers
• View saved papers in the sidebar
• Click Download BibTeX to export citations

Keyboard Shortcuts

Key	Action
←	Pass on current paper
→	See more details / Next paper
Space	Superlike and save paper

Mobile Support

On touch devices:

• Swipe Left: Pass paper
• Swipe Right: More details
• Tap Star Button: Save paper

Project Structure

yuzu/
├── backend/
│   ├── app/
│   │   ├── main.py          # FastAPI app
│   │   ├── arxiv_client.py  # ArXiv API client
│   │   └── openai_client.py # Gemini integration
│   ├── Dockerfile
│   └── requirements.txt
├── frontend/
│   ├── app/
│   │   ├── page.tsx         # Main app
│   │   └── layout.tsx       # Root layout
│   ├── components/
│   │   ├── PaperCard.tsx    # Paper display
│   │   ├── SwipeInterface.tsx # Swipe logic
│   │   ├── Toast.tsx        # Notifications
│   │   └── ...
│   ├── lib/
│   │   ├── confetti.ts      # Celebration animations
│   │   ├── api.ts           # Backend API client
│   │   └── types.ts         # TypeScript types
│   ├── Dockerfile
│   └── Dockerfile.prod      # Production build
├── docker-compose.yml       # Development setup
└── docker-compose.prod.yml  # Production setup

Environment Variables

Backend

• GOOGLE_API_KEY: Your Google API key for Gemini
• ENVIRONMENT: Set to production for production builds

Frontend

• NEXT_PUBLIC_API_URL: Backend API URL (default: https://api-yuzu.epicrunze.com)

Development

Running Locally

# Backend (requires Python 3.11+)
cd backend
pip install -r requirements.txt
uvicorn app.main:app --reload --port 8282

# Frontend (requires Node.js 18+)
cd frontend
npm install
npm run dev

Installing Dependencies

The frontend requires additional packages. Run:

cd frontend
npm install

This will install:

• canvas-confetti - Celebration animations
• react-swipeable - Mobile touch gestures

License

MIT

Acknowledgments

Built with 🍋 for Duke AI Hackathon

• ArXiv for open research access
• Google Gemini for AI summaries
• Next.js and React communities

---

Made with 🍋 by Yuzu Team