arXade

AI-powered arXiv paper discovery using semantic search and intelligent summaries.

Demo

Watch the demo video here: arXade demo

Dataset

This project uses the arXiv CS Embeddings (Gemini-004-int8) dataset: https://www.kaggle.com/datasets/yxpx237/arxiv-cs-embeddings-gemini-004-int8

Features

Intelligent Search

• Vector embeddings for semantic paper discovery
• Search by concepts rather than keywords
• Relevance scoring based on paper content

AI Summaries

• Gemini AI generates research summaries
• Contextual recommendations for related work
• Deep research analysis for paper collections

Data Visualization

• Word clouds from paper abstracts
• Category distribution charts
• Timeline views of research trends
• Interactive filtering and exploration

Tech Stack

Frontend     │ Next.js, TypeScript, Tailwind CSS
Backend      │ FastAPI, MongoDB Atlas, Vector Search  
AI           │ Google Gemini API
Charts       │ Recharts, Visx, Tremor
Deployment   │ Docker, Google Cloud Run

Local Development

# Clone the repository
git clone https://github.com/yxpx/arXade.git
cd arXade

# Set up environment variables
cp .env.example .env
# Edit .env with your MongoDB URI, Gemini API key, etc.

# Start both frontend and backend
docker-compose up

This runs both services automatically:

• Frontend at http://localhost:3000
• Backend API at http://localhost:8000

Requirements

• Docker and Docker Compose
• MongoDB Atlas account
• Google Gemini API key

Project Structure

arXade/
├── frontend/          # Next.js application
├── backend/           # FastAPI server
├── data/              # Embedding generation scripts
└── docker-compose.yml # Development environment

License

MIT License - see LICENSE file for details.