Juryline

The AI-Native Hackathon Judging Platform

Juryline is a modern, production-grade platform designed to fix the broken hackathon judging process. It replaces spreadsheets and manual entry with a streamlined, AI-orchestrated workflow that handles everything from submission ingestion to final leaderboards.

Live Demo: juryline.vercel.app

Demo Video: Demo Video Link

---

Key Features

• Dynamic Form Builder: Create custom submission forms (Text, URL, File Upload, etc.) instantly.
• AI-Powered Orchestration: Powered by Archestra, a swarm of 5 specialized agents:
  • Ingest Agent: Validates submission metadata for spam/quality.
  • Assignment Agent: Intelligently balances judging load across available reviewers.
  • Progress Agent: Tracks judging velocity and identifies bottlenecks.
  • Aggregation Agent: Calculates final weighted scores for unbiased rankings.
  • Feedback Agent: Synthesizes scores into constructive feedback for hackers.
• Magic Link Auth: Passwordless, secure login for judges (via Supabase Auth).
• R2 Storage Integration: Fast, scalable asset storage for submission videos and banners.
• Real-time Leaderboard: Instant, auto-calculating results dashboard.

---

Architecture

Built for scale and reliability:

• Frontend: Next.js 14, TypeScript, Chakra UI, Framer Motion
• Backend: Python 3.12, FastAPI, Pydantic
• Database: PostgreSQL (Supabase) with Row Level Security (RLS)
• AI Runtime: Self-hosted Archestra Agent Swarm (JSON-RPC 2.0 A2A Protocol)
• Storage: Cloudflare R2 (S3-compatible)

---

Getting Started

Prerequisites

• Docker & Docker Compose
• Supabase Project
• Cloudflare R2 Bucket (Optional, for uploads)
• Archestra API Key (Optional, for AI features)

1. Clone & Setup

git clone https://github.com/tusharkhatri/juryline.git
cd juryline

2. Environment Configuration

Create a .env file in the backend/ directory:

# Supabase
SUPABASE_URL=https://your-project.supabase.co
SUPABASE_SERVICE_KEY=your-service-role-key  # For backend admin access
SUPABASE_JWT_SECRET=your-jwt-secret          # For token verification

# Cloudflare R2 (File Storage)
R2_ACCOUNT_ID=your-account-id
R2_ACCESS_KEY_ID=your-access-key
R2_SECRET_ACCESS_KEY=your-secret-key
R2_BUCKET_NAME=your-bucket-name
R2_PUBLIC_URL=https://your-r2-public-domain.com

# Archestra AI (Self-Hosted)
ARCHESTRA_API_KEY=your-archestra-key
ARCHESTRA_BASE_URL=https://baserul.com
# Agent Prompt IDs (if running own swarm)
ARCHESTRA_INGEST_PROMPT_ID=...
ARCHESTRA_ASSIGN_PROMPT_ID=...

3. Run with Docker

Launch the full stack (Frontend + Backend):

docker compose up -d --build

• Frontend: http://localhost:4000
• Backend: http://localhost:8888
• API Docs: http://localhost:8888/docs

4. Database Seeding

Populate the database with a production-grade demo dataset (Events, Submissions, Videos, Reviews):

db/seed.py

This script will:

1. Clear existing data (safely).
2. Create an "Organizer" and multiple "Judges".
3. Create a live event ("Global AI Hackathon 2026").
4. Submit 20+ realistic projects with video links.
5. Simulate judge activity to populate the leaderboard.

---

Deployment

The project is containerized and ready for deployment.

• Backend: Deploy the backend/Dockerfile to any container runtime (DigitalOcean App Platform, Fly.io, Railway, AWS ECS).
• Frontend: Deploy frontend/ to Vercel or Netlify.
• Archestra: The AI agents run on a dedicated DigitalOcean Droplet for high-performance inference.

---

License

MIT License. Built with ❤️ by Tushar Khatri.