Snorlax

AI-powered GitHub issue triage assistant using Claude and semantic code search.

What It Does

Automatically categorizes GitHub issues, detects duplicates, finds related code, and generates response suggestions.

Setup

Prerequisites:

• Docker Desktop
• Python 3.9+
• Node.js 18+
• Anthropic API key

Install:

git clone <your-repo-url>
cd snorlax
make setup
make dev

Configure:

1. Open http://localhost:3000
2. Add Anthropic API key in Settings
3. Add repository and import issues
4. Start triaging

Commands

make setup      # Initial setup
make dev        # Start all services
make db-reset   # Reset database
make clean      # Stop all services

Tech Stack

• Frontend: Next.js 15, React 19, TypeScript
• Backend: FastAPI, Python
• Database: PostgreSQL with pgvector
• AI: Claude Sonnet 4.5
• Embeddings: CocoIndex (automatic real-time sync)

How It Works

Automatic Issue Sync:

1. User adds repository → Snorlax clones and indexes code
2. CocoIndex starts watching the database for changes
3. When issues are fetched from GitHub → PostgreSQL triggers fire
4. CocoIndex automatically generates embeddings in real-time
5. UI always shows fresh data (no manual sync needed)

Benefits:

• 🔄 Real-time: Embeddings update within seconds
• 🤖 Automatic: No manual "Sync" button needed
• ⚡ Efficient: Only changed issues are reprocessed
• 🎯 Consistent: Uses same embedding model as code

Configuration

Backend (backend/.env):

APP_DATABASE_URL=postgresql://snorlax:snorlax_password@localhost:5432/snorlax

Frontend (frontend/.env.local):

NEXT_PUBLIC_API_URL=http://localhost:8000

Troubleshooting

Stop services:

lsof -ti:8000 | xargs kill -9  # Backend
lsof -ti:3000 | xargs kill -9  # Frontend
make db-down                    # Database

Reset database:

make db-reset