MC-Bench

MC-Bench is a benchmarking platform for evaluating and comparing model performance.

Project Structure

• mc-bench-backend/: Python backend services including API and workers
• mc-bench-frontend/: React+TypeScript frontend application

Prerequisites

• Python 3.12.7
• Bun
• Docker and Docker Compose
• Access to Github OAuth Client ID and Secret (see Setting Up Oauth Pre-Reqs.

Quick Start

1. Clone both repositories:

# Clone backend
git clone https://github.com/mc-bench/mc-bench-backend.git
cd mc-bench-backend

# Clone frontend
git clone https://github.com/mc-bench/mc-bench-frontend.git

2. Set up the backend:

cd mc-bench-backend
# Create and activate a Python virtual environment (see backend README for options)
cp .env.template .env
pip install -e .[dev,api,worker]

Be sure to copy over the Github Client ID and Client Secret from the pre-requisites into your new .env file

3. Set up the frontend:

cd mc-bench-frontend
cp .env.template .env
bun install

Be sure to copy over the Github Client ID from the pre-requisites into your new .env file

4. Start the services:

cd mc-bench-backend
docker-compose up -d

This will start:

• PostgreSQL database
• Redis for message queuing
• MinIO for object storage
• API services
• Worker services

5. Start the frontend development server:

cd mc-bench-frontend
source .env-local
bun run dev

The application will be available at:

• Frontend: http://localhost:5173
• Main API: http://localhost:8000
• Admin API: http://localhost:8001