Skip to content

the first decentralized effort to train and test upcoming computer-use and game-playing AI agents.

Notifications You must be signed in to change notification settings

viralmind-ai/viralmind-agents

Repository files navigation

ViralMind.ai 🧠

viralmind.ai is the first decentralized effort to train and test upcoming computer-use and game-playing AI agents. Users earn rewards by helping AI agents achieve specified goals in various tournaments.

🔗 Useful Links

🎯 Our Vision

We're building a decentralized platform where:

  • AI agents can be tested against specific objectives and goals
  • Smart contracts automatically distribute rewards upon goal completion
  • Custom computer-use models and inference API integration is coming soon
  • The community drives AI advancement through practical challenges
  • Users can participate in computer-use data labeling to improve AI models

🚀 Development Quickstart

Prerequisites

  • Docker and Docker Compose
  • Node.js and npm
  • OpenAI API key
  • Solana CLI tools and wallet

Environment Setup

  1. Create a .env file in the root directory:
cp .env.example .env

Then edit the .env file and add your API keys.

  1. Set up Solana keypair:

Install the Solana CLI and the Anchor CLI.

# Create a directory for solana secrets
mkdir -p ./backend/secrets/solana
solana-keygen new -o ./backend/secrets/solana-keypair.json

Start Development Environment

# Make sure solana is on devnet
solana config set --url https://api.devnet.solana.com/

# Build and Deploy Contract
cd backend/jailbreak-pool 
anchor build
# To deploy make sure that your previously created keypair has a SOL balance.
anchor deploy

# Start all services
docker compose up --build

# Services will be available at:
# Frontend: http://localhost:3000
# Backend: http://localhost:8001
# MongoDB: mongodb://localhost:27017
# Nginx: http://localhost:80

🤖 How It Works

1. Choose a Tournament

  • Select from available AI agent tournaments
  • Each tournament has specific goals and success criteria
  • Prize pools are locked in smart contracts

2. Help AI Agents Achieve Goals

  • Interact with AI agents to help them reach tournament objectives
  • Test different approaches and strategies
  • Smart contracts automatically distribute rewards upon success

3. Upcoming Features

  • Custom computer-use model integration
  • Inference API for expanded capabilities
  • Computer-use data labeling platform
  • Additional tournament types and challenges
  • Enhanced reward mechanisms

📊 Tournament System

  • Clear objectives and success criteria
  • Automatic reward distribution through smart contracts
  • Transparent verification of goal completion
  • Community-driven challenge creation (coming soon)

📞 Contact & Support

For technical support or partnership inquiries, reach out at [email protected]

About

the first decentralized effort to train and test upcoming computer-use and game-playing AI agents.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published