ComputeChain SDK

Build decentralized GPU marketplaces in 3 commands.

npm install @compute-chain/sdk
node sdk/example.js
# Your GPU marketplace is live!

⚡ Quick Start

# Install SDK
npm install @compute-chain/sdk

# Run example
cd node_modules/@compute-chain/sdk
node example.js

# Output:
# ✅ Job 1 submitted (CPU)
# ✅ Job 2 submitted (GPU)
# ✅ Provider registered
# 🎉 Your marketplace is running!

Or use Docker:

docker run -d -p 9944:9944 skills003/substrate-ipfs-node:v1.1.0 --dev

🎯 What You Get

• ✅ Full Substrate blockchain - Production-ready consensus
• ✅ Docker execution engine - Run any containerized workload
• ✅ Payment & escrow - Automatic token escrow and release
• ✅ Fraud detection - Challenge period for verification
• ✅ Provider network - Register GPU/CPU providers
• ✅ React frontend - User dashboard included
• ✅ CLI tools - Command-line interface

📖 API Usage

Submit a Job

const ComputeChain = require('@compute-chain/sdk');
const chain = new ComputeChain('ws://localhost:9944');

const jobId = await chain.submitJob({
  image: 'pytorch/pytorch:2.0.1-gpu',
  command: 'python train.py',
  ram: 16,      // GB
  cpu: 8,       // cores
  gpus: 2,      // count
  duration: 120 // minutes
});

console.log(`Job ${jobId} submitted`);

Register as Provider

await chain.registerProvider({
  name: 'my-gpu-farm',
  ram: 128,
  cpu: 64,
  gpus: 8,
  endpointUrl: 'http://your-server:7682'
});

Check Status

const job = await chain.getJob(jobId);
console.log(`Status: ${job.status}`);
console.log(`Cost: ${job.escrowed} tokens`);

List Providers

const providers = await chain.getProviders();
providers.forEach(p => {
  console.log(`${p.peerId}: ${p.resources.gpus} GPUs available`);
});

🛠️ CLI Commands

# Submit job
node compute-chain.js submit ubuntu:22.04 "echo hello" 0

# Register provider
node compute-chain.js register my-provider 1

# Check status
node compute-chain.js status 0

# List providers
node compute-chain.js providers

💰 Pricing

Default rates:

• RAM: 10 tokens/GB/hour
• CPU: 5 tokens/core/hour
• GPU: 50 tokens/GPU/hour

Example:

• 16GB RAM + 4 CPU + 1 GPU for 2 hours
• Cost: (16×10 + 4×5 + 1×50) × 2 = 500 tokens

Customize pricing in your chain config!

🏗️ Architecture

┌─────────────┐
│   Client    │  Submit job + escrow tokens
└──────┬──────┘
       │
┌──────▼──────────┐
│   Blockchain    │  Substrate chain with payment logic
└──────┬──────────┘
       │
┌──────▼──────────┐
│    Provider     │  Claim job, execute container
└──────┬──────────┘
       │
┌──────▼──────────┐
│ Docker + GPU    │  Return results + proof
└─────────────────┘

📂 Templates

Pre-configured chains for specific use cases:

AI Training

# Optimized for ML workloads
# - GPU: 30 tokens/hr (cheap)
# - Max duration: 7 days
# - Max RAM: 256GB

Render Farm

# For 3D rendering and video
# - GPU: 80 tokens/hr (premium)
# - Blender, Maya support

Scientific Compute

# Long-running simulations
# - Max duration: 14 days
# - High CPU allocation

See /templates/ for configurations.

🚀 Full Deployment

Deploy complete marketplace with all services:

git clone https://github.com/your-org/compute-chain-sdk
cd compute-chain-sdk
./deploy.sh

# Starts:
# - Blockchain (port 9944)
# - Docker service (port 7682)
# - Provider service
# - Frontend (port 3000)

Production setup:

# 1. Get Ubuntu server with GPU
# 2. Install Docker + NVIDIA runtime
# 3. Clone and deploy
docker-compose up -d

📊 Cost Comparison

| Workload | Duration | Your Chain | AWS p3.2xlarge | Savings | |----------|----------|------------|----------------|---------| | GPT Fine-tune | 8hr | 3,232 tokens (~$32) | ~$240 | 87% | | Stable Diffusion | 1hr | 202 tokens (~$2) | ~$8 | 75% | | Video Render | 4hr | 1,664 tokens (~$17) | ~$32 | 48% |

At $0.01/token vs AWS on-demand pricing

🔒 Security Features

• ✅ Payment escrow - Funds locked until job verified
• ✅ Challenge period - 100 blocks for disputes
• ✅ Fraud detection - Validators re-execute suspicious jobs
• ✅ Reputation system - Track provider quality
• ✅ Slashing - Penalties for malicious providers

🧪 Testing

Tested and verified:

npm test

# Output:
# ✅ Test 1: Connection (Substrate IPFS Compute Network)
# ✅ Test 2: Submit CPU job (Job 15)
# ✅ Test 3: Submit GPU job (Job 16)
# ✅ Test 4: Get job status
# ✅ Test 5: List providers
# ✅ Test 6: Register provider
# ✅ Test 7: Verify registration
# 🎉 All tests passed!

See /SDK_TEST_REPORT.md for detailed test results.

🎓 Examples

AI Training:

await chain.submitJob({
  image: 'huggingface/transformers-pytorch-gpu',
  command: 'python train_gpt.py --epochs 100',
  ram: 64, cpu: 16, gpus: 4, duration: 480
});

Stable Diffusion:

await chain.submitJob({
  image: 'stability-ai/stable-diffusion',
  command: 'python generate.py --prompt "blockchain GPU"',
  ram: 16, cpu: 4, gpus: 1, duration: 30
});

Batch Processing:

for (let i = 0; i < 100; i++) {
  await chain.submitJob({
    image: 'my-processor',
    command: `process.py --input data-${i}.csv`,
    ram: 4, cpu: 2
  });
}

More examples in /sdk/ai-training-testnet/test-ai-workloads.js

📚 Documentation

• API Reference: See code comments in compute-chain.js
• Templates: Configuration examples in /templates/
• Deployment: Instructions in deploy.sh
• Architecture: Details in README.md

🤝 Contributing

Built with progressive enhancement:

1. Make it work ✅
2. Ship it ✅
3. Enhance later ← We are here

Contributions welcome! Please:

1. Fork the repo
2. Create a feature branch
3. Submit a pull request

🐛 Troubleshooting

Blockchain won't start:

rm -rf /tmp/substrate-*
./deploy.sh

Jobs not executing:

# Check provider is registered
node compute-chain.js providers

# Check Docker service
curl http://localhost:7682/health

Out of tokens:

# Use pre-funded dev accounts
account: '//Alice'   # Has 1M tokens

📜 License

MIT - Build whatever you want!

🔗 Links

• npm: @compute-chain/sdk
• Docker: skills003/substrate-ipfs-node
• Live Demo: [Coming soon]

⭐ Star History

If this helps you build your GPU marketplace, please star the repo!

Built with: Substrate · Docker · Polkadot.js · React · NVIDIA CUDA

Tested on: Ubuntu 22.04 · NVIDIA GPUs · Docker 24+ · Node.js 18+