MeshVision

MeshVision

Decentralized Visual Compute Network

---

🌌 Introduction

MeshVision is a decentralized infrastructure for verifiable AI inference. It enables developers to offload complex computer vision tasks—such as object detection, facial recognition (privacy-preserving), and anomaly detection—to a distributed network of TEE-enabled compute nodes.

By leveraging Zero-Knowledge Proofs (ZKPs) and Optimistic Verification, MeshVision ensures that every pixel processed is cryptographically verifiable, bridging the gap between high-performance AI and trustless Web3 systems.

🏗 Architecture

The MeshVision monorepo is organized into a modular microservices architecture:

Package	Path	Description	Tech Stack
dApp	packages/dapp	The user interface for submitting tasks and exploring the network.	React, Vite, Wagmi, RainbowKit
Node	packages/node	The Python-based compute engine. Runs ONNX models and generates ZK proofs.	Python 3.10, FastAPI, ONNX, YOLOv8
Router	packages/router	The orchestrator. Listens to on-chain events and dispatches tasks to nodes.	TypeScript, Node.js, Ethers.js
Contracts	packages/contracts	Smart contracts handling escrow, task lifecycle, and verification.	Solidity, Hardhat

✨ Key Features

• Verifiable Inference: Every computation comes with a proof of execution.
• Privacy-First: Support for ZK-Blur models to redact sensitive information at the edge.
• Decentralized Storage: Input/Output data is secured via IPFS (Pinata).
• Cost Efficient: Fraction of the cost compared to centralized cloud vision APIs.
• Seamless UX: One-click deployment and intuitive dashboard for task management.

🚀 Getting Started

Prerequisites

• Node.js: v18+
• pnpm: v8+
• Python: v3.10+
• MetaMask: Or any Web3 wallet

One-Command Launch ⚡️

We have provided a unified startup script that handles dependency installation, model setup, and service orchestration.

./start-dev.sh

This will:

1. Install Node.js dependencies (pnpm install).
2. Setup the Python environment and download the YOLOv8 model.
3. Start the dApp (frontend), Node (compute), and Router (backend) concurrently.

Access the dApp at: http://localhost:5173

⚙️ Configuration

Before running in production, you must configure the environment variables for each service. We have provided .env.example files in each package.

1. Compute Node (packages/node/.env)

Controls the AI inference engine and wallet identity.

WALLET_ADDRESS=0x...          # Your node's public address
NODE_PRIVATE_KEY=0x...        # Private key for signing results
MAX_CONCURRENT_TASKS=5        # Max parallel inferences

2. Router (packages/router/.env)

Manages task scheduling and blockchain interaction.

RPC_URL=https://...           # BSC RPC Endpoint
ROUTER_PRIVATE_KEY=...        # Key for submitting transactions

3. dApp (packages/dapp/.env)

Frontend configuration and IPFS keys.

VITE_PINATA_JWT=...           # Pinata JWT for IPFS uploads

🛠 Usage Flow

1. Connect Wallet: Open the dApp and connect your MetaMask.
2. Select Model: Choose from available models (e.g., YOLOv8, ZK-Blur).
3. Upload & Submit: Upload an image. The dApp uploads it to IPFS and creates an on-chain task.
4. Processing: The Router detects the event and assigns it to a Compute Node.
5. Verification: The Node processes the image, generates a proof, and submits the result.
6. Result: The result appears in the "Explorer" tab as a Verified Inference.

🤝 Contributing

We welcome contributions to the MeshVision ecosystem!

1. Fork the repository.
2. Create your feature branch (git checkout -b feature/amazing-feature).
3. Commit your changes (git commit -m 'Add some amazing feature').
4. Push to the branch (git push origin feature/amazing-feature).
5. Open a Pull Request.

📄 License

Distributed under the MIT License. See LICENSE for more information.

---

Built with ❤️ by the MeshVision Team