Orcha-net

Orchanet is a decentralized orchestration protocol that spawns specialized multi-agent committees via 0G Compute. It transforms parallel, trustless inference into automated DeFi execution through Uniswap's Universal Router, packaging the entire verifiable execution graph into immutable 0G Storage artifacts.

It is a trustless multi-agent orchestration layer bridging 0G Compute parallel inference, 0G Storage provenance, and automated Uniswap Permit2 execution.

Orchanet utilizes ENS-registered intelligent NFTs (iNFTs) to route parallel inference tasks across the 0G Network, employing adversarial critic agents for trustless consensus. Upon validation, the orchestrator seamlessly bridges AI logic into smart contract automation by signing off-chain Permit2 approvals and executing swaps on Ethereum.

Orchanet is built as a full-stack Next.js (App Router) application, written entirely in TypeScript. The core architecture relies heavily on bridging natural language processing with Web3 infrastructure, utilizing Ethers.js (v6) for seamless wallet interaction and transaction signing on the Ethereum Sepolia network.

Instead of treating AI as a traditional centralized backend, we engineered an Orchestrator Node that relies directly on our sponsor track technologies to handle intelligence, identity, provenance, and execution:

0G Compute (The Intelligence Layer): Traditional LLM integration wasn't going to cut it. We needed our multi-agent committee (Tokenomics Modeler, Smart Contract Auditor, and the Critic) to run in a decentralized environment. We utilized the 0G Compute SDK to route our specialized inference tasks. The Orchestrator parallelizes these requests, handles the 0G compute fee settlements via the ledger, and pipes the specialized outputs into an adversarial debate—all powered by 0G nodes.

0G Storage (The Provenance Layer): A core problem with AI is the lack of verifiability. Once the Orchestrator reaches a consensus and executes a task, it compiles the entire "Run Record"—which includes the raw user prompt, the individual agent responses, the critic's debate, and any executed transaction hashes—into a JSON artifact. We then integrated the 0G Storage SDK to fragment and upload this artifact directly to 0G Storage nodes, returning an immutable Root Hash to the UI so users can audit exactly how the AI arrived at its decision.

Uniswap API & Permit2 (The Execution Layer): We didn't want Orchanet to just give advice; we wanted it to take action. We integrated the Uniswap Trade API directly into our Orchestrator pipeline.

The Orchestrator uses regex to extract "trade intents" from the user's natural language prompt (e.g., "buy 0.0001 ETH"). We query the Uniswap API for the optimal routing quote. Instead of forcing the user to do multiple on-chain approval transactions, we implemented Permit2. The Orchestrator requests a single off-chain EIP-712 signature from the user's wallet. The system packages the Permit2 signature with the swap calldata and broadcasts it through the Uniswap Universal Router on Sepolia.

ENS (The Identity Layer): To ensure users are interacting with genuine, specialized agents, we treat our AI agents as Intelligent NFTs (iNFTs). We integrated ENS (Ethereum Name Service) via Ethers.js to resolve on-chain identities. Before an agent is allowed to participate in a debate, the Orchestrator resolves its .eth domain (e.g., tokenomics.orchanet.eth) to verify its smart contract address, capabilities, and ownership, ensuring trustless participation.

Technical Challenges: We Overcame Our biggest hurdle was marrying asynchronous, long-running AI inference streams with strict on-chain execution requirements. Destructuring the nested Uniswap Quote API responses and correctly mapping the EIP-712 Permit2 domains for the Sepolia network required meticulous debugging. Furthermore, managing the asynchronous upload of large JSON artifacts to 0G Storage while keeping the UI responsive led us to build a custom event-emitter pipeline that streams live updates to the frontend Studio UI.