ARYA

ARYA is a swarm of four specialized AI agents that work together to discover, evaluate, and execute yield farming opportunities across DeFi protocols. Unlike fully autonomous fund managers, ARYA keeps humans in control: agents analyze and recommend, users approve, and smart contracts enforce the approval gate on-chain. AI agents are powerful but keeping Human in the loop is the ethical way.

I had an idea to utilise AI to come up with strategies to maximise yield potential. But I need visualization on top of that for explainability so that it is not a black box to user and the actions of the AI are made aware to the user. ARYA is built as a multi-agent orchestration to execute DeFi yield farming strategies - with the human always in the loop before any capital moves.

As I'm solo, I have leveraged ChatGPT and GitHub Copilot to validate my ideas and come up with a system. There were several iterations to adjust the overall implementaion process.

The agents are written using TypeScript with LangGraph.js Scout Agent - Gathers data from DefiLlama eg. APY, TVL, token pairs etc.

Risk Agent - Evaluate yield farming strategy with Multi-factor Risk Assessment and Portfolio Correlation Analysis.

Executor Agent - Activates only after a user explicitly approves a strategy. Perform swaps and initiates monitoring.

Orchestrator Agent - Acts as the conductor. Routing messages between agents, managing shared state etc.

Relevant Smart Contracts Smart Contract I : Implements the ERC-7857 iNFT standard. Agents mint an iNFT with encrypted metadata stored on-chain. This means every recommendation in ARYA is traceable back to a specific, identifiable agent (not an anonymous black box).

Smart Contract II : On-chain enforcement of Human-in-the-Loop Design. Agents can propose strategies anytime but the user ultimately decides.

Smart Contract III : Tracks agent accuracy over time by comparing predicted APY against actual outcomes. It helps mitigate the trust problem in AI-driven finance. Users are actively made to verify transactions made by AI agents instead of blindly trusting AI agents.

Partner Technology Integrations 0G is foundational to ARYA. We use 0G chain as the deployment target for all smart contracts and ERC-7857 agent identity. 0G Storage serves as the persistent memory layer for the entire swarm.

Uniswap's Trading API is the execution backbone. The API's structured response format made it straightforward to pipe swap calldata directly into our StrategyVault contract for on-chain execution.

KeeperHub is tightly coupled with our Executor Agent via REST API calls. Once user approves and a position is opened, the agent creates monitoring workflows. KeeperHub does the heavy lifting, so our agents don't have to babysit positions and thereby, freeing up resources. This was critical for the "set it and monitor it" experience we wanted.

The frontend consists of a live strategy feed with agent status indicators, a strategy detail view with a risk radar chart, an agent registry with iNFT identities and an execution history with KeeperHub workflow status.

The most notable design decision was making the "human-in-the-loop" pattern architecturally mandatory rather than optional. ARYA's data is heavily reliant on the blockchain rather than a vector database. A knowledge base with a vector database would have introduced extra complexity with embedding pipelines, chunking logic etc.

Tech Stack Summary Frontend - TypeScript, Next.js, shadcn/ui, Recharts Agent Orchestration - LangGraph.js Smart Contracts - Solidity + Hardhat Wallet connectivity - wagmi + viem + RainbowKit Yield Data - DefiLlama API and CoinGecko API Trading API - Uniswap Automation - KeeperHub Agent Memory - 0G Storage