NeuraFi
Unlock DeFi Yields with AI-Powered Optimization
Demo
About
The problem it solves # Yield Optimizer Protocol     **AI-Powered DeFi Yield Optimization Protocol** *Intelligently allocates user capital across DeFi strategies using off-chain AI agents and Chainlink Automation* --- ## 📖 Overview The **Yield Optimizer Protocol** is a sophisticated DeFi platform that combines the power of artificial intelligence with blockchain automation to optimize yield generation across multiple DeFi strategies. Built on Ethereum with ERC4626-style vaults, the protocol intelligently allocates user capital based on risk preferences and real-time market conditions. ### 🌟 Key Features - **🧠 AI-Powered Strategy Selection**: Off-chain AI agent analyzes market data from CoinGecko, DeFiLlama, and protocol APIs - **⚙️ Chainlink Automation**: Automated on-chain execution of AI recommendations - **🎯 Risk-Based Allocation**: Support for Low Risk (Aave) and High Risk (Morpho/Curve) strategies - **📊 Real-Time Market Analysis**: Dynamic strategy selection based on APY trends and market conditions --- ## Deployed Contracts (Sepolia) ``` MockAavePool: 0x6deb956da4300Fe538efB475D05e818e7D4F3699 MockMorpho: 0x779D4755eF3574467Da5B4909629106794A9dd28 VaultFactory: 0xd0954eA81Aa99f320C0cB2A6703e1319c196441c LowRiskVault: 0xb32a6FF65dcC2099513970EA5c1eaA87fe564253 HighRiskVault: 0x721bF349E453cbFB68536d3a5757A70B74D84279 Low Vault Owner: 0xcdd0a2f1Ed2Fcf90486f510F85d9fc1249C21D2e High Vault Owner: 0x204554c6B13f436a30e77B986de525f436b7Cf1C LowRiskStrategy: 0x5E7F3A1b83ecC5Bed41F8CC2b13DcD75BeD525Ee HighRiskStrategy: 0xBf84a91eDD1455b7Df150038F2004d00C27ef5c0 USDC_SEPOLIA = 0x1c7D4B196Cb0C7B01d743Fbc6116a902379C7238 ``` ## Chainlink Autoamtion Upkeep https://automation.chain.link/sepolia/371634965542298094432024965539145112188713635505165377177798693816686038288 https://automation.chain.link/sepolia/6159270574854941854967137437462153565675183198460466722719413083942550900961 ## 🏗️ Architecture ### Core Components ```mermaid graph TB A[User Deposit] --> B[YVault] B --> C[VaultFactory] C --> D[Strategy Selection] D --> E[Low Risk: Aave] D --> F[High Risk: Morpho] G[AI Agent] --> H[Market Data Analysis] H --> I[Strategy Recommendation] I --> J[Chainlink Automation] J --> K[On-Chain Execution] B --> K E --> K F --> K ``` ### 1. YVault (Yearn V2-Style Vault) The core vault contract implementing ERC4626-like functionality: - **Asset Management**: Accepts USDC deposits and tracks user shares - **Strategy Integration**: Routes funds to active yield strategies - **Security Features**: Reentrancy protection, admin controls, emergency pause - **Automation Ready**: Chainlink keeper integration for automated execution ### 2. VaultFactory - **Dynamic Deployment**: Creates vaults based on risk profiles - **Strategy Association**: Links vaults with appropriate strategies at creation - **Registry Management**: Tracks all deployed vault instances ### 3. Strategy Contracts - **LowRiskAaveStrategy.sol**: For conservative users, integrates with mock Aave pool. - **HighRiskMorphoStrategy.sol**: For aggressive users, integrates with mock Morpho pool. - **MockStrategy.sol**: For testing vault mechanics. ### 4. AI Agent (ElizaOS) - **Data Sources**: CoinGecko, DeFiLlama, Aave/Morpho APIs - **Analysis**: Market trend analysis, APY comparison, risk assessment - **Output**: Strategy recommendations stored in ElizaOS KV store - **Integration**: Seamless connection with Chainlink Automation ### 5. Chainlink Automation - **Trigger**: Scheduled or event-based strategy execution - **Security**: Role-based access control for automation functions - **Integration**: Direct connection with AI agent recommendations ### Made With Love ❤️ By 1. Yash Sharma 2. Nagateja 3. Krishna ---
How it's made
Challenges we ran into Primary challenge was of simulating the real world protocol and their pool, we tackled that in our project by creating mocks of them Another challenge was of integrating two automation in a single vault using chainlink automation, one to report strategy and allocate funds and other to automate set strategy which is called based on eliza agent's response. In the end we also faced the problem of weaving all these functional parts into this project in which shed blood, sweat and tears