Autonomous AI agent monitoring Arbitrum DeFi (DeFiLlama, Uniswap v3, Aave v3), producing a 0-100 sentiment score and recording each report immutably on-chain via a Solidity contract.
ArbiSense is an autonomous AI agent that continuously monitors Arbitrum DeFi protocols and records market intelligence on-chain — with zero human intervention.
Collect: Fetches live data from DeFiLlama (Arbitrum chain TVL), Uniswap v3 (volume & TVL), and Aave v3 (pool utilization rates) using public APIs.
Analyze: Produces a 0–100 sentiment score — TVL momentum 40%, Uniswap volume/TVL ratio 30%, Aave avg utilization 30%.
Record: Calls submitReport() on the deployed SentinelRegistry Solidity contract, storing each report immutably on Arbitrum Sepolia.
DeFi participants need timely, unbiased, on-chain intelligence. ArbiSense removes the human bottleneck — operating 24/7 and writing structured reports directly to the blockchain without any manual trigger.
Python · web3.py · Solidity 0.8.20 · Hardhat · DeFiLlama API · Arbitrum Sepolia
<p>Built ArbiSense from scratch during the hackathon:</p><ul><li><p>Designed three data collectors (DeFiLlama, Uniswap v3, Aave v3) using public APIs</p></li><li><p>Built weighted sentiment scoring engine (0–100 scale, configurable weights)</p></li><li><p>Wrote and compiled <strong>SentinelRegistry.sol</strong> (Solidity 0.8.20) with <code data-inline="true" spellcheck="false">submitReport()</code>, <code data-inline="true" spellcheck="false">onlyAgent</code> access control, and full report history</p></li><li><p>Integrated web3.py for on-chain transaction signing and submission</p></li><li><p>Created CLI with run-once, run-loop, reports, onchain-status commands</p></li><li><p>Validated full dry-run: Score 35/100 (Fear), Arbitrum TVL $1.72B, UniV3 TVL $223M</p></li><li><p>Published to GitHub: github.com/Steffdarz/arbisense</p></li></ul>
<p>Not fundraising. ArbiSense is an open-source hackathon project. Seeking ecosystem grant from Arbitrum Foundation.</p>
