Analysis-AVS

This project is an AI-powered agent built on EigenLayer AVS that analyzes Ethereum transaction histories in a fun, insightful way. The agent listens for new task requests triggered by smart contract events. When a task is created, it retrieves the transaction history and details for a specified wallet address from the The graph and Etherscan API.

The fetched transaction data is processed using OpenAI’s GPT-4 model to generate detailed insights. The analysis covers common transaction patterns, detection of suspicious activities, identification of high-value transactions, and an examination of unusual gas usage—all delivered with a unique, playful roast-style commentary that makes blockchain data both informative and entertaining.

The analysis is then rendered into a dynamic PDF report that combines comprehensive textual insights with colorful, engaging charts and visuals generated using pdfkit and Chart.js. In addition, a creative, “Spotify Wrapped”-style image is generated using DALL-E to further enhance the visual presentation of the report. This signed PDF report is subsequently submitted back on-chain as the task response, ensuring full transparency and verifiability within the EigenLayer ecosystem.

Leveraging technologies such as Node.js, TypeScript, Ethers.js, Solidity, and the OpenAI API, this project provides a robust and interactive tool for blockchain analysis that not only delivers valuable insights but also presents them in an appealing, user-friendly format.

Our project leverages Node.js with TypeScript for the off-chain agent, which uses Ethers.js to interact with Ethereum smart contracts deployed on the Base Sepolia network. The agent fetches transaction histories from the Etherscan API using Axios and supplements this data by querying The Graph for additional insights—including Uniswap swap data, Hop Protocol bridge transfers, and ENS domain information. All of this comprehensive data is then processed with OpenAI’s GPT-4 model to generate deep, insightful analysis. Additionally, DALL-E is used to generate a creative “Spotify Wrapped”-style image, while pdfkit combined with Chart.js produces dynamic PDF reports that blend textual insights with engaging charts. On-chain, Solidity smart contracts integrated with EigenLayer AVS and secured via ECDSA signature verification manage task creation and response submissions.

When a new task is triggered by a smart contract event, the off-chain agent retrieves the designated wallet’s transaction data from Etherscan and the supplementary data from The Graph. This combined dataset is fed into GPT-4, which generates detailed insights covering transaction patterns, anomalies, and risk metrics. Simultaneously, DALL-E produces a visually appealing image to complement the analysis. The agent then compiles this information into a dynamic PDF report that incorporates both the textual analysis and colorful charts. The final report, along with its associated metadata (such as the DALL-E image URL), is signed by the operator and submitted back on-chain through the smart contract’s respond function, ensuring full transparency and verifiability.

To optimize on-chain storage, we designed the system to store only minimal task data and cryptographic signatures on-chain, while keeping the rich analysis and visual assets off-chain, accessible via IPFS. We also tackled the challenge of integrating multiple AI services—carefully engineering our prompts for GPT-4 and DALL-E to ensure complete and accurate JSON outputs and image generation. This combination of efficient on-chain verification with dynamic off-chain reporting creates an innovative and interactive tool for blockchain analysis, transforming raw Ethereum data into engaging, actionable insights.