Larisa Assistant

Larisa is an AI-powered Ethereum transaction interpreter designed for beginners, builders, and curious users who are overwhelmed by traditional block explorers like Etherscan or Blockscout. Instead of presenting raw blockchain data, contract calls, and hexadecimal gibberish, Larisa takes a single Ethereum transaction hash and produces a clear, plain-English explanation of what happened — who sent what, to whom, and why it might matter. At its core, Larisa solves a major onboarding pain in crypto: understanding what’s going on. New users struggle to read transaction details, understand DeFi interactions, or even distinguish between token swaps and contract approvals. Larisa simplifies this.

How It’s Made This project was built using a combination of blockchain data, language models, and lightweight frontend tooling. The goal was to create a conversational AI interface that turns raw Ethereum transactions into understandable summaries — directly in the browser.

Stack & Architecture 🔸 Frontend Framework: Next.js with React UI: TailwindCSS, Heroicons for modern, responsive design Chat Interface: Built from scratch with custom message rendering, scroll control, and input parsing

🔸 Wallet Integration Scaffold-ETH gave us fast wallet hooks and signer access Users can analyze their own transactions or paste any hash

🔸 Blockchain Data Blockscout API: Fetches detailed transactions Rich metadata helps AI generate more accurate and relevant summaries

🔸 AI Layer LLM Used: DeepSeek model through API

Logic: Get user input -> pass to deepseek -> generate blockscout api request -> Return the result back to the user in the chat UI

🤯 Hacky Wins Minimal prompt tuning still produces readable summaries Scaffold-ETH saved hours on wallet boilerplate and EVM setup

🚀 Partner Tech Highlights DeepSeek: Gave us reliable, cost-free LLM access with API Blockscout: Made it easy to enrich txs Scaffold-ETH: Let us go from idea to wallet-ready dapp in one night