Panda Trade

Problem Statement

Panda Trade is an AI-assisted multi-token trading tool designed for crypto users who want to execute smarter, more personalized trades without complex research.The user starts by connecting their wallet and choosing a blockchain network. Then they input the amount they want to trade in USDC, select their preferred risk level (safe or risky), and choose interest categories from a multi-select list — such as Meme, DeFi, RWA, Blue Chip, Gaming, AI, Infrastructure, DAO, Low Cap, Short Term, Long Term, or High Liquidity.Our system uses this data as input for an AI model, which returns a recommended multi-token trade – including selected tokens and suggested allocation percentages.The user can review the proposed trade, and if satisfied, confirm it to execute the entire multi-token purchase in one step.Panda Trade saves time, reduces decision fatigue, and offers a fun and simple way to explore crypto trading based on personal interests and risk profile — with help from the most intelligent panda in Web3.

Solution

Github repos: https://github.com/lebrande/panda-trade https://github.com/Cezary24/ETHGlobal-Prague-vlayerWe used VLayer for we proing that the answer really comes from LLM and it's not spoofed. We used Blockscout to get tokens data to feed our LLM engine, also we used Blockscout SDK to provide best in class UX and transactions feedback.Other technologies:VercelOdos aggregatorRainbowkitOpenAINext.jsviem + wagmiAlchemyProduct presentation:User starts the journey with selecting:USDC amountrisk appetitetags or topics of their today investmentsIn the second step Panda finds the best matchusing Blockscout API data about tokenssends that data to LLMLLM answers with 5 matches combining Blockscout data and user preferencesUser sees the breakout of selected tokensPanda uses Odos agregator to create a multiswapUSDC in the inputall selected tokens on outputwith proportion of Panda's choiceIn meantime we use VLayer to generate Web Proofwe proove that the answer really comes from LLM and it's not spoofedThen transaction is supposed to run only if the ZK proof is attached to the calldataFinally user approves the ERC20 transfer and executes the tradePanda uses Blockscout SDK to display feedback for all the transactionstoast messages pops out everytime users sings the transactionlast screen shows the output from Blockscout API showing transfer detailsMoreover uses can lookup their all recent transactionsPanda Trade uses Blockscout SDK to present that in the popup