All projects

LinkForge AI

AI-powered Web3 portfolio assistant where every AI decision is verifiable on-chain. Uses Chainlink Functions to execute AI analysis on the DON, Chainlink CRE for orchestration, and commits reasoning results directly as on-chain smart contract events on Base Sepolia.

CRE & AI

What it is

LinkForge AI solves the core problem of unverifiable AI in Web3: every decision the AI makes is provable on-chain.
 The system works in three layers:
 1. User sets risk profile, ESG preference, and automation settings — stored on-chain via smart contract on Base Sepolia.
 2. Clicking "Request AI Analysis" triggers Chainlink Functions on the DON. The DON fetches live market data (Fear & Greed Index, CoinGecko), runs EigenAI Deepseek V3.1, and commits the result (action, risk score, sentiment score) directly on-chain as a ReasoningCommitted event.
 3. A Chainlink CRE orchestration workflow ties everything together: reads on-chain profile, aggregates off-chain signals with consensusMedianAggregation, and produces HOLD / SHIFT_TO_STABLE / INCREASE_EXPOSURE / DIVERSIFY recommendations.
No black-box AI. No off-chain trust assumptions. Full audit trail on Base Sepolia.

How it Works

LinkForge AI is a full-stack Web3 application built as a monorepo with four layers:                                                                                                                                                                                 
 1. Smart Contract (Solidity + Foundry) — LinkForgeAI.sol on Base Sepolia inherits Chainlink FunctionsClient and integrates Chainlink Data Feeds. It stores user risk profiles on-chain and exposes requestAIAnalysis() which triggers Chainlink Functions DON execution.
 2. Chainlink Functions (JavaScript) — The DON-executed source fetches live market data from Fear & Greed Index and CoinGecko, calls EigenAI (Deepseek V3.1) via OpenAI-compatible API, and returns encoded AI reasoning (action, risk score, sentiment score) back to the contract via fulfillRequest callback.
 3. Chainlink CRE Workflow (TypeScript) — An orchestration layer in the Chainlink Runtime Environment reads the user profile directly from the smart contract via EVMClient, aggregates off-chain signals using consensusMedianAggregation via HTTPClient, and runs the full AI decision pipeline locally for simulation and testing.
 4. Frontend + Backend (Next.js + Fastify) — Wallet-gated UI built with wagmi and RainbowKit. Users set their profile (risk level, ESG priority, automation), trigger AI analysis from the dashboard, and see results committed on-chain in real time. Backend provides supplementary market signal endpoints.

Links

Created by

  • Imanuel Putra