ChainPilot

This project is an AI-powered on-chain trading assistant that helps users make smarter portfolio decisions without executing trades on their behalf. By connecting a user’s wallet, the system analyzes real-time token balances, portfolio allocation, and on-chain signals to generate actionable insights and swap recommendations.

Instead of acting as a traditional trading interface, the platform focuses on decision intelligence. It identifies portfolio imbalances, risk exposure, and market opportunities, then suggests optimized token swaps along with clear reasoning, confidence scores, and expected impact. Users remain fully in control and can execute trades externally via platforms like Uniswap.

A key differentiator is its memory layer: once a user makes a decision, the context is stored on decentralized infrastructure (0G), enabling transparency, auditability, and future personalization. Over time, the system adapts to user behavior, improving the relevance of recommendations.

The result is a clean, AI-driven experience that bridges wallet analytics, market intelligence, and guided decision-making turning complex DeFi data into simple, actionable insights.

The system is built as a modular backend using NestJS, with a focus on scalability and real-time data processing. Wallet data is fetched using Alchemy APIs, which provide token balances and metadata directly from the blockchain. A custom normalization layer filters spam tokens, standardizes symbols, and prepares clean inputs for further processing.

A dynamic pricing engine combines batched CoinGecko requests, Redis caching, and fallback sources like Binance to avoid rate limits and ensure reliable USD valuation. Portfolio metrics such as allocation percentages, risk exposure, and diversification are computed internally without relying on external aggregation APIs.

The core intelligence comes from a rule-based opportunity engine enhanced with AI reasoning. It detects overexposure, underutilized capital, and market signals to generate swap suggestions. These are enriched with confidence scores derived from portfolio imbalance, market trends, and on-chain activity. For advanced insights, Uniswap’s smart order router is used to simulate swap routes, estimate price impact, and provide realistic trade previews.

A stateful AI agent layer manages user interactions, guiding them through decision-making steps without executing trades. Once a decision is made, the context is stored using a 0G-compatible memory layer, enabling auditability and future personalization.

The frontend is designed using a modern component-based approach (Lovable), focusing on a clean, AI-first UX with minimal clutter and strong visual hierarchy.