Sentinel Oracle

Demo

About

**Sentinel Oracle** is an AI-powered, privacy-preserving oracle system that makes DeFi safer by detecting anomalies in real-time price feeds and automatically executing protective actions. It's a complete solution that combines three cutting-edge technologies to solve a critical problem in decentralized finance. ### The Problem We Solve DeFi protocols depend on oracles for pricing data, but these oracles can be: - **Manipulated** through flash loan attacks or MEV strategies - **Delayed** due to network congestion - **Temporarily incorrect** due to API failures This leads to cascading liquidations and millions in financial losses. Currently, there's no way to automatically detect and respond to anomalous price data while preserving user privacy. ### Our Solution Sentinel Oracle continuously monitors asset prices through **Pyth Network's Pull Oracles**, uses **AI-powered anomaly detection** (via ASI Alliance) to identify suspicious patterns, and executes **privacy-preserving automated protections** through Lit Protocol's Vincent framework. **Example Flow:** 1. BTC price drops 10% in 2 minutes ⚠️ 2. AI agent detects anomaly (z-score > 2.5) 🤖 3. Lit Protocol decrypts user's private trigger conditions 🔓 4. Vincent executes stop-loss automatically ⚡ 5. User's position is protected 🛡️ ### Key Features **Real-Time Price Monitoring** - Fetches live price data from Pyth Network via Hermes API - Updates on-chain every 5 seconds using Pull Oracle method - Supports multiple assets (currently BTC/USD, easily expandable) **AI-Powered Anomaly Detection** - Uses statistical analysis (z-score) to detect unusual price movements - Sliding window approach with 30-sample history - MeTTa reasoning engine provides explainable AI decisions - Configurable thresholds **Privacy-Preserving Automation** - User trigger conditions encrypted with Lit Protocol - Threshold encryption ensures privacy until execution - Vincent framework handles automated transaction execution - No one can see user's private strategies until conditions are met **Beautiful Dashboard** - Live price visualization with Recharts - Real-time anomaly alerts - AI agent chat interface (ASI:One simulation) - User deposit and trigger management - System statistics and monitoring ### Technical Architecture The system consists of four main components: 1. **Smart Contract** (`SentinelOracle.sol`) - Stores prices, manages deposits, handles anomaly flags 2. **AI Agent** (`agent`) - Monitors prices, detects anomalies, provides reasoning 3. **Privacy Layer** (`lit_vincent/`) - Encrypts triggers, executes automated actions 4. **Frontend** (`frontend/`) - User interface and real-time visualization ### Real-World Impact This system could protect millions of dollars in DeFi positions by: - Detecting oracle manipulation attempts before they cause damage - Automatically executing stop-losses during market anomalies - Providing transparency through AI reasoning - Maintaining user privacy while enabling automation

How it's made

## How It’s Made – Sentinel Oracle ### Overview * Full-stack DeFi intelligence system. * Integrates **Pyth Network** (data), **ASI/uAgents** (AI reasoning), and **Lit Protocol** (privacy & automation). ### Smart Contracts (Solidity + Hardhat) * Built using Solidity 0.8.20, Hardhat, and Pyth SDK. * Implements **dual price layers**: * Fresh Pyth data for accuracy. * Cached data for fast AI access. * Uses dynamic asset mapping (`keccak256(asset)` → Pyth feed ID). * Optimized for low gas and easy extensibility. ### AI Agent (Python + uAgents) * Developed with Python, Flask, Web3.py, NumPy, and uAgents. * Uses **z-score anomaly detection** over a sliding price window. * Incorporates a **MeTTa-inspired reasoning engine** for explainable insights. * Includes **30-second cooldown** to prevent redundant anomaly alerts. ### Privacy & Automation (Lit Protocol + Vincent) * Uses Lit Protocol SDK v3 and Vincent framework on Datil testnet. * Encrypts user triggers off-chain; stores only hash on-chain for gas savings. * Implements a **StopLossAbility** that decrypts and executes automated actions. * Uses **access control conditions** tied to on-chain states for secure automation. ### Frontend (Next.js + Ethers.js) * Built with Next.js 14, React 18, Tailwind CSS, Recharts, and Ethers.js v6. * 5-second smart polling for real-time updates. * Multi-asset dashboard with z-score–based color indicators for anomaly visualization. ### Integration Flow * Data pipeline: Hermes API → Smart Contract → AI Agent → Flask API → Frontend → Lit/Vincent Execution. * End-to-end latency: ~15–30 seconds per detection cycle. ### Key Highlights * Dynamic asset mapping without contract redeployment. * Hybrid on-chain/off-chain price layers. * Gas-optimized (hash-only) on-chain data storage. * Mock data fallback ensures system resilience. * Python–JS bridge for cross-language data flow. ### Performance * Update interval: every 5 seconds. * Uptime: ~99.9% during testing. * Gas cost: ~$3–8 per full cycle. * Adaptive 30-sample AI window for stability and responsiveness. ### Why It’s Special * Combines **accurate data**, **explainable AI**, and **privacy-preserving automation**. * Creates an oracle that is intelligent, efficient, and transparent—ideal for secure real-time DeFi operations.