RecallOS

RecallOS is your personal second brain - a blockchain-verified AI memory system that captures, organizes, and retrieves your digital experiences with the intelligence and reliability of a human memory, but with the power of artificial intelligence and blockchain verification. It addresses the fundamental problem of information overload in our digital lives by creating an intelligent, decentralized personal knowledge graph that never forgets and always remembers. In today's digital age, we consume massive amounts of content daily but retain very little of it, just like how our biological brain struggles to remember everything we encounter. Traditional bookmarking systems are static and disconnected, while search engines provide generic results that don't understand our personal context. There's no way to verify when or where we learned something, and no system to connect related information across time. RecallOS solves these fundamental issues by creating a personal AI memory system that functions as your digital second brain, capturing everything we encounter online and making it searchable, verifiable, and interconnected.

The system combines cutting-edge AI, blockchain technology, and semantic search to create an intelligent memory vault that works like your biological brain but with superhuman capabilities. RecallOS automatically captures web content as users browse through a browser extension, processes it with AI to extract key insights and generate semantic embeddings, and stores it with blockchain verification for immutable proof of capture - just like how your brain processes and stores memories, but with perfect recall and verification. Users can then search their memories using natural language queries and receive AI-powered answers with source citations, while exploring their knowledge through an interactive 3D memory graph that shows relationships between different pieces of information, mirroring the neural networks in your brain. The system is built as a full-stack application with multiple interfaces, featuring a Node.js backend with Express and TypeScript, PostgreSQL database with pgvector extension for vector similarity search, Redis for job queues, and Prisma ORM for database management. The frontend includes a React-based web dashboard with Three.js for 3D visualization, a browser extension for automatic content capture, and an SDK for programmatic access. The AI processing pipeline uses Google Gemini API with Ollama as a local fallback, while blockchain integration is handled through Ethereum Sepolia testnet with custom smart contracts for memory verification.

The system's core features include intelligent content capture through browser extension with smart deduplication, AI-powered processing that summarizes content and extracts metadata, semantic search with vector similarity and AI-generated answers, blockchain verification for immutable proof of capture timestamps, and an interactive 3D memory graph for visual exploration. The hybrid AI system ensures reliability with multiple fallback options, while the gas deposit system allows users to pay for blockchain transactions without managing gas fees directly. Users simply install the browser extension and browse the web normally - content is automatically captured and processed in the background, just like how your brain automatically processes and stores information as you experience the world. Through the web dashboard, users can search their memories using natural language queries like "What did I learn about blockchain?" and receive comprehensive AI answers with source citations. The 3D memory graph provides an intuitive way to explore connections between different pieces of information, while blockchain verification ensures users can prove when and where they learned something, giving you the confidence of a perfect memory with verifiable proof.

The project demonstrates significant technical innovation through its hybrid AI system that combines cloud and local processing, advanced vector search capabilities using pgvector, sophisticated blockchain integration with gas deposit management, and cutting-edge 3D visualization. The system handles async processing for AI tasks, implements comprehensive caching for performance optimization, and supports massive scale with 1M+ memory capacity and 10M+ relationship edges, giving you a second brain that can store and process information at a scale far beyond human capacity. The development process involved setting up the React frontend and Express backend, integrating AI processing and vector search, implementing blockchain functionality, and building the browser extension with 3D visualization. The result is a production-ready system with comprehensive API endpoints, robust error handling, and seamless user experience across all interfaces.

RecallOS represents a fundamental shift in personal knowledge management, providing users with an AI-powered memory system that's blockchain-verified and always accessible - essentially giving you a second brain that never forgets, always remembers, and can prove what it knows. The immediate value includes never forgetting important web content, connecting related information across time, and verifying learning sources, just like having a perfect memory with photographic recall. Future potential includes personal AI training on individual memory data, collaborative team knowledge graphs, enterprise-wide memory systems, and multi-chain blockchain support for broader accessibility. The live demo showcases the complete system functionality, from automatic content capture through the browser extension to intelligent search and 3D visualization. Users can install the extension, browse any website, and immediately see their content captured and processed. The search functionality demonstrates AI-powered answers with citations, while the 3D memory graph shows the interconnected nature of captured information. The system is fully functional with real blockchain transactions on Sepolia testnet and production-ready architecture. RecallOS represents the future of personal knowledge management - a blockchain-verified AI memory system that captures, connects, and preserves your digital experiences forever, giving you the power of a second brain that's smarter, more reliable, and more verifiable than your biological memory.


Core Architecture & Memory Storage System RecallOS combines blockchain verification, AI processing, and semantic search to create an intelligent personal memory system. Content is captured through browser extension, web client, or SDK, then processed by AI to generate summaries and extract metadata. A cryptographic hash is created for blockchain verification, with smart deduplication preventing duplicate storage.

AI-Powered Processing with Hybrid Fallback The system uses a sophisticated hybrid approach with multiple fallback mechanisms. Primary processing uses cloud-based AI services for summarization and vector embedding generation, with fallbacks to local AI processing and deterministic embedding systems. The multi-tier approach tries cloud services first, then multiple local AI models, and finally falls back to custom embedding generators using text analysis and semantic clustering.

Vector Search & Memory Mesh The system uses PostgreSQL with vector extensions for high-performance similarity search. The memory mesh creates a knowledge graph by generating embeddings and finding semantic relationships using cosine similarity. Three relationship types are created: semantic (content similarity), topical (metadata overlap), and temporal (time-based proximity with decay). Dimensionality reduction techniques create 3D force-directed graphs showing natural memory clusters.

Blockchain Integration & Verification Smart contracts use upgradeable proxy patterns with a gas deposit mechanism where users deposit cryptocurrency to cover transaction costs. A relayer system submits transactions on behalf of users with retry logic and gas estimation. Each memory is verified on-chain with content hash, URL hash, and timestamp, providing immutable proof of capture.

Envio Hyperindex for Event Processing The system uses Envio Hyperindex for blockchain event indexing and processing. Event handlers automatically process all smart contract events including memory storage, gas transactions, and authorization events. The indexing system updates user statistics, system statistics, and maintains real-time analytics. This provides a complete audit trail of all blockchain interactions and enables complex analytics queries across the entire system.

Blockscout Integration for Transaction Monitoring The system integrates with Blockscout for comprehensive transaction monitoring and data prefetching. An intelligent caching system prefetches transaction data from multiple networks including Sepolia, Mainnet, Polygon, Arbitrum, and Optimism with network-specific finality rules. An in-memory queue system handles retry logic and exponential backoff, automatically handling rate limiting and network-specific block times. Transaction data is cached with finality status tracking, allowing the system to provide real-time transaction status updates.

Advanced Search & Retrieval The search system implements multiple strategies: pure vector similarity search, hybrid search combining keyword matching with semantic search, and context-aware search with AI-generated answers. Optimized database queries with vector operations provide maximum performance, while sophisticated scoring algorithms blend keyword relevance with semantic similarity. Comprehensive caching and search job tracking enable efficient long-running AI answer generation.

Database Schema & Performance The database schema is designed for high performance with proper indexing on vector columns, user relationships, and timestamp fields. Modern ORM tools provide type-safe database operations with connection pooling for scalability. The system supports millions of memories and relationship edges with sub-100ms query performance for complex semantic searches.

Development & Deployment The system uses Hardhat for smart contract development with comprehensive testing using Foundry, and multi-environment support for local development, Sepolia testnet, and production deployments. Automated deployment scripts, contract verification on Etherscan, and comprehensive monitoring provide transaction tracking, gas usage analytics, and performance metrics. The architecture supports horizontal scaling with Redis job queues, PostgreSQL databases, and multiple AI providers for redundancy.