MemoryPlus

This project is an intelligent Telegram bot that revolutionizes how users interact with blockchain data and personal information management. The bot combines three core functionalities:

🧠 AI-Powered Memory System: Users can naturally store personal experiences, preferences, and information through conversational AI. The system uses dynamic schema inference to automatically categorize and structure data, allowing users to recall information with natural language queries like "What restaurants did I like in Tokyo?"

⛓️ Multi-Chain Token Integration: Real-time token data integration across three major blockchains (Ethereum Mainnet, Flow EVM, and Rootstock) using Blockscout APIs. Users can search tokens, view prices, holder counts, trading volumes, and market charts through simple commands.

💬 Smart Conversation Interface: Powered by GPT-4, the bot understands context and intent, automatically determining whether users want to store memories, recall information, fetch blockchain data, or simply chat. The AI maintains conversation history and provides contextual responses.

Key features include persistent memory storage, natural language processing, multi-chain blockchain data aggregation, and seamless user experience through Telegram's familiar interface.

This project leverages a modern tech stack optimized for rapid development and scalability:

Backend Architecture (Node.js/TypeScript):

• Telegraf.js for Telegram Bot API integration
• OpenAI GPT-4 for natural language processing and intent classification
• SQLite database for rapid prototyping with structured memory storage
• Express.js server for health checks and potential API expansion

AI & Memory System:

• Custom prompt engineering to classify user intents (store/recall/blockchain/chat)
• Dynamic schema inference that creates data structures on-the-fly
• Contextual memory recall using text search and categorization
• Chat history maintenance for conversational context

Blockchain Integration:

• Blockscout APIs across three chains: Ethereum Mainnet, Flow EVM, and Rootstock
• RESTful API integration for token search, details, and market charts
• Error handling and fallback mechanisms for multiple chain queries
• Custom response formatting for optimal mobile display

Deployment:

• Railway for backend hosting with automatic deployments
• Environment-based configuration for secure API key management
• TypeScript compilation with proper type safety

Notable Technical Achievements:

• Single conversational interface that intelligently routes to different functionalities
• Real-time multi-chain data aggregation without requiring users to specify chains
• Memory system that adapts to user's personal data patterns
• Responsive error handling across all external API dependencies

The hackiest part was implementing dynamic schema inference - the AI analyzes user input patterns and automatically creates appropriate data categories and metadata structures, making the memory system truly adaptive to each user's needs without predefined schemas.