🧠 RecallMatrix

"Never Forget What Matters" - An AI-powered memory management application with intelligent RAG search capabilities and multi-agent orchestration.

🎯 Overview

RecallMatrix is a sophisticated web application designed to help people remember and retrieve important information effortlessly. Built with cutting-edge AI technology, it combines Retrieval-Augmented Generation (RAG), vector embeddings, and a multi-agent AI system to provide intelligent memory management.

Live Application: https://recall-matrix.vercel.app

✨ Key Features

🔍 Smart Memory Storage & Search

• RAG-Powered Search: Natural language queries using semantic similarity
• AI Embeddings: Automatic indexing with GitHub Models API (text-embedding-3-small)
• Intelligent Retrieval: Vector similarity search for contextual memory matching
• Store text memories and file attachments (images/videos up to 10MB)

🤖 Multi-Agent Orchestration System

A modular AI agent framework with autonomous decision-making:

• Master Orchestrator Agent: Analyzes queries, creates execution plans, and coordinates agents
• Search Agent: Multi-strategy search with automatic fallback mechanisms
• Organization Agent: Auto-categorization, clustering, duplicate detection, tag generation
• Insight Agent: Pattern discovery, trend analysis, relationship mapping
• Reminder Agent: Context-based alerts, time-sensitive notifications, proactive suggestions

🔗 Knowledge Graph Visualization

• Canvas-based visual representation of memory connections
• Force-directed layout with intelligent clustering
• Interactive exploration of relationship networks

💬 Telegram Bot Integration

• Manage memories directly from Telegram
• Add, search, and view memories via chat interface
• Python FastAPI backend for bot operations

🎨 Modern User Experience

• Responsive design optimized for all devices
• Light and dark mode with system preference detection
• Beautiful, minimalistic UI with smooth animations
• Built with shadcn/ui components and Tailwind CSS

🔐 Security & Privacy

• Secure email/password authentication via Supabase Auth
• Row Level Security (RLS) on all database tables
• User data isolation with secure file storage paths
• 100MB storage limit per account with usage tracking

🏗️ Tech Stack

Frontend

• Framework: Next.js 16 (App Router), React 18
• Language: TypeScript
• Styling: Tailwind CSS, shadcn/ui components
• Visualization: Canvas-based Knowledge Graph

Backend

• Database: Supabase (PostgreSQL with pgvector extension)
• Authentication: Supabase Auth
• AI/ML: GitHub Models API
  • text-embedding-3-small for embeddings
  • gpt-4o-mini for text generation
• Storage: Supabase Storage for file uploads
• Telegram Bot: Python FastAPI backend

AI Architecture

• RAG Pipeline: Query → Embeddings → Vector Search → Ranked Results
• Agentic AI: Multi-agent system with autonomous planning and tool use
• Vector Database: PostgreSQL with pgvector for semantic search

🎯 Core Capabilities

RAG Search Flow

1. User enters natural language query (e.g., "Where did I keep my passport?")
2. Query converted to embeddings using OpenAI's text-embedding-3-large
3. Vector similarity search finds semantically related memories
4. Results ranked by cosine similarity
5. Top matches displayed with confidence scores

Multi-Agent Workflow

┌─────────────────────────────────────────┐
│        Master Orchestrator Agent        │
│     (Analyzes, Plans, Coordinates)      │
└─────────────────┬───────────────────────┘
                  │
        ┌─────────┼─────────┬─────────┐
        ▼         ▼         ▼         ▼
    [Search]  [Organize] [Insight] [Reminder]
     Agent     Agent      Agent     Agent

Example Query: "Help me plan my Japan trip"

1. Orchestrator detects search + organize + insight intents
2. Search Agent finds Japan-related memories
3. Organization Agent categorizes as "Travel" and creates clusters
4. Insight Agent analyzes patterns and generates recommendations
5. Synthesized results with actionable insights

💡 Unique Selling Points

1. True Agentic AI Architecture

Not just RAG - implements a complete multi-agent system with:

• Autonomous decision-making
• Multi-step reasoning and planning
• Dynamic tool selection
• Inter-agent coordination

2. Transparent Intelligence

• Visible reasoning process for each agent
• Execution plans displayed to users
• Confidence scoring for results
• Tool usage tracking

3. Modular & Extensible

• Easy to add new agents and capabilities
• Clean separation of concerns
• Well-documented codebase
• Type-safe TypeScript implementation

4. Production-Ready

• Comprehensive error handling
• Security best practices (RLS, data isolation)
• Performance optimized (concurrent agent execution)
• Scalable architecture

📊 Database Schema

Tables

• profiles: User account information and storage tracking
• memories: Text and file memories with vector embeddings
• memory_files: File attachments linked to memories

Storage

• Supabase Storage bucket for secure file uploads
• 10MB max per file, 100MB total per account

🚀 Future Enhancements

• [ ] Third-party integrations (Google Drive, Notion, Evernote, Apple Photos)
• [ ] Advanced memory organization with tags and categories
• [ ] Memory sharing and collaboration features
• [ ] Export functionality (JSON, CSV, PDF)
• [ ] Mobile applications (iOS, Android)
• [ ] Voice memo support with transcription
• [ ] OCR for text extraction from images
• [ ] Agent-to-agent communication
• [ ] Persistent agent memory and learning
• [ ] Background agent execution and scheduling

🎓 Use Cases

Personal Information Management

• Store and retrieve important documents (passport, licenses, certificates)
• Remember where you kept physical items
• Track personal notes and reminders

Travel Planning

• Organize travel-related memories and documents
• Get insights from past trips
• Proactive reminders for visa renewals, passport expiration

Knowledge Management

• Save and retrieve book notes, articles, research
• Auto-categorization of knowledge resources
• Pattern detection in learning habits

Life Organization

• Visual knowledge graph of life events and connections
• Behavioral insights from memory patterns
• Context-aware reminders for time-sensitive information

🏆 Technical Achievements

✅ Advanced AI Integration: RAG + Multi-Agent system
✅ Vector Search: PostgreSQL pgvector with semantic similarity
✅ Real-time Orchestration: Dynamic agent coordination
✅ Knowledge Graph: Force-directed layout visualization
✅ Full-Stack TypeScript: Type-safe end-to-end
✅ Modern Architecture: Next.js 16 App Router
✅ Secure by Design: RLS, authentication, data isolation
✅ Cross-Platform: Web app + Telegram bot

📝 License

MIT License - See LICENSE for details.

🔗 Links

• Live Demo: https://recall-matrix.vercel.app
• Repository: https://github.com/akhilathuluri/RecallMatrix
• Documentation: See /lib/agents/README.md for multi-agent system details

Built by @akhilathuluri | Created: December 2025 | Last Updated: January 2026