RecallMatrix

An AI-powered memory management web application with intelligent RAG (Retrieval-Augmented Generation) search capabilities. Built for people who need help remembering important information, MemoryVault uses advanced embedding technology to help you find memories through natural language queries.

🎉 New: Telegram Bot Integration!

Manage your memories directly from Telegram!

✨ Add memories, search, and view your collection via our Telegram bot. See TELEGRAM_INTEGRATION.md for setup instructions.

Features

• Smart Memory Storage: Save text memories and file attachments (images/videos)
• RAG-Powered Search: Search your memories using natural language queries
• AI Embeddings: Automatic indexing using GitHub Models API (text-embedding-3-small)
• Intelligent Retrieval: Vector similarity search for semantic memory matching
• Knowledge Graph: Visual representation of memory connections with clustering
• Telegram Bot: Manage memories from Telegram (see TELEGRAM_INTEGRATION.md)
• User Authentication: Secure email/password authentication via Supabase
• Theme Support: Light and dark mode with system preference detection
• Storage Management: Track storage usage with 100MB total limit
• Responsive Design: Optimized for all devices and screen sizes
• Modern UI: Beautiful, minimalistic interface with smooth animations

Tech Stack

Frontend

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

Backend

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

Integrations

• Telegram Bot: Full-featured bot for memory management
• More coming soon: Google Drive, Notion, Evernote, Apple Photos

Setup Instructions

1. Prerequisites

• Node.js 18+ installed
• A Supabase account and project
• GitHub Models API access

2. Supabase Setup

1. Create a new Supabase project at https://supabase.com
2. The database migrations have already been applied via the MCP tools
3. Enable the vector extension in your database if not already enabled
4. Note your project URL and anon key from Settings > API

3. GitHub Models API Setup

1. Get access to GitHub Models at https://github.com/marketplace/models
2. Generate an API key for accessing the models
3. This app uses:
   • text-embedding-3-large for generating embeddings
   • gpt-4o for text generation (future features)

4. Environment Variables

Update the .env.local file with your credentials:

NEXT_PUBLIC_SUPABASE_URL=your_supabase_project_url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your_supabase_anon_key
GITHUB_MODELS_API_KEY=your_github_models_api_key

5. Install Dependencies

npm install

6. Run Development Server

npm run dev

Open http://localhost:3000 in your browser.

7. Build for Production

npm run build
npm run start

Usage

Creating Memories

1. Click "Add Memory" in the navigation bar
2. Choose between "Text Memory" or "File Memory"
3. For text memories: Enter a title and detailed content
4. For file memories: Enter a title and upload images/videos (max 10MB per file)
5. Memories are automatically indexed with AI embeddings

Searching Memories

1. Use the search bar on the home page
2. Enter natural language queries like:
   • "Where did I keep my passport?"
   • "Photos from last vacation"
   • "Notes about the book I wanted to read"
3. The RAG system will find semantically similar memories

Managing Your Account

• Settings: Update your profile and monitor storage usage
• Theme: Toggle between light/dark mode via the profile menu
• Integrations: Coming soon - sync with Google Drive, Notion, etc.

File Storage Limits

• Maximum file size: 10MB per file
• Total storage limit: 100MB per account
• Supported formats: Images and videos only

Database Schema

profiles

• User account information and storage tracking

memories

• Text and file memories with vector embeddings
• Indexed for efficient retrieval

memory_files

• File attachments linked to memories

Storage

• Supabase Storage bucket for file uploads

Architecture

RAG Search Flow

1. User enters a search query
2. Query is converted to embeddings using OpenAI's text-embedding-3-large
3. Vector similarity search finds relevant memories
4. Results are ranked by cosine similarity
5. Top matches are displayed to the user

Security

• Row Level Security (RLS) enabled on all tables
• Users can only access their own data
• File storage paths include user ID for isolation
• API keys are stored securely on the server side

Future Enhancements

• Third-party integrations (Google Drive, Notion, Evernote)
• Advanced memory organization with tags and categories
• Memory sharing and collaboration
• Export functionality
• Mobile applications
• Voice memo support
• OCR for text extraction from images

License

This project is private and proprietary.

Support

For issues or questions, please contact the development team.