Back to Zahid's profile
50views

MemoryVerse AI — Personal Digital Second Brain

Zahid Hamdule
  • Expert reviewed
MemoryVerse AI — Personal Digital Second Brain

Overview

Overview

MemoryVerse AI is an AI-powered personal digital second brain that automatically extracts, indexes, and visually maps connections between your digital memories (resumes, project reports, certificates, notes, etc.). It features an interactive 3D/2D Knowledge Graph and a citations-backed RAG Chat Assistant.

Key Design Decisions

  1. Hybrid Vector Database: Combines structured PostgreSQL (Neon DB) for transactional data and document metadata, with ChromaDB for storing and querying text embeddings locally.

  2. Multi-Model AI Pipeline: Utilizes Groq API (using Llama 3) for lightning-fast metadata, topic, and entity extraction. We implemented a Groq key-pool rotation manager with automatic exponential backoff to completely bypass API rate limits.

  3. RAG (Retrieval-Augmented Generation): Uses sentence-transformers (all-MiniLM-L6-v2) for generating semantic vector embeddings, ensuring highly relevant context matches during searches.

  4. Modern UI/UX: Built with React 18, Vite, TailwindCSS, Framer Motion for smooth transitions, and ReactFlow for dynamic connection rendering. Features an interactive guided spotlight onboarding tour.

  5. Cloud-Native Storage: Uses Cloudinary for secure file storage and document signing.

How to Test / Review

  1. Interactive Demo Dataset: Go to the "Upload Files" tab and click the "Import Demo Dataset" button. This will automatically populate the database with pre-packaged resume, certificate, and internship documents so you don't have to upload your own files to see the graph working.

  2. Knowledge Graph: Visit the "Relationships" tab to explore nodes mapping people, locations, organizations, and files. Click on the nodes to expand connections.

  3. Citations Chat: Visit the "Chat Assistant" tab. Ask questions about the uploaded files (e.g. "What projects has Zahid worked on?"). You will see responses containing exact source documents and match percentages.