MemoryVerse AI '26

MemoryVerse is an AI-powered digital identity system that ingests academic/professional documents (PDF, images, DOCX, TXT), auto-classifies them into categories (Project, Skill, Certification, etc.), extracts text, generates embeddings, and enables natural language semantic search. It also discovers relationships between documents (e.g., a certification connects to a skill, which connects to a project) and visualizes them on a timeline and relationship graph. Built with FastAPI + React + SQLite + ChromaDB + Google Gemini Flash, deployable via Docker or free on Hugging Face Spaces. Setup: backend/.env needs GEMINI_API_KEY. Run with Docker: docker compose up --build (port 7860). Or local: cd backend && pip install -r requirements.txt && uvicorn app.main:app --reload + cd frontend && npm install && npm run dev. Test credentials: Register a new account at /register (any email/username/password works). No pre-seeded users. Design decisions: SQLite + ChromaDB for zero-setup local dev. Google Gemini 2.5 Flash for classification/embeddings. Hybrid search (semantic + category-keyword boost). JWT auth with data isolation per user. Single-container Dockerfile for HF Spaces deployment.