MemoryVerse AI '26

This project implements all 5 modules end-to-end: AI ingestion, categorization, a relationship knowledge graph, growth timeline, and natural-language semantic search. Note on technical approach: given time and offline-reliability constraints, categorization and search use TF-IDF vector similarity + cosine similarity rather than transformer-based embeddings — a genuine vector-space technique, chosen for zero API-key dependency and instant, demo-safe startup. The relationship engine uses networkx to build a real knowledge graph connecting documents through shared skills. Full reasoning and tradeoffs are documented in the Thought Process Sheet. To try it: clone the GitHub repo, run pip install -r requirements.txt then streamlit run app.py, and upload the sample files in /sample_data for an instant populated demo. No login or credentials required.