MS Abhishek
Ask your own notes in English, Hindi or Hinglish — cited answers from your notes only, or an honest refusal.
Built at Pitch 2026
Team MS Abhishek: MS Abhishek
Built with Python 3.12 (standard library only), WSGI, Vercel, BharatCode Model API (Qwen 3.8 27B, DeepSeek fallback), HTML/CSS/JavaScript
About the project
NoteVault Bharat lets students and professionals ask their own notes — in English, Hindi or Hinglish — and answers only from those notes. The problem: knowledge sits scattered across notebooks, apps and files; keyword search finds words, not answers, and generic chatbots answer from the open internet in English and guess when they don't know. NoteVault keeps notes on the user's machine; when a question is asked, only the matching excerpts go to the model (BharatCode API — Qwen 3.8 27B with a DeepSeek fallback). Every answer cites the notes it used, and shows its route, model and budget. If the notes don't contain the answer, it refuses before any model call, at zero cost, instead of guessing. A 3D constellation maps the vault, with a plain list fallback. It is live today, MIT-licensed and open source, with 52 passing tests. The public demo currently runs in a clearly labelled mock mode; the real model path activates with an API key. Built solo by M S Abhishek, a BTech CSE student at Shobhit University, Meerut.