Building a semantic search engine for personal notes and documents. Instead of scanning files for exact keyword matches like standard search engines, it maps the conceptual meaning of your queries to find the right information, even if you use completely different words.
what is it: a semantic search tool for text files.
expected workflow:
better than ctrl+f keyword matching 
tech stack:
Next.js + Tailwind frontend
FastAPI + ChromaDB backend
embeddings run locally with all-MiniLM-L6-v2 (no external API calls, everything stays on device)
what’s actually working rn:
upload endpoint that reads .txt/.md, chunks it, embeds it, stores in ChromaDB
search endpoint that does semantic query against the collection
basic CORS + FastAPI scaffolding
frontend is still literally the create-next-app starter page lol, haven’t touched the UI yet
what we do next?
real sentence boundary chunking so we stop cropping mid sentence
reranking layer to actually highlight the matching sentence
then build an actual UI (upload + search) instead of the default next.js landing page
