Paste a GitHub link, get a README that actually describes the repository.
Every README generator works on a toy repo and falls over on a real one, because real repositories do not fit in a context window. The problem is not writing prose — it is deciding what the model gets to read.
The reader and the chunker are the project. Everything else is plumbing around that decision.
FastAPI backend, Next.js frontend. Users authenticate and manage their own API keys, so they supply their own model access.
Per-user token accounting. The only way a tool wrapping a paid model stays solvent is knowing who spent what. The ledger is not an analytics nicety; it is the business model.
Tone presets — professional, startup, meme. The README a company wants and the one a side project wants are not the same document.
backend/ FastAPI
services/ generation, auth
utils/
reader.py context-aware code reader ← decides what matters
chunker.py splits what survives ← makes it fit
parser.py
routers/ API + auth endpoints
models/ user, api_key
frontend/ Next.js App Router
Built by Uday Parmar