NoteSmith
Turns a YouTube lecture playlist into a published, searchable notes site. It fetches the transcripts, drafts structured notes with an LLM key you bring yourself, and publishes straight to GitHub Pages.
Turning a long lecture playlist into organized study notes is slow, repetitive manual work, and most 'AI notes' tools stop at a single ungrounded summary.
A Python CLI pulls transcripts from a YouTube playlist, generates sectioned notes titled from the real video title using whichever provider you point it at (Anthropic, Gemini, or a rotating bucket of free-tier Grok keys), then publishes them to a Jekyll-based GitHub Pages site with full-text search and a custom glassmorphic theme, with a React and Express web UI wrapping GitHub OAuth and job tracking on top.
Built the transcript-to-notes CLI pipeline, the rate-limiting and incremental-update logic, the custom glass Jekyll theme, and the React and Express web frontend for connecting a GitHub repo and tracking publish jobs.
Skipping already-published transcripts instead of regenerating everything made incremental runs fast enough to actually use daily, not just as a one-off demo.