LLM-powered web app for natural-language Q&A on YouTube channel data
The project lets anyone ask plain‑English questions about a YouTube channel and receives cited answers drawn from video transcripts. It builds a searchable corpus from local JSON reports, serves a React+TypeScript front‑end on GitHub Pages, and proxies Anthropic API calls through a Cloudflare Worker so secrets stay hidden. It’s useful for non‑technical teammates who need quick insights without manual transcript digging, and its server‑less architecture makes deployment cheap and simple. Compared to ad‑hoc scripts, it offers a polished UI, secure secret handling, and instant, grounded responses.
View on GitHub →John-Iles/YTAnalyser