Static-site RSS reader that makes LLMs do the skimming

What it does FeedMe collects RSS feeds and uses a configurable LLM to generate article summaries during a build step, outputting a React-based static site deployable to GitHub Pages, Vercel, Alibaba Cloud ESA, or a local Docker container. There are no user accounts, no runtime database, and no persistent server—just HTML and JSON refreshed on a schedule.

The interesting bit The entire “backend” is either a GitHub Actions workflow or a Docker cron job that fetches feeds, calls your LLM API, and rebuilds the site. The browser receives a fully static app that happens to contain AI-generated digests. It is essentially a build-time RSS aggregator with an LLM preprocessor.

Key highlights

• Zero hosting cost on GitHub Pages; the output is pure static files with no server to maintain
• Works with any OpenAI-compatible LLM API, letting you choose the model and control prompts, temperature, and token limits
• Supports multi-language summaries via locale configuration
• Docker deployment includes an internal cron scheduler for self-hosted builds without relying on GitHub Actions
• Responsive UI with dark and light themes built on shadcn/ui and Tailwind

Caveats

• Requires your own LLM API key and base URL; there is no built-in model or free inference tier
• Feed updates only arrive when the scheduled build runs, so this is not a live, push-based RSS client
• The README does not mention support for authenticated or private RSS feeds

Verdict A solid fit for developers who want a low-maintenance, self-hosted news digest without running a persistent backend. Look elsewhere if you need real-time notifications, social features, or a native mobile app.