2026.10.17
FeedMe: A Self-Hosted RSS Reader with AI Summaries
FeedMe is an MIT-licensed lightweight RSS reader that aggregates feeds and generates summaries through an LLM. It supports GitHub Pages, Vercel, Alibaba Cloud ESA Pages, and Docker, with deployment trade-offs to review.
What it is for
FeedMe is designed for people who want to read multiple RSS feeds in one responsive web interface without adopting a large client or a mandatory account workflow. It aggregates feed content and uses an LLM to generate article summaries. The repository is released under the MIT License and is tagged with RSS, LLM, GitHub Actions, and self-hosted topics.
The project is closer to a static-site generation workflow with a reading interface than to a full server-side RSS platform with user accounts. Updates, summary generation, and site builds can run through GitHub Actions, or through scheduled tasks inside a Docker deployment.
Known features
FeedMe can combine multiple RSS sources and display them by category. Its documented configuration file lets users change source names, URLs, categories, the default source, and the number of retained items per feed. Summary settings include the prompt, input truncation length, temperature, maxTokens, fallback text, and output locales. The documentation lists zh,en as the default summary locales.
For deployment, the GitHub Pages workflow can fetch RSS content, generate summaries, and build the static site. The documentation says the workflow runs hourly by default and can also be triggered by a push or manually. Build output can be pushed to a deploy branch for deployment by Vercel or Alibaba Cloud ESA Pages. The Docker option runs update and build tasks inside the container. The interface includes responsive design and light and dark themes.
Who may benefit
FeedMe is a reasonable fit for technically comfortable individuals who want to maintain their own feed list, as well as developers who prefer deploying a reading page on a static hosting platform. It is particularly relevant if you already have access to an LLM service and are comfortable managing source and summary settings in configuration files.
It may be less suitable if you only need conventional RSS reading without AI, or if maintaining Actions, scheduled container tasks, and an external model service would add unwanted operational work.
Limitations and deployment considerations
AI summarization is not self-contained. The documented GitHub Pages setup requires an LLM API key, base URL, model name, and summary locales; the Docker setup also requires corresponding environment variables. Users therefore need to evaluate the selected model provider’s availability, cost, quotas, and output quality. The supplied project materials do not establish universal pricing, performance, or reliability guarantees.
RSS fetching and scheduled updates also depend on either a workflow or a container cron task. Changing the schedule requires editing the relevant configuration. Feed access restrictions, format differences, and source stability should be checked before production use. When deploying through a static hosting pipeline, teams should understand the role of the deploy branch and limit branch permissions and secret exposure appropriately.
The supplied sources confirm the MIT License and several deployment paths, but do not confirm minimum hardware requirements, a specific compatibility list for LLM providers, or a security audit. Before adoption, review the current repository configuration, workflow permissions, logs for accidental secret exposure, and whether feed content is appropriate to send to the chosen model service.
Sources
Sources checked: 2026-10-17