Tiendil/feeds.fun

What it solves

Feeds Fun is a self-hosted news reader designed to solve the problem of information overload. It allows users to manage numerous news feeds while ensuring they only read the most interesting and important content by using AI-driven tagging and custom scoring rules.

How it works

The system uses a "Librarian" worker to analyze news entries and assign tags using various processors. While it supports simple processors (like extracting domains), its most powerful feature is the use of LLMs (OpenAI and Gemini) to automatically detect relevant tags. These raw tags are then passed through a chain of normalizers to ensure consistency (e.g., removing articles or splitting "vs" terms). Users can then create rules to score news based on these tags, allowing them to filter and sort their feed to prioritize specific topics.

Who it’s for

It is for power users who subscribe to many RSS/news feeds and want a highly customizable, self-hosted way to organize and filter their content using AI.

Highlights

• AI-Powered Tagging: Automatically assigns tags to news entries using ChatGPT or Gemini.
• Custom Scoring: Users can define rules to score news by tags to prioritize content.
• Flexible Deployment: Supports both single-user and multi-user setups via Docker.
• Broad Source Support: Includes improved integration for major sources like Reddit, GitHub, YouTube, ArXiv, and Hacker News.
• Extensible Architecture: Allows for custom tag processors, normalizers, and source-specific plugins.