Argus Overwatch

Argus Overwatch is a news aggregator that ranks stories by coverage breadth instead of clicks. It pulls 26 feeds from 18 news organizations, groups the articles into stories, and scores each story by how many outlets reported it, how recently, and how heavily. The result is a daily edition of what newsrooms collectively decided mattered, split into World, U.S., Business, Technology and Science.

A language model writes each story's description from the headlines and summaries of the underlying articles. Before anything publishes, automated checks verify that every claim in the description is attributable to one of the source articles. The point is a summary you can trace, not a summary you have to trust.

Earlier versions of Argus were an LLM eval and observability platform, and before that a RAG-based media-bias tool with a vector database behind it. v1.0 threw most of that away. It runs in a single Docker container, collects from a few hundred candidate sources, and is built to run on a laptop or the cheapest host that will have it.