YOR.AI

Every sysadmin starts the day with the same three questions: is anything wrong, is it new, and what do I do about it? Answering them means digging through metrics, scrolling logs, and remembering what you tried last week. Foreman does that digging for you. It’s a crew of four. Chip collects metrics and runs the Muse CLI locally on the server for a first read. Chuck scrapes the logs with no model at all, so the Reviewer works from facts. The Reviewer, on Muse Wonder, reads today against seven days of history and writes a one-page report: what’s wrong, whether it’s new or recurring, how bad. The Fixer, on Muse Spark, writes an executable fix for anything that needs one, with the exact commands, the risk, the side effects, and how to undo it. Then a person decides. The report lands on the dashboard and on your phone through Telegram, with Apply and Rerun under each proposed fix. Nothing runs until you tap Apply. Rerun asks for a different approach and explains how the new one compares. Foreman remembers. A fix that failed shows up in the next review, and the Fixer won’t propose it again. When a problem is on a host it can’t reach, it says so and hands you a manual step instead of inventing a command. Guardrails live in code, not in the prompt: no destructive commands, no touching its own containers, no acting without approval. It runs with docker compose up and one API key. Everything in the repo was built during the hackathon.