It's 2am. A consumer is lagging. You're three nats commands deep, piping into jq, trying to remember if it's --last-per-subject or --deliver-last-per-subject.

Hyerix is the GUI that ends that ritual — and ships an AI that tells you which consumer is broken.

The visual layer covers everything you'd otherwise piece together from five CLI commands: stream configs, consumer lag, pending counts, KV buckets, Object Store, server resource usage, cluster topology. One screen, no jq.

The piece I'm most interested in feedback on is Signal AI — a natural-language layer over your live cluster state. You can ask things like:

- "Which consumers have growing pending counts in the orders stream?"

- "Show me streams whose retention policy doesn't match their discard policy.”

and get answers drawn from real cluster data, not training-time docs. It also runs background anomaly detection — for example, flagging a consumer whose ack-pending grew 4× in an hour and pointing at the subject filter that changed — and can do root-cause analysis when something looks off. Your NATS data never leaves your machine. Signal AI runs locally against the cluster state pulled to your desktop. Nothing shipped to a third-party cloud.

Runs on macOS, Windows, and Linux. Built on Tauri v2.

Try it: hyerix.ai — 14-day trial, no card.

No NATS cluster handy? https://github.com/hyerix/hyerix-demo-cluster spins one up via Docker Compose, pre-loaded with streams, consumers, and synthetic activity.

I'm building this for operators running NATS in production. Tell me what's missing.