System Monitor

Real failures as they happened, rather than what a five-minute poll happened to catch. Retried means the automatic retry fixed it and nobody noticed. Fell back means it couldn't be fixed, so the fallback took over and the customer was still served — those are the ones that were emailed to you.

Calls and chats run on OpenRouter. If it stops answering, everything switches to a standby model on OpenAI automatically — a different vendor, so one outage can't take both — and switches back on its own once OpenRouter recovers. You get an email either way.

URGENT issue emails (and RESOLVED follow-ups) go here. Change it anytime — takes effect immediately, no restart.

This host blocks outbound SMTP (ports 25/465/587), so alerts and reports send through the Gmail API over HTTPS instead. Connect the mailbox they should come from — it only ever gets permission to send, never to read your inbox.

Every business has a daily email allowance. Booking confirmations to their customers are never held back by it — only non-urgent mail (review requests, summaries) pauses at the ceiling. The owner is warned on their own dashboard at 80%, 95% and again when it's used up. Change any limit here: blank for the platform default, 0 for unlimited.

A full copy of every business, booking, review and call log. Snapshots are kept on the data volume, which a deploy never replaces — so a deploy that breaks something still has a restore point from just before it. Restoring saves the current state first, so it can always be undone. Download keeps a copy somewhere the volume can't take with it.