Monitoring

Health

curl -fsS http://127.0.0.1:8787/health

answers when the node is up. Use it from a timer, your monitoring system, or after a deploy.

The service

systemctl status mandala
journalctl -u mandala --since "1 hour ago" | grep -iE "error|warn"

Restart=always brings Mandala back after a crash or a deliberate restart; frequent restarts in the journal mean something is wrong.

Disk

Every confirmed write waits for a disk sync, and a full disk stops writes. Alert at 80% on the data directory's volume.

In the app

• Settings → Overview flags expired sign-ins and extensions that are not connected.
• Settings → Devices and sync shows each peer's state and how many of its changes are held here.
• Initiatives shows background token use against its daily limit.
• Activity shows failed turns with their errors.

Coming soon — operations dashboards: node health, sync lag, queue depths and model spend in one page.