Added
- CPU and memory graphs on the device page. The Health card is now the Metrics card: alongside temperature, voltage and fan, it graphs CPU load and memory usage over 1h / 6h / 24h / 7d / 30d. Your routers were already reporting both on every poll — now the readings are kept instead of discarded, so the graphs cost nothing extra on the device.
- Devices polled over SNMP now have history too. Previously they were graphed not at all; they now get the same CPU, memory, temperature, voltage and power history as SSH-polled devices. Switching a device between SSH and SNMP polling keeps its graphs continuous.
- The card no longer hides itself on devices without sensors. A CHR has no temperature sensor, and that used to hide the whole card — CPU and memory history included. It now appears whenever there is something to draw, including for a device that is currently offline, which is exactly when its last hours of CPU are worth looking at.
- History is kept for 90 days by default, and the limit is now adjustable in Settings → Traffic instead of only through an environment variable.
- First part of Discussion #47. Threshold alerting and Grafana/Prometheus export are planned as separate releases.