Stats
The Stats panel shows live resource usage for the selected sandbox.
Opening it
Click on any sandbox row in the Sandbox table to open the slide panel, then choose the Stats tab.
What you see
| Metric | Description |
|---|---|
| CPU | Current CPU load as a percentage, plotted as a rolling sparkline. |
| Memory | RSS memory used by the shmastra pm2 process, plotted as a sparkline. |
| Disk | Filesystem usage for the sandbox root partition. |
All three are polled every few seconds while the panel is open.
When to use it
- Diagnosing slow responses — an agent stuck in a loop often shows CPU pegged at 100 %.
- Checking memory pressure — if memory climbs toward the cgroup limit the healer will eventually trigger an emergency restart. You can pre-empt it by restarting from the Terminal tab.
- Verifying a sandbox restarted — after an update, confirm the process came back online and memory settled to a baseline.
Automatic healer threshold
The healer agent monitors the same signals in the background:
- Memory > 90 % of the cgroup limit, sustained for ≈ 1 minute → emergency heal.
- Load average > 2.5 per CPU core, sustained for ≈ 1 minute → emergency heal.
See Healer agent for details.