Skip to main content

Documentation Index

Fetch the complete documentation index at: https://www.rumus.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

A Monitor card is a real-time chart of system metrics for a host, sitting right alongside your terminals on a workspace canvas. Use it to watch CPU spike during a deployment, see memory drift over an afternoon, or tell at a glance whether a GPU is doing any work.

Add a monitor

Monitor cards live on the canvas, so you need an open workspace.
1

Open or create a workspace

From the tab bar, switch to a workspace tab — or create one via the workspace switcher.
2

Add a Monitor card

Click New Monitor in the canvas toolbar, or right-click empty canvas space and pick New Monitor.
3

Pick a host

The card defaults to monitoring the local machine. Open the card’s settings to point it at a saved SSH host instead — one card per host.
4

Resize and place it

Default size is 400 × 400 (minimum 300 × 200). Drag the corner to resize; place it next to the terminals you care about.

Metrics shown

The card streams these metrics in real time and keeps the last 60 data points for a rolling chart:
MetricDetail
CPUOverall usage (0–100%) as a line chart.
MemoryUsed / total, with a percent-used line chart.
DiskUsed / total (human-readable), with usage percent.
GPUOne sub-section per detected GPU — name, memory used / total, utilization %, and temperature in °C.
When the host has no GPU detected, the GPU section shows Not detected and the rest of the card carries on as usual.

Refresh interval

The card pulls fresh metrics every 2 seconds by default. The interval is configurable per-card — bump it down for tighter granularity during an incident, or up for less network chatter on a constrained connection.

Connection status

Monitor cards have a small status pill showing what’s going on:
  • Connecting — the card is establishing the metrics stream.
  • Waiting for data — connected, but no samples have arrived yet.
  • Connected — live samples flowing.
  • Disconnected — no data for ~10 seconds, or an error occurred. A Retry button appears; the card retries automatically up to a few times before giving up.

What it doesn’t do (yet)

  • Per-process breakdown — Monitor cards show host-level metrics. For per-process detail, run htop / top in a terminal next to the card.
  • Historical persistence — the chart keeps the last ~60 samples in memory and resets when you close the workspace. For long-term graphs, point a real metrics system at the host (Grafana, Datadog, etc.).
  • Alerts — Monitor cards don’t fire notifications on threshold crosses today.
  • Custom metrics — you can’t yet add bespoke graphs (network throughput, custom systemd unit health, etc.).

Tips

  • Lay out one Monitor card per host alongside the terminals you have open to that host. Visual cause-and-effect is fast feedback when troubleshooting.
  • For a fleet, combine Monitor cards with multi-terminal broadcast: broadcast the action, then watch every host react in parallel.
  • Keep a Monitor card on your local machine in your “scratch” workspace — it’s a surprisingly useful baseline before you go suspect a remote host.

Next steps

Workspaces & canvas

Where Monitor cards live; how to arrange them next to terminals.

SSH host management

Save the hosts you’ll be monitoring.