Use the web UI

The Lumen web UI is the fastest way to check fleet health, drill into one host, and manage agents. It is served by the hub, so there is no separate frontend service to deploy.

Open the dashboard

After the hub is running, open the hub URL in your browser:

http://<hub-host>:8090

On first run, Lumen asks you to create the admin account. After that, sign in with that account to reach the dashboard.

Dashboard at a glance

The dashboard summarizes the current state of every host that has sent metrics to the hub.

Area	What it means
WebSocket status	Shows whether the browser is receiving live updates from `/api/stream`.
Hosts	Total number of known hosts and how many are currently online.
Stale	Hosts that have not reported recently. Check the agent service or network if this is non-zero.
Average CPU / RAM	Fleet-wide average from the latest snapshots.
Search	Filters the host cards by host name.
Host cards	Show current CPU, RAM, disk, load, and recent activity for each monitored host.

Click any host card to open its detail page.

Add a host

Go to Settings → Hosts.
Enter a stable host name, for example pve-01, lxc-postgres, or nas.
Click Create.
Copy the lum_... token immediately. The token is only shown once.
Install or configure the agent on that machine with the copied token.

For installation commands, see Agent — Docker Compose or Agent — Linux.

Read a host detail page

The host detail page combines live values with historical charts.

Status: Up means the host is reporting now. Stale means the hub has seen it before but not recently. Waiting means no metrics have arrived yet.
System line: shows the primary IP or hostname, operating system, and uptime when the agent reports them.
Time range: switch charts between 1h, 6h, and 24h.
Per-core CPU: appears when the agent reports per-core data.
Charts: CPU, RAM, disk, load average, network throughput, disk I/O, and temperature when available.
Containers: appears when the agent can read Docker container state from the target host.

Charts refresh automatically. The live snapshot updates over WebSocket, while historical chart data refreshes periodically.

Manage tokens and agents

Use Settings → Hosts for token and host lifecycle tasks.

Action	Use it when
Rotate token	A token was lost, exposed, or copied incorrectly. The old token stops working immediately.
Delete host	You no longer want the hub to accept or display that host.
Create host	You are onboarding a new monitored machine.

Do not create a new host just to update an agent binary. For updates, keep the same host and token; see Update Docker agents if you run agents with Docker.

Configure runtime behavior

The Settings page contains several sub-tabs:

Hosts: create hosts, rotate tokens, and delete hosts.
Account: view the signed-in user and change the password.
Runtime: change the agent collection interval shown by the UI.
Retention: control how long raw metrics are kept.
Downsample: configure bucket size and hot/archive windows for lower-resolution history.
Logs: preview log-management behavior and limits.

After saving runtime, retention, or downsample settings, verify that agents and charts still update at the expected cadence.

Change language or theme

Use the header controls to switch language and theme. The dashboard is designed for both desktop and mobile widths, so the same web UI can be used from a phone browser.

Troubleshooting

The dashboard shows no hosts : Create a host in Settings → Hosts, then start an agent with that host’s token. The first card appears after the agent’s next collection tick.

A host is stale : Check that the agent service is running, the hub URL is reachable from the target, and the token has not been rotated.

WebSocket status is disconnected or error : Confirm the browser can reach the hub and that any reverse proxy forwards WebSocket upgrades to /api/stream.

Charts have live values but little history : Wait for more collection ticks, then try a shorter range such as 1h. If history remains empty, check retention and downsample settings.

Container data is missing : The agent only reports containers when it can read Docker state on the target. For Docker agents, make sure the Docker socket is mounted read-only as documented in the install guide.