System

The System page reports runtime details: update status (current / available), database and proxy state, the log-write queue, and GPU poller / device status.

See Updates & health and Log retention.

Top-level status

The first row summarizes:

• Process uptime and Node version.
• Request-log queue depth and dropped log count.
• GPU vendor, device count, and last snapshot age.
• Proxy upstream host and whether insecure TLS is enabled.

Runtime panels

The page then breaks down the process state:

• Control Bus - active backend, upstream URL, upstream host, /metrics, direct-upstream count, and allowed direct targets.
• Inference Backend - capability flags for models, running-model state, lifecycle actions, logs, and config editing.
• Loading Queue - queued and dropped request-log writes. Dropped rows mean logging was shed to avoid blocking proxy traffic.
• GPU Telemetry - detected driver, device count, primary device, memory, utilization, temperature, power, and cores when available.
• Runtime - uptime, Node version, current git commit, and update check status.
• Database - SQLite path and whether it is a special path such as :memory: or a file: URI.

Maintenance

Manual prune and compaction controls live on the Settings page. The System page is the place to confirm whether the queue, database, proxy, or GPU poller needs attention before running maintenance.