Updates & health

llama-dash checks for updates against the GitHub main branch head commit (with a short in-memory cache) and surfaces current / available / error state in the System runtime panel. The same panel reports database, proxy, log-queue, and GPU poller health.

Update checks

To install an update, follow Updating.

The update checker compares the running git commit to the main branch head from:

https://api.github.com/repos/ndom91/llama-dash/branches/main

Results are cached in memory for 1 hour per running commit.

Statuses:

• unknown - the running commit is unknown, so llama-dash cannot compare versions.
• current - the running commit matches the latest branch commit, or the latest branch commit starts with the running short SHA.
• available - GitHub reports a different latest commit.
• error - the GitHub check failed.

When the status changes, llama-dash publishes a system-change event so the dashboard can refresh.

Runtime health

The System page reports:

• Runtime uptime and Node version.
• Current git commit.
• Update status and latest commit URL when available.
• Database path and whether it is a special SQLite path.
• Inference backend capabilities.
• Proxy upstream URL, host, direct-target count, and TLS mode.
• Request-log queue depth and dropped rows.
• GPU poller status and device telemetry.

Interpreting health

Use the queue and GPU panels as operational signals:

• Non-zero log queue depth can be normal during bursts.
• Increasing dropped log count means llama-dash is shedding request-log writes to protect proxy traffic.
• GPU unavailable means no supported GPU telemetry source was detected, or polling failed for the detected source.
• Update check errors usually mean GitHub was unreachable from the process, not that the app is unhealthy.