Logs

The Logs page streams raw llama-swap, proxy, and upstream logs into one viewer over server-sent events. The stream is only subscribed while the page is mounted, so it adds no overhead to the rest of the dashboard.

Sources

The viewer labels each line by source:

• llama-swap - proxy/control-plane log lines.
• llama.cpp - upstream model server log lines.

The page is available only when the active inference backend exposes log streaming. With llama-swap, llama-dash uses the backend log stream capability shown on the System page.

Controls

• Filter by all, upstream, or proxy lines.
• Search with plain text or regex (.*).
• Toggle wrapping for long model-server lines.
• Clear the in-browser buffer without affecting backend logs.
• Auto-scroll follows the tail until you scroll away from the bottom.

Log lines are virtualized in the browser, so long sessions remain usable without rendering every row at once.