Try it
Talk to the actual server.
Pick a model, ask a question. This calls the real home LLM server over an authenticated bridge — if it's offline, you'll get a clearly-labeled saved example instead of a broken page.
Ask about the architecture, the roadmap, or why this project runs locally. Requests try the real home server first, and fall back gracefully if it's offline. Deep reasoning replies can take a while — that's expected, not broken.
Powered by a home Apple Silicon workstation over an authenticated bridge — expect occasional downtime, and always a graceful fallback instead of an error.