A Venue-Expert CTA and Wiring a Temps Indicator Into ArcLLM

Enabling the Whitney Venue Expert CTA

I enabled the Whitney venue-expert call to action on the venue sidebar (de06d04c). The CTA had existed in the markup but was effectively disabled, so visitors browsing a venue had no obvious path to talk to the venue expert for that location. Turning it on is a small change with a direct business consequence: the sidebar now does the job it was designed for. I mirrored it to the Next.js frontend (330ea2d7) so the experience is consistent across both surfaces.

Adding a Temps Indicator to ArcLLM

The other half of the day was plumbing the triple-temps-indicator into ArcLLM (b472b0c3). Local model servers are opaque about thermal and utilization state, and I wanted a lightweight way to see whether the hardware was saturated or just idle-waiting. The indicator surfaces temperature and utilization together, which is the pair you actually need to tell the difference.

Wiring it into ArcLLM means the API stack can report the hardware context alongside its own queue depth. When a request is slow, I can now distinguish a busy model from a throttling one.

The Pattern

Two unrelated projects, both about connecting something that already existed to the place it was needed. The CTA was present but switched off; the indicator existed but wasn’t visible from the API. A lot of useful work is just completing circuits.