What it shows
Open AI Activity with Cmd+K, or click the pulse icon in the titlebar; it pulses with an amber dot whenever AI work is running. The page shows what is using the AI engine on your Mac, as plain activities rather than logs:
- A status line: whether the engine is idle, running tasks, or loading a model, and which model is currently held in memory.
- Happening now: requests in flight, with a live timer. Routines waiting their turn behind your chat show as "waiting its turn".
- Recent: the last activities with status, model, duration, and when they ran. Rows marked automatic are work the app started for you.
Why is the GPU busy when I'm not chatting?
nenva runs AI in the background on your behalf: email insights and bundling after a mail sync, routines, memory tidy-ups after a chat, project-progress updates, and model warm-ups (loading the model so your first message answers fast). Loading a model into memory is itself the heaviest moment for the GPU, even before any request runs. All of it shows on this page, so a busy GPU always has a visible reason.
Each activity happens on the model set for it: a new install starts on nenva Cloud, and you can switch to a model on your Mac, your own server, or a provider you added a key for, in Settings → AI Assistant → Models.
From activity to full detail
Click a row to open the exact request in LLM Logs (full prompt, response, and token counts), or use Open detailed logs for the whole list. The activity list is per-Mac and does not sync; Clear empties it without touching the logs.