Voice commands reference
The Core App is voice-first — there's no UI to navigate beyond a small status display on the glasses.
The Core App is voice-first — there's no UI to navigate beyond a small status display on the glasses. Almost everything inside a session is conversational; the session lifecycle itself is gated on a tiny handful of explicit commands.
Session lifecycle — explicit commands
| Say | Effect |
|---|---|
| "Start training session" | Opens the voice channel, starts POV recording, joins the AI participant. |
| "Stop training session" | Closes the session cleanly. Uploads kick off; the post-session pipeline extracts procedures (trainer mode) or seals the trainee session. |
These are the only two explicit lifecycle commands today. There's no pause, resume, or cancel — once a session is open you finish it with "Stop training session" and the recording is committed.
Inside a session — conversational
There is no wake word. Once the session is open, the AI is always listening on a server-side voice channel. Speak naturally — examples (all interpreted by the AI, not pattern-matched to fixed commands):
- "What's the gas valve part number for this Worcester 30CDi?"
- "Show me the exploded view for the heat exchanger."
- "Open the manual for this boiler."
- "How do I bleed the system?"
- "Look at this fault code and tell me what it means."
The AI decides which skill to call based on your phrasing — search_knowledge, capture_camera, or any of the platform / tenant skills available on your seat. See The AI assistant for the orchestration model.
In trainer mode
In trainer mode the AI listens but generally doesn't intervene — there are no extra fixed commands. Just narrate naturally as you work. Procedure extraction is fully automated by the post-session pipeline; you don't need to mark steps, flag safety, or signal step boundaries by voice.
When the AI can't hear you
| Symptom | Likely cause | Fix |
|---|---|---|
| No response at all | Voice channel dropped | Say "Stop training session" then "Start training session" again. |
| Repeated misrecognition | Background noise / accent | Move somewhere quieter, or pair a Bluetooth mic (see Troubleshooting). |
| AI replies in the wrong language | Tenant locale mis-set | Dashboard → Settings → Locality → Language. |
Where to next
- Trainee mode — the conversational model in detail
- The AI assistant — what it can call
- Troubleshooting — connectivity, audio, pairing fixes