🎙️ Voice
Talk to your mesh — and get an answer short enough to listen to.
Voice ships a spoken-answer agent tuned for text-to-speech surfaces. It answers in at most two speakable sentences, runs on the chat model tier, reaches for mesh lookups or web search only when the question actually needs them, and hands long work to the portal instead of reading it aloud.
What ships
- The
Voiceagent — registered for you when you install the package, and reached by name from every speech surface. - Radio stations — a handful of ready-made
Stationnodes to ask for by name. - A prompt you can edit — the agent's speaking rules are a node, not a compiled constant.
Where it speaks
- A voice satellite on your LAN, through the memex voice gateway — wake word, streaming replies, conversation mode, and fire-and-forget delegation to the standard agents.
- The mobile app — the chat microphone records and transcribes through the portal.
- The portal — the composer's mic uses the same centralized speech service.
Speech-to-text runs on the centralized Whisper container; Swiss German transcribes out as Standard German. Deployment, model builds and licensing are in the Guide.