Docs menuVoice & observability

Voice & observability

Voice input

Composer microphone: capture on-device, then transcribe with Whisper (@xenova/transformers + bundled tiny model). Works offline after the first model load. Needs system mic permission. STT is available in the desktop app.

Light observability

  • Usage analytics: Settings → Usage — totals and a 14-day trend; breakdown tabs for Members (per assistant / squad) and Models; gateway tokens also expose usage per assistant / squad
  • Message trace: assistant bubbles can open a waterfall timeline (model / tools / approvals / compression; Developer mode entry points)
  • Error logs: daily jsonl under the data directory logs/

Enough to see what an assistant did. Light usage and tracing stay on the gateway host — not a full APM.

Content is derived from the OkBot product repo and evolves with the product.

PreviousSquadsNextFAQ