A connected room does not prove the browser captured a microphone, and a microphone indicator does not prove the agent received usable audio. Diagnose each boundary separately instead of showing one generic connection error.
LiveKit's camera and microphone guide describes permission and publication behavior. Check the relevant client SDK and browser environment for the implementation being reviewed. This is a diagnostic design, not a tested browser script.
Model the frontend states explicitly
Separate permission pending, permission denied, device unavailable, capture active, track published, room connecting, room connected, agent joined, and remote playback. Show the user the state that matters and a real supported recovery action.
Keep authentication and network failures distinct from device access. Repeatedly asking for microphone permission will not repair an invalid connection token or an unavailable agent. Likewise, refreshing a room connection may not change a denied device permission.
Rehearse a fictional first visit
A user opens a browser voice widget and denies microphone access. The application should explain the capture problem and offer the approved help or alternative route. It should not claim the business is offline if the actual issue is local capture.
A second fixture grants capture but makes the agent unavailable. The UI should distinguish that service state and avoid a misleading microphone troubleshooting loop. Logs retain diagnostic categories without unnecessary personal audio.
| Browser fixture | Expected distinction |
|---|---|
| Permission declined | Capture failure, not business-action failure |
| No usable device | Device-selection or support route |
| Room connection rejected | Authentication or connection diagnosis |
| Agent absent | Service availability state |
| Audio published, no response | Inspect downstream media and agent evidence |
Preserve the request when voice is unavailable
Offer an actual callback form, telephone route, or staff contact if the product supports it. Do not invent a fallback that merely opens another broken voice session. Make clear whether any customer request has been submitted.
Detailed frontend states improve diagnosis but can overwhelm users. Keep technical categories in controlled logs while using concise user-facing explanations. Avoid exposing tokens, internal URLs, or raw exception details in the interface.
Test denied and later-restored permissions, interrupted connections, device changes, and repeated join attempts with controlled accounts. Review debugging voice failures for connecting frontend evidence to backend outcomes.
Questions about microphone failures
Is joining a room enough to show “ready to speak”?
Use the actual capture, publication, and agent-readiness evidence required by the product rather than one connection flag.
Should the app keep retrying permission prompts?
Use supported browser behavior and a clear user choice. Repeated prompts are not a substitute for explaining the problem.
What should support receive?
Client version, browser context, diagnostic stage, non-sensitive identifiers, and reproducible steps, with private audio collected only through the approved process.