Thanks — I’m responding to your question about using a SIP trunk to dial out and having your Web UI join only to monitor/stream the room transcript without being an active audio participant. Yes — you can join a room for monitoring and receive the room transcript without publishing an audio track. The recommended approach is to disable audio output for the agent session and consume transcript text from the

lk.transcription

text stream. This keeps the SIP participant (dialed via your trunk) as the active audio party while your Web UI only receives transcripts and other data. What to do (high level) • Start your agent session with audio output disabled so no agent audio track is published (example: set

audio_output=False

in

RoomOutputOptions

). • Have your UI join the room but do not publish audio/video tracks — join as a receive-only participant. • Listen to the

lk.transcription

text stream topic from the room; the agent publishes its text responses there. • Filter messages by the

lk.transcribed_track_id

attribute to distinguish live speech transcriptions (transcribed from audio tracks) from agent text responses (agent text responses will not include

lk.transcribed_track_id

). Minimal code pointers 1. Disable audio output when starting the agent session (Python example):

await session.start(
    agent=MyAgent(),
    room=ctx.room,
    room_options=room_io.RoomOptions(
        audio_output=False,  # disable TTS/audio output
    ),
)

1. Register a text stream handler on the frontend (TypeScript example):

room.registerTextStreamHandler('lk.transcription', async (reader, participantInfo) => {
    const message = await reader.readAll();
    // if no lk.transcribed_track_id attribute, this is agent text response
    if (!reader.info.attributes['lk.transcribed_track_id']) {
        console.log(`Agent response: ${message}`);
    } else {
        console.log(`Speech transcription: ${message}`);
    }
});

Notes and gotchas • The built-in playground UI uses legacy transcription events and may not show these text-streamed agent responses when audio tracks are not published. You’ll need a custom receiver in your UI to read

lk.transcription

. • Joining the room without publishing audio/video still allows you to receive text streams and monitor the call. • This approach is suitable for integration testing where the SIP trunk handles the live call participant and the Web UI is used purely for monitoring/transcript consumption. If you’d like, tell me: • Which SDK/browser/client you’re using for the Web UI (JS/TS, React, etc.), and • Whether you want agent responses shown alongside live transcriptions or merged into a single transcript view. I can then provide a short, specific code example for your client and a suggested UI pattern for displaying agent vs. transcribed-speech messages. Sources: How to build an agent with speech-to-text input and text-only output | Live... | Outbound Calling Script | LiveKit Documentation