Thanks — I’m responding to your question about collecting a call transcript after a session when the default webhook doesn’t include it. Short answer: use the agent’s conversation events (specifically the

conversation_item_added

event) to capture the transcript. Room webhooks (like

room_finished

or

participant_left

) do not include conversation content. Why: LiveKit room webhooks are lifecycle/operational signals and do not carry full conversation text. The agent emits

conversation_item_added

each time a message (transcribed user speech or agent response) is added, so it provides a chronological stream of all transcript items. Recommended approach • Subscribe to

conversation_item_added

in your agents-js session and persist each item to your backend as it arrives. ◦ Example (agents-js):

session.on('conversation_item_added', (item) => {
      // item.message.content and item.message.role contain the transcript and who spoke
      // store or forward item to your backend for the full transcript
    });

• If you need timestamps/precise timing, enable TTS-aligned transcriptions (

useTtsAlignedTranscript: true

) and use the transcription nodes/TimedString objects to record

start_time

and

end_time

. • If a call may end abruptly, explicitly close the agent session (

await session.aclose()

) to flush final items into the conversation history before you finalize storage. Additional options • Forward

conversation_item_added

events to your DB or file storage in real time so you can reconstruct the full transcript after the room finishes. • After

room_finished

, pull the stored transcript from your backend (not from the room webhook). If you’d like, I can: • Show a complete agents-js example that stores items to a simple backend endpoint. • Show how to enable and parse the TTS-aligned transcription objects. • Recommend a minimal schema for storing timestamps, role, and content. Which of those would help you next? Sources: Text and transcriptions | LiveKit Documentation | How to capture real-time agent transcriptions during conversations | Best practices for managing webhook event streams | LiveKit | Server-side | livekit/livekit_composite | DeepWiki | Understanding Transcript Timing in Voice Conversations