I'm glad you're looking to save the full call transcript between the user and agent to your database when the call ends or the user leaves the room. LiveKit's transcription features can definitely help with that! Here's a friendly outline to guide you through the process: • First, enable transcription by using the

VoicePipelineAgent

or

MultimodalAgent

. Once set up, transcriptions will be automatically forwarded to clients if your frontend supports it. • Next, listen for transcription events in your application. For example, if you're using JavaScript, you can use this snippet:

room.on(RoomEvent.TranscriptionReceived, (segments) => {
  for (const segment of segments) {
    console.log(`New transcription from ${segment.senderIdentity}: ${segment.text}`);
    // Save the segment.text to your database here
  }
});

• Each transcription segment comes with useful metadata like the sender's identity and the text itself. You can store these details in your database as you see fit. • If you want to tailor the transcription behavior, there are options to disable transcription output or enable TTS-aligned transcriptions—feel free to customize it to what works best for you. One important note: the

RoomEvent.TranscriptionReceived

event is deprecated and will be removed in a future version. To keep your implementation future-proof, it’s best to use text streams on the

lk.chat

topic instead. Please don’t hesitate to reach out if you have any questions or need a hand with the implementation—I’m here to help! Sources: Transcriptions | LiveKit Docs | Text and transcriptions | LiveKit Docs | Docs::Agents

I'm glad you're interested in saving the full call transcript using

session.history

in the Python SDK! Let me walk you through how you can do this step-by-step: 1. Save Full Transcript at Session End A great way to save the transcript is by using the

add_shutdown_callback

method, which triggers when the session ends. Here's a simple example for you:

from datetime import datetime
import json

def entrypoint(ctx):
    session = AgentSession()

    async def write_transcript():
        current_date = datetime.now().strftime("%Y%m%d_%H%M%S")
        filename = f"/tmp/transcript_{ctx.room.name}_{current_date}.json"

        with open(filename, 'w') as f:
            json.dump(session.history.to_dict(), f, indent=2)

        print(f"Transcript for {ctx.room.name} saved to {filename}")

    ctx.add_shutdown_callback(write_transcript)

2. Real-Time Transcript Access If you're interested in real-time updates, you can listen to events like

user_input_transcribed

. Here's how you might do that:

@session.on("user_input_transcribed")
def on_transcript(transcript):
    if transcript.is_final:
        print(f"Final transcript: {transcript.transcript}")
        # You can save this to a database or process it further here

3. Structure of

session.history

Just so you know, the

session.history

object includes: • role: Who's speaking (user or assistant). • text_content: The actual message text. • timestamp: When the message happened (if available). 4. Additional Options • You can enable or disable transcription using

RoomOutputOptions

. • For word-level synchronization, try enabling TTS-aligned transcription with

use_tts_aligned_transcript=True

. If you'd like, I’m here to help with more details or even a complete working example. Just let me know! Sources: Session recording and transcripts | LiveKit Docs | complex-agents/teleprompter/README.md | Text and transcriptions | LiveKit Docs