How do you send and receive messages from the livekit agent when text-only mode is enabled?

Thanks for reaching out, Matin. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

Why does it seem like I still have to wait for the agent to finish thinking? Is there any way to stream the chunks? Or is it generating it at a conversational pace? Also should I set sync_transcript to false, or just not specify it at all?

Hey there! I totally get your concerns about waiting for the agent to finish thinking and streaming chunks. Let me walk you through some helpful details and settings that can make things smoother for you: 1. Streaming Chunks: You can actually stream text chunks from the agent by setting up a text stream handler. This way, you can process and show responses in real-time as they come in:

// JavaScript Example
room.registerTextStreamHandler('lk.transcription', async (reader, participantInfo) => {
    const message = await reader.readAll();
    if (reader.info.attributes['lk.transcribed_track_id']) {
        console.log(`New transcription from ${participantInfo.identity}: ${message}`);
    } else {
        console.log(`New message from ${participantInfo.identity}: ${message}`);
    }
});

And on the backend, you can do something similar to handle chunks as they arrive:

async def on_text_received(reader, participant_identity):
    async for chunk in reader:
        print(f"Received chunk: {chunk}")

2. sync_transcription Setting: This

sync_transcription

option controls whether transcriptions sync up with the audio playback: • By default: Transcriptions match the audio, showing text word by word as it’s spoken. • Want it faster? Set

sync_transcription=False

to get transcriptions right away, without waiting for sync. Here’s a quick example:

await session.start(
    agent=MyAgent(),
    room=ctx.room,
    room_output_options=RoomOutputOptions(sync_transcription=False),
)

3. Key Takeaways: • Make sure you have a text stream handler registered for the

lk.transcription

topic to stream responses live. • Set

sync_transcription=False

if you want to skip delays caused by syncing with audio playback. • If you don’t set

sync_transcription

, it defaults to synchronized transcription. I hope this helps you get faster response streaming and cuts down on waiting! If you have any more questions or want me to clarify anything, just let me know—I'm here to help! Sources: Text and transcriptions | LiveKit Docs | examples/other/transcription/translator.py | Text and transcriptions | LiveKit Docs

Can you please create a python example please?

Here's a friendly Python example to help you set up a LiveKit agent in text-only mode, stream text chunks as they come in real-time, and manage the

sync_transcription

setting: 1. Agent Configuration: Let's configure the agent to work in text-only mode and turn off audio:

import asyncio
from livekit.agents import (
    AgentSession,
    RoomInputOptions,
    RoomOutputOptions,
)
from livekit.plugins import openai

class MyAgent:
    def __init__(self):
        self.instructions = "You are a helpful assistant."

async def entrypoint(ctx):
    session = AgentSession(
        llm=openai.LLM(model="gpt-4o-mini"),
    )
    await session.start(
        agent=MyAgent(),
        room=ctx.room,
        room_input_options=RoomInputOptions(text_enabled=True, audio_enabled=False),
        room_output_options=RoomOutputOptions(transcription_enabled=True, audio_enabled=False, sync_transcription=False),
    )

2. Streaming Text Chunks: Here’s how you can register a text stream handler to catch and display text chunks as they arrive:

from livekit import rtc
from livekit.agents.types import (
    ATTRIBUTE_TRANSCRIPTION_FINAL,
    ATTRIBUTE_TRANSCRIPTION_SEGMENT_ID,
    ATTRIBUTE_TRANSCRIPTION_TRACK_ID,
    TOPIC_TRANSCRIPTION,
)

class TextStreamPrinter:
    def __init__(self):
        self._text_chunk_queue = asyncio.Queue()

    async def on_text_received(self, reader: rtc.TextStreamReader, participant_identity: str):
        async for chunk in reader:
            print(f"Received chunk from {participant_identity}: {chunk}")

        # Check the final flag
        final = reader.info.attributes.get(ATTRIBUTE_TRANSCRIPTION_FINAL, "null")
        print(f"Final flag: {final}")

# Register the handler
text_printer = TextStreamPrinter()
room.register_text_stream_handler(
    topic=TOPIC_TRANSCRIPTION, 
    handler=text_printer.on_text_received
)

3. Key Points to Keep in Mind: • Setting

sync_transcription=False

lets you get transcriptions right away, without waiting for synchronization. • Use the

lk.transcription

topic to catch text responses from the agent. • Process those text chunks as they come in by using the handler you registered. This example walks you through setting up and using a LiveKit agent for real-time text streaming in Python. If you want me to explain anything further or share more examples, just let me know — I’m here to help!