How can I persist chat messages? I know how to load them, but not how to save them after receiving them.

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

Sure.

Hey there! Could you share which storage solution you're using (like local storage, a database, or Firebase) or which programming language you're working with? That way, I can whip up a detailed example that fits perfectly with what you need.

Let's say an in-memory database for simplicity.

Using an in-memory database like Redis is a fantastic choice when you want something simple and fast. Let me walk you through how you can save chat messages with Redis: 1. Set Up Redis: First, install and run a Redis server. You can use libraries like

ioredis

or

redis

in Node.js to connect and interact with Redis easily. 2. Save Messages: Whenever a new message comes in, just store it in Redis using a unique key—something like the room ID or user ID works great. 3. Retrieve Messages: When you need to get messages, like when someone joins a chat room, you can fetch them straight from Redis. Here’s a simple example in Node.js to get you started:

const Redis = require('ioredis');
const redis = new Redis();

// Save a message
function saveMessage(roomId, message) {
  const key = `chat:${roomId}`;
  redis.rpush(key, JSON.stringify(message));
}

// Retrieve messages
async function getMessages(roomId) {
  const key = `chat:${roomId}`;
  const messages = await redis.lrange(key, 0, -1);
  return messages.map(msg => JSON.parse(msg));
}

// Example usage
saveMessage('room1', { sender: 'user1', content: 'Hello!', timestamp: Date.now() });
getMessages('room1').then(messages => console.log(messages));

If you want, I’m here to help you dive deeper or troubleshoot anything along the way. Just let me know!

Yes, but how do you save them in an AgentSession?

To save chat messages in an

AgentSession

, there are a few different ways you can go about it, depending on what fits your needs best: 1. Using a Key-Value Store: You could save messages in a key-value store like Redis or another in-memory database. For instance, if you’re working with TypeScript,

kv.hmset

and

kv.zadd

are handy for storing messages and linking them to a user or session. 2. Event Listeners: If you’re using Python, you can tap into events like

user_speech_committed

or

agent_stopped_speaking

to catch and log messages as they happen. 3. ChatContext: Another option is to add messages to a

ChatContext

object, which keeps track of the conversation history during the session—great for temporary storage. If you want, I can share some detailed examples for any of these methods. Or if you have a particular programming language or storage option in mind, just let me know! Sources: Sending text | LiveKit Docs | Sending text | LiveKit Docs | client/app/actions.ts | examples/voice_agents/realtime_load_chat_history.py | VoicePipelineAgent | LiveKit Docs

The VoicePipelineAgent is from v0.x, I'm on v1.x

In LiveKit v1.x, you can save chat messages in an

AgentSession

by setting up a custom solution. Here are a couple of ways you might do this: 1. Using ChatContext: You can load chat history into the session with

ChatContext

. Here’s a quick example:

chat_ctx = llm.ChatContext.empty()
for item in chat_history:
    chat_ctx.add_message(role=item["role"], content=item["content"])

session = AgentSession()
agent = Agent(
    instructions="You are a helpful assistant.",
    llm=openai.realtime.RealtimeModel(modalities=["text"]),
    chat_ctx=chat_ctx,
)
await session.start(agent=agent, room=ctx.room)

1. Save Chat History to a File: Another way is to use the

add_shutdown_callback

method to save the chat history when the session ends:

def save_chat_history(session: AgentSession):
    history = session.history
    chat_history = [
        {"role": msg.role.name, "content": msg.text_content}
        for msg in history if isinstance(msg, llm.ChatMessage)
    ]
    with open("chat_history.json", "w") as f:
        json.dump(chat_history, f, indent=2)

session.add_shutdown_callback(save_chat_history)

If you want, I’d be happy to walk you through these methods in more detail or help out with any specific part of the implementation. Just let me know! Sources: Sending text | LiveKit Docs | ChatContext | LiveKit Agents | Session recording and transcripts | LiveKit Docs

i want to save the message as soon as I receive them

Hey there! To save messages right as they come in with LiveKit v1.x, you can set up event listeners that catch and store those messages in real-time. Here’s a simple way to do it: 1. Listen for Incoming Messages: Just use the right event listener to know when a message pops up. 2. Save the Message: Then, save that message right away in whatever storage you prefer—whether that’s a database, a file, or even in-memory. Here’s a quick example in Python to get you started:

@agent.on("user_speech_committed")
def on_user_message(msg: llm.ChatMessage):
    # Save the message immediately
    save_message({
        "role": "user",
        "content": msg.text_content,
        "timestamp": datetime.now().isoformat()
    })

@agent.on("agent_stopped_speaking")
def on_agent_message(msg: llm.ChatMessage):
    # Save the message immediately
    save_message({
        "role": "assistant",
        "content": msg.text_content,
        "timestamp": datetime.now().isoformat()
    })

def save_message(message):
    # Example: Save to a file or database
    with open("chat_history.json", "a") as f:
        f.write(json.dumps(message) + "\n")

If you want, I’d be happy to help you tweak this for your specific storage setup or programming language. Just let me know!

Does it work with chat message received on

lk.chat

topic?

Absolutely! You can save chat messages received on the

lk.chat

topic. Just use the

setupDataMessageHandler

function to listen for incoming messages and then store them wherever works best for you. Here’s a simple example in JavaScript to get you started:

const { messageObservable } = setupDataMessageHandler(room, 'lk.chat');

messageObservable.subscribe((message) => {
  // Save the message to your storage solution
  saveMessageToDatabase({
    id: message.id,
    sender: message.from?.identity,
    content: message.message,
    timestamp: new Date(message.timestamp)
  });
});

If you want, I’d be happy to help you tweak this for a specific storage option or programming language. Just let me know! Sources: Sending text | LiveKit Docs

Agent in v1.x does not have

on

I don't have the answer you're looking for. You could also try asking your question: • in one of the other Slack channels or • to https://deepwiki.com/livekit/livekit_composite which is trained on all LiveKit source code If you find the answer, please post it here to help others!

There isn't any info about it