Can I access to the room from the LLM step? I have the chat_context, but it does not seem to have access to the room. Using the agent library I want to trigger an RPC method there.

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

I have an instance of LLM: class MyLLM(LLM): … How can I do it there?

Hey there! Let me break down how you can work with rooms and RPC methods in your custom LLM class. Think of it like setting up a friendly communication system between your code and other participants. 😊

class MyLLM(LLM):
    def __init__(self, room):
        self.room = room
        self._register_rpc_methods()

    def _register_rpc_methods(self):
        # Here's where we set up our communication channels
        self.room.local_participant.register_rpc_method(
            "agent.flipFlashCard",
            self.handle_flip_flash_card
        )

    async def handle_flip_flash_card(self, rpc_data):
        # We'll handle this RPC call with a friendly response
        return "Card flipped successfully"

    async def trigger_client_method(self, client_identity, payload):
        # And here's how we can reach out to other clients
        await self.room.local_participant.perform_rpc(
            destination_identity=client_identity,
            method="client.eou_update",
            payload=json.dumps(payload)
        )

### Here's the Scoop: • When you create your LLM class, just pass in the

room

object - it's like introducing a new friend to the group! • Use

room.local_participant.register_rpc_method

to set up methods that other participants can call. It's like creating open communication channels. • And with

room.local_participant.perform_rpc

, you can easily reach out and call methods on other participants. This approach lets your custom LLM class be super flexible - you can both listen for and initiate conversations with different parts of your application. Pretty cool, right? 🚀 Sources: examples/rpc/rpc-demo.ts | examples/rpc/index.ts | examples/rpc.py | complex-agents/turn-taking/agent.py

But that’s defining the RPC, I want to invoke it

Hey there! Let me walk you through how to call an RPC method from your custom LLM class. It's actually pretty straightforward – think of it like sending a quick message to a friend in the room. 😊

class MyLLM(LLM):
    def __init__(self, room):
        self.room = room

    async def invoke_rpc(self, destination_identity, method, payload):
        try:
            response = await self.room.local_participant.perform_rpc(
                destination_identity=destination_identity,
                method=method,
                payload=json.dumps(payload)
            )
            print(f"RPC response: {response}")
            return response
        except Exception as e:
            print(f"RPC call failed: {e}")
            raise

### Key Parameters: • destination_identity: Who you want to send the message to. • method: The specific action you want them to do (like saying "greet"). • payload: The actual message you're sending (usually in JSON). ### Example Usage:

my_llm = MyLLM(room)
await my_llm.invoke_rpc(
    destination_identity="participant-identity",
    method="greet",
    payload={"message": "Hello from RPC!"}
)

### Quick Tips: • Make sure the destination participant has the method you're trying to call registered first. • Keep your message under 15KiB – we don't want any data overflow! • By default, the system will wait 10 seconds for a response, but you can adjust this if needed. And there you have it! This method lets you easily communicate and trigger actions with other participants in the room. Pretty cool, right? 🚀 Sources: RPC | LiveKit Docs | Function calling with Voice Agents | LiveKit Docs | Runtime/Scripts/Rpc.cs | Remote method calls (RPC) | LiveKit Docs | examples/rpc/index.ts