Is there any way to have the agent not speak if an LLM response had a tool call?

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

Can you give me a code example of how to do that?

Here’s a simple example of how you can set up the agent so it stays quiet whenever a tool call happens:

class FunctionAgent(Agent):
    """A LiveKit agent that uses MCP tools but doesn’t speak on tool calls."""

    def __init__(self):
        super().__init__(
            instructions="You are a helpful assistant.",
            stt=deepgram.STT(),
            llm=openai.LLM(model="gpt-4"),
            tts=openai.TTS(),
            vad=silero.VAD.load(),
            allow_interruptions=True
        )

    async def llm_node(self, chat_ctx, tools, model_settings):
        """Keep an eye out for tool calls and stay silent when they happen."""
        tool_call_detected = False
        async for chunk in super().llm_node(chat_ctx, tools, model_settings):
            if isinstance(chunk, ChatChunk) and chunk.delta and chunk.delta.tool_calls:
                tool_call_detected = True
                # Skip speaking if a tool call is detected
                continue
            if tool_call_detected and not chunk.delta.tool_calls:
                tool_call_detected = False
            yield chunk

Quick rundown: • The

llm_node

method watches each response chunk to spot tool calls. • We use a flag (

tool_call_detected

) to remember when a tool call happens. • When it does, the agent skips speaking that part of the response. If you want me to walk you through this or help with anything else, just let me know!

how do i import ChatChunk

Hey there! To bring

ChatChunk

into your LiveKit agent setup, here’s how you can do it: For Python:

from livekit.agents.llm import ChatChunk

For JavaScript/TypeScript:

import { ChatChunk } from '@livekit/agents-llm';

Just make sure you’ve got the right LiveKit package added to your project. If you run into any hiccups or want a hand with anything else, I’m here to help! Sources: agents/src/llm/index.ts | livekit-agents/livekit/agents/llm/__init__.py

current_speech.done() returns False even when the agent isn't speaking

Hey there! Just a quick heads-up about the

current_speech.done()

method — sometimes it might return

False

even when the agent isn’t actually speaking. Here’s why that can happen: 1. Pending Tasks: The underlying

Future

that tracks if the speech is done might still be waiting on some tasks or generation steps to finish. 2. Interrupted Speech: If the speech got interrupted, it might take a moment before it officially marks itself as done because it’s still cleaning up active tasks. 3. Unmarked Completion: The

_mark_done()

method, which updates the

done()

status once everything’s finished or interrupted, hasn’t been called yet. Here’s how you can handle this smoothly: • First, check if there’s an active speech handle before you look at `done()`:

if agent_session.current_speech is None:
    # Agent isn't speaking
    pass

• If you need to wait for the speech to finish, just do this:

await agent_session.current_speech.wait_for_playout()

• And if you want to clear any lingering tasks that might keep the speech active, try this:

if agent_session.current_speech:
    agent_session.current_speech.interrupt()

Feel free to reach out if you want me to explain anything further or help out with this! Sources: livekit-agents/livekit/agents/voice/agent_session.py