how to make sure the agent always requests customer to give it a moment while it performs some actions using tool calling?

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

give me the complete code

Here’s a complete code example that shows how to politely ask the customer to hang tight while the agent is working on something:

import asyncio
from livekit.agents import function_tool, RunContext

# Define the function with appropriate delay and speech handling
@function_tool
async def perform_slow_operation(rc: RunContext):
    # Create a speech handle to let the user know we’re processing their request
    handle = await rc.session.generate_reply(
        instructions="Please wait while I process your request."
    )

    try:
        # Perform the slow operation (simulated with asyncio.sleep)
        await asyncio.sleep(5)  # Simulate a slow operation

        # Wait for the speech to finish
        await handle

        # Let the user know the operation is done
        await rc.session.generate_reply(
            instructions="The operation is complete. Here are the results."
        )

    except asyncio.CancelledError:
        # Handle if the user stops the agent
        if handle.interrupted:
            await rc.session.generate_reply(
                instructions="Operation was cancelled."
            )

Key Features: • Speech Handling: Uses

generate_reply()

to gently keep the user informed. • Background Operations: Simulates a slow operation with

asyncio.sleep()

. • Error Handling: Includes try-except blocks to gracefully manage interruptions. • Interrupt Check: Checks

handle.interrupted

to see if the user stopped the process. If you want me to walk you through any part of this or need a hand with something else, just let me know!

is it not possible to handle this via a generic system instruction so that the LLM can generate a response to customer before invoking tool?

Absolutely, you can handle this using a generic system instruction to make sure the LLM responds to the customer before calling any tools. Here’s a simple way to do it: Steps: 1. Start by using the

update_instructions

method to set system-level guidance for the LLM’s behavior. For example:

session.update_instructions(
    "First, provide a response to the user. Then, invoke any necessary tools."
)

2. Next, generate the initial response with the

tool_choice

parameter set to

"none"

so no tools are triggered right away:

session.generate_reply(
    instructions="First, provide a response to the user.",
    tool_choice="none"
)

3. After that first response, you can allow tool invocation by setting

tool_choice

to

"auto"

or

"required"

, depending on what you need:

session.generate_reply(
    tool_choice="auto"
)

Key Features: • System Instructions: The

update_instructions

method helps you guide the LLM to respond first before using any tools. • Tool Choice Parameter: This ensures tools only kick in after the initial reply, keeping things smooth and clear. If you’d like me to walk you through it more or help set it up, just let me know—I’m here to help! Sources: livekit-agents/livekit/agents/voice/agent_activity.py | Function calling with Voice Agents | LiveKit Docs | common/chat.cpp