To get function tools results in a LiveKit realtimeSession, follow these steps: 1. Define your function tools using the

@function_tool

decorator. For example:

@function_tool
async def add_numbers(self, context: RunContext, a: float, b: float):
    print(f"Adding {a} and {b}")
    return str(a + b)

1. Create an instance of the

RealtimeModel

class and update the tools using `update_tools`:

realtime_model = RealtimeModel()
tools = [add_numbers]
realtime_model.update_tools(tools)

1. Create a

RealtimeSession

instance and start the session:

session = realtime_model.session()
session.start()

1. Handle the tool output content event to retrieve function tool results by implementing

_handle_tool_output_content_event

asynchronously:

async def _handle_tool_output_content_event(self, event_data: dict) -> None:
    tool_use_id = event_data["event"]["toolUse"]["toolUseId"]
    tool_name = event_data["event"]["toolUse"]["toolName"]
    args = event_data["event"]["toolUse"]["content"]
    tool_result = await self.tools.function_tools[tool_name](**json.loads(args))
    print(f"Tool {tool_name} result: {tool_result}")

Key points: • Use

update_tools

on

RealtimeModel

to register your function tools. • Start the realtime session. • Implement and await

_handle_tool_output_content_event

on

RealtimeSession

to process results. This way, you get the function tools' output in realtime in your session. Sources: livekit-agents/livekit/agents/voice/agent_activity.py | realtime/openai-realtime-tools.py