https://livekit.io logo
Join Slack
Powered by
can you give me an example of function calling wit...
# ask-ai
m
can you give me an example of function calling with any anthropic LLM model?
t
Thanks for reaching out, Kush. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
Here's a friendly example of how to use function calling with an Anthropic LLM model: Step 1: Define the Function Context
Copy code
from livekit.agents.llm import FunctionContext, ai_callable
import aiohttp

class WeatherFunctionContext(FunctionContext):
    @ai_callable(description="Get the current weather in a given location")
    async def get_weather(
        self,
        location: Annotated[str, TypeInfo(description="The city and state, e.g., San Francisco, CA")],
    ) -> str:
        """Get the current weather for a given location."""
        url = f"<https://wttr.in/{location}?format=j1>"
        async with aiohttp.ClientSession() as session:
            async with session.get(url) as response:
                if response.status == 200:
                    weather_data = await response.json()
                    return f"The weather in {location} is {weather_data['current_condition'][0]['weatherDesc']}."
                raise Exception(f"Failed to get weather for {location}")
Step 2: Create the Agent with Anthropic LLM
Copy code
from livekit.agents import Agent, AgentSession
from livekit.plugins.anthropic import LLM

class WeatherAgent(Agent):
    def __init__(self) -> None:
        super().__init__(
            instructions="You are a weather assistant. Use the get_weather function to provide accurate weather information.",
            llm=LLM(),
        )

    async def on_enter(self):
        self.session.generate_reply()

async def entrypoint(ctx: JobContext):
    session = AgentSession()

    await session.start(
        agent=WeatherAgent(),
        room=ctx.room
    )
Step 3: Use the Function in the LLM When someone asks about the weather, the LLM will naturally pick up on the need to call the 
get_weather
function and share the result with them. Step 4: Handle Function Results The LLM then uses the info from the function call to craft a helpful response, like this:
Copy code
The weather in New York is partly cloudy.
This example shows you how to smoothly integrate function calling with an Anthropic LLM in LiveKit, making your agents even smarter and more useful. Sources: livekit-agents/livekit/agents/voice/run_result.py | examples/pydantic_models_to_grammar_examples.py
m
can you give me an example in 0.x agent?
t
Here’s a friendly example showing how to use function calling with an Anthropic LLM model in LiveKit's 0.x agent:
Copy code
from livekit.agents import llm
from livekit.agents.pipeline import VoicePipelineAgent
from livekit.plugins.anthropic import LLM

# Define a function context with the tool
class WeatherFnc(llm.FunctionContext):
    @llm.ai_callable()
    async def get_weather(
        self,
        location: Annotated[str, llm.TypeInfo(description="Location to get weather for")]
    ):
        """Get the current weather for a given location."""
        # Simulate an API call to get weather data
        url = f"<WEATHER_API_URL>/{location}?format=%C+%t"
        async with aiohttp.ClientSession() as session:
            async with session.get(url) as response:
                if response.status == 200:
                    weather_data = await response.text()
                    return f"The weather in {location} is {weather_data}."
                else:
                    raise f"Failed to get weather data, status code: {response.status}"

# Initialize the Anthropic LLM
anthropic_llm = LLM(model="claude-3-sonnet")

# Create the function context
weather_ctx = WeatherFnc()

# Initialize the agent with the LLM and function context
agent = VoicePipelineAgent(
    llm=anthropic_llm,
    fnc_ctx=weather_ctx,
    # Add other necessary components like STT, TTS, VAD, etc.
)
This little example shows you how the 
get_weather
function is wrapped with 
ai_callable()
, so the LLM knows it can use it. When someone asks about the weather, the LLM calls this function to fetch the latest weather info and includes it right in the reply. If you want me to walk you through any part of this or need more examples, just let me know—I’m here to help!
Open in Slack
PreviousNext
Powered by