To use

function_tool

with LLM for calling custom functions and managing chat context, follow these steps: 1. Define Custom Functions: Use the

@function_tool

decorator to define your custom functions. For example:

from livekit.agents import function_tool

@function_tool
def get_current_time(timezone: str = "UTC") -> str:
    """Get the current time in the specified timezone"""
    from datetime import datetime
    import pytz
    
    tz = pytz.timezone(timezone)
    return datetime.now(tz).strftime("%Y-%m-%d %H:%M:%S %Z")

2. Use Functions with LLM: Pass your function tools to the LLM's chat method along with the chat context:

# Create chat context
chat_ctx = ChatContext()
chat_ctx.add_message(role="user", content="What time is it in New York?")

# Get response with function calling
async with llm.chat(chat_ctx=chat_ctx, tools=[get_current_time]) as stream:
    async for chunk in stream:
        # Handle function calls and response chunks
        pass

3. Automatic Chat Context Management: Function calls and their outputs are automatically added to the chat context. The system tracks: • FunctionCall: Represents the LLM's request to call a function • FunctionCallOutput: Contains the result of the function execution • ChatMessage: Regular messages from users and assistants 4. Function Call Flow: 1. LLM receives chat context with function definitions 2. LLM decides to call a function and sends a function call request 3. Your application executes the function 4. Function result is sent back to the LLM 5. LLM continues with the function result to generate the final response 5. Adding Function Messages to Chat Context: Function calls and outputs are automatically included in the chat context:

# Function calls and outputs are automatically added to chat_ctx.items
tool_messages = new_calls + new_fnc_outputs
if generate_tool_reply:
    chat_ctx.items.extend(tool_messages)

By following these steps, you can effectively integrate custom functions with LLM and manage their interactions within the chat context. Sources: LLM Integrations | livekit/livekit_composite | DeepWiki | livekit-agents/livekit/agents/llm/_provider_format/openai.py | complex-agents/personal_shopper/personal_shopper.py