https://livekit.io logo
Join Slack
Powered by
In my agent class BaseAgent(Agent) I have a number...
# ask-ai
c
In my agent class BaseAgent(Agent) I have a number of tools with annotation @function_tool() If I want to disable some of those tools during initialization of the BaseAgent, how can I best do that?
t
Thanks for reaching out, Vahid. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
To disable specific tools in your 
BaseAgent
class during initialization, you can use one of the following methods: 1. Override Tools in the ConstructorYou can filter out unwanted tools by overriding the 
tools
parameter in your agent's constructor:
Copy code
from livekit.agents import Agent, function_tool
from livekit.agents.llm import find_function_tools

class BaseAgent(Agent):
    def __init__(self, **kwargs):
        all_tools = find_function_tools(self)
        disabled_tool_names = {"unwanted_tool", "another_disabled_tool"}
        filtered_tools = [
            tool for tool in all_tools 
            if getattr(tool, '__name__', '') not in disabled_tool_names
        ]
        super().__init__(tools=filtered_tools, **kwargs)

    @function_tool()
    async def useful_tool(self, context, param: str):
        return "result"

    @function_tool()
    async def unwanted_tool(self, context, param: str):
        return "disabled"
2. Conditional Tool RegistrationYou can register tools conditionally based on configuration:
Copy code
class BaseAgent(Agent):
    def __init__(self, disabled_tools=None, **kwargs):
        disabled_tools = disabled_tools or set()
        tools = []
        if "weather_tool" not in disabled_tools:
            tools.append(self.weather_tool)
        if "calendar_tool" not in disabled_tools:
            tools.append(self.calendar_tool)
        super().__init__(tools=tools, **kwargs)

    @function_tool()
    async def weather_tool(self, context, location: str):
        return {"weather": "sunny"}

    @function_tool()
    async def calendar_tool(self, context, date: str):
        return {"events": []}
3. Use 
update_tools
After InitializationYou can modify the tool set dynamically after the agent is created:
Copy code
# Create agent normally
agent = BaseAgent()

# Filter out unwanted tools
current_tools = agent.tools
disabled_tool_names = {"unwanted_tool"}
filtered_tools = [
    tool for tool in current_tools 
    if getattr(tool, '__name__', '') not in disabled_tool_names
]

# Update agent with filtered tools
await agent.update_tools(filtered_tools)
4. Tool Name FilteringFor more advanced filtering, you can check tool names dynamically:
Copy code
from livekit.agents.llm import is_function_tool, get_function_info

class BaseAgent(Agent):
    def __init__(self, disabled_tools=None, **kwargs):
        disabled_tools = disabled_tools or set()
        all_tools = find_function_tools(self)
        filtered_tools = [
            tool for tool in all_tools
            if get_function_info(tool).name not in disabled_tools
        ]
        super().__init__(tools=filtered_tools, **kwargs)
Choose the method that best fits your use case. Constructor filtering is ideal for permanent disabling, while 
update_tools
allows runtime control. Sources: Tool definition and use | LiveKit Docs | livekit-agents/livekit/agents/llm/tool_context.py
f
gettign error find_function_tools not found.
t
The error you're seeing suggests that 
find_function_tools
might not be available in your current LiveKit setup. This function is usually used to automatically discover tools in an agent class. If it’s missing, no worries—you can manually define the tools instead of relying on auto-discovery. Here’s a simple example to guide you:
Copy code
from livekit.agents import Agent, function_tool

class BaseAgent(Agent):
    def __init__(self, **kwargs):
        # Manually list the tools you want to include
        tools = [
            self.useful_tool,
            # Add other tools here
        ]
        super().__init__(tools=tools, **kwargs)

    @function_tool()
    async def useful_tool(self, context, param: str):
        return "result"
If auto-discovery is something you really want to use, it might help to check if your LiveKit version supports it or if there are any updates available. Please feel free to reach out if you’d like more help—I’m here to support you!
Open in Slack
PreviousNext
Powered by