How to integrate custom tools

Custom tools in an agent can be defined in a handful of ways, depending on what works best for your scenario. You may find that your company already has Azure Functions implemented for your agent to use, or a public OpenAPI specification gives your agent the functionality you’re looking for.

Function Calling

Function calling allows agents to execute predefined functions dynamically based on user input. This feature is ideal for scenarios where agents need to perform specific tasks, such as retrieving data or processing user queries, and can be done in code from within the agent. Your function may call out to other APIs to get additional information or initiate a program.

Example: Defining and using a function

Start by defining a function that the agent can call. For instance, here’s a fake snowfall tracking function:

PythonCopy

import json

def recent_snowfall(location: str) -> str:
    """
    Fetches recent snowfall totals for a given location.
    :param location: The city name.
    :return: Snowfall details as a JSON string.
    """
    mock_snow_data = {"Seattle": "0 inches", "Denver": "2 inches"}
    snow = mock_snow_data.get(location, "Data not available.")
    return json.dumps({"location": location, "snowfall": snow})

user_functions: Set[Callable[..., Any]] = {
    recent_snowfall,
}

Register the function with your agent using the Azure AI SDK:

PythonCopy

# Initialize agent toolset with user functions
functions = FunctionTool(user_functions)
toolset = ToolSet()
toolset.add(functions)
agent_client.enable_auto_function_calls(toolset=toolset)

# Create your agent with the toolset
agent = agent_client.create_agent(
    model="gpt-4o-mini",
    name="snowfall-agent",
    instructions="You are a weather assistant tracking snowfall. Use the provided functions to answer questions.",
    toolset=toolset
)

The agent can now call recent_snowfall dynamically when it determines that the prompt requires information that can be retrieved by the function.

Azure Functions

Azure Functions provide serverless computing capabilities for real-time processing. This integration is ideal for event-driven workflows, enabling agents to respond to triggers such as HTTP requests or queue messages.

Example: Using Azure Functions with a queue trigger

First, develop and deploy your Azure Function. In this example, imagine we have a function in our Azure subscription to fetch the snowfall for a given location.

When your Azure Function is in place, integrate add it to the agent definition as an Azure Function tool:

PythonCopy

storage_service_endpoint = "https://<your-storage>.queue.core.windows.net"

azure_function_tool = AzureFunctionTool(
    name="get_snowfall",
    description="Get snowfall information using Azure Function",
    parameters={
            "type": "object",
            "properties": {
                "location": {"type": "string", "description": "The location to check snowfall."},
            },
            "required": ["location"],
        },
    input_queue=AzureFunctionStorageQueue(
        queue_name="input",
        storage_service_endpoint=storage_service_endpoint,
    ),
    output_queue=AzureFunctionStorageQueue(
        queue_name="output",
        storage_service_endpoint=storage_service_endpoint,
    ),
)

agent = agent_client.create_agent(
    model=os.environ["MODEL_DEPLOYMENT_NAME"],
    name="azure-function-agent",
    instructions="You are a snowfall tracking agent. Use the provided Azure Function to fetch snowfall based on location.",
    tools=azure_function_tool.definitions,
)

citrix training courses malaysia