IronaAI supports function calling, allowing language models to interact with external tools, APIs, or custom functions. This feature enhances the model’s capabilities by enabling actions such as retrieving real-time data, performing computations, or accessing databases.

​
When to use function calling

Use function calling when you need the model to:

  • Retrieve real-time or external data (e.g., weather info, stock prices)
  • Perform computations not natively supported by the model
  • Interact with APIs or databases
  • Execute custom functions in your application

This is ideal for building interactive agents, automating workflows, or integrating AI into systems. Defining tools Tools can be defined in multiple ways:

  • As dictionaries: Specify the tool’s name, description, and parameters.

  • As LangChain tools: Integrate tools from the LangChain library.

​
Example: Defining a tool as a dictionary

tools = [
    {
        "type": "function",
        "function": {
            "name": "get_current_weather",
            "description": "Get the current weather in a given location",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {"type": "string", "description": "The city, e.g., San Francisco"},
                    "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]}
                },
                "required": ["location"]
            }
        }
    }
]

​
Example: Using a LangChain tool

from langchain_core.tools import tool

@tool
def multiply(a: int, b: int) -> int:
    """Multiply two integers."""
    return a * b

​
Make a function call using the client

# Bind the tool to the client
client = IronaAI()
completion_with_tools = client.bind_tools([multiply])
response = client.completion_with_tools(
    messages=[{"role": "user", "content": "What's the weather like in San Francisco?"}],
    tools=tools,
    tool_choice="auto"
)

The response includes a tool_calls attribute if the model opts to call a tool.

​
Considerations and limitations

  • Model Support: Verify support using client.supports_function_calling(model).
  • Tool Definitions: Ensure clarity to avoid incorrect calls.
  • Execution: Handle tool execution and errors in your code.