Function Calling in LLMs

What is Function Calling? Function calling allows LLMs to output structured requests to call external functions/tools, enabling them to take actions and access real-time information.

How It Works

User Query: "What is the weather in Tokyo?"
     |
     v
LLM: {"function": "get_weather", "arguments": {"location": "Tokyo"}}
     |
     v
System: Execute function with arguments
     |
     v
Function Result: {"temp": 22, "condition": "sunny"}
     |
     v
LLM: "The weather in Tokyo is 22C and sunny."

OpenAI Function Calling

Define Functions

tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "Get current weather for a location",
        "parameters": {
            "type": "object",
            "properties": {
                "location": {"type": "string", "description": "City name"},
                "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]}
            },
            "required": ["location"]
        }
    }
}]

Call API

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Weather in Tokyo?"}],
    tools=tools,
    tool_choice="auto"
)

# Check if model wants to call a function
if response.choices[0].message.tool_calls:
    tool_call = response.choices[0].message.tool_calls[0]
    function_name = tool_call.function.name
    arguments = json.loads(tool_call.function.arguments)

    # Execute function
    result = execute_function(function_name, arguments)

    # Send result back to model for final response
    messages.append(response.choices[0].message)
    messages.append({
        "role": "tool",
        "tool_call_id": tool_call.id,
        "content": json.dumps(result)
    })

    final = client.chat.completions.create(model="gpt-4o", messages=messages)

Common Function Types

Best Practices

• Clear, specific function descriptions
• Validate function arguments before execution
• Handle function errors gracefully
• Limit number of available functions (reduce confusion)
• Test with adversarial inputs

Open Source Alternatives