Function Calling

Enable LLMs to interact with external systems in a controlled, production-ready way

Learning goals

•Understand function calling and tool use
•Learn to define function schemas
•Implement safe and effective function execution

What Is Function Calling?

Function calling allows LLMs to:

Recognize when a function should be called
Extract the correct parameters from natural language
Return a structured function call for your code to execute

The model doesn't execute functions—it tells your code what to call.

Defining Functions

Functions are defined using JSON Schema:

const functions = [{
  name: "get_weather",
  description: "Get the current weather for a location",
  parameters: {
    type: "object",
    properties: {
      location: {
        type: "string",
        description: "City and state, e.g., 'San Francisco, CA'"
      },
      unit: {
        type: "string",
        enum: ["celsius", "fahrenheit"],
        description: "Temperature unit"
      }
    },
    required: ["location"]
  }
}];

Clear descriptions help the model choose correctly.

The Function Calling Loop

The typical pattern:

User sends a message
Model decides to call a function
Your code executes the function
Send the result back to the model
Model generates a final response

User: "What's the weather in Tokyo?"
↓
Model: {function: "get_weather", arguments: {location: "Tokyo, Japan"}}
↓
Your code: fetch weather API → {temp: 22, condition: "sunny"}
↓
Model: "It's currently 22°C and sunny in Tokyo."

Common mistakes

×Poor function descriptions—the model relies on descriptions to choose functions

×Missing parameter descriptions—leads to incorrect parameter extraction

×Not handling function errors—always have fallback behavior for failed calls

×Trusting function calls blindly—validate parameters before executing

Key takeaways

+Function calling bridges natural language and structured API calls

+Clear function and parameter descriptions are critical for accuracy

+Always validate parameters before executing functions

+The model suggests calls—your code executes them safely

Defining Functions

Functions are defined using JSON Schema:

const functions = [{
  name: "get_weather",
  description: "Get the current weather for a location",
  parameters: {
    type: "object",
    properties: {
      location: {
        type: "string",
        description: "City and state, e.g., 'San Francisco, CA'"
      },
      unit: {
        type: "string",
        enum: ["celsius", "fahrenheit"],
        description: "Temperature unit"
      }
    },
    required: ["location"]
  }
}];

Clear descriptions help the model choose correctly.

The Function Calling Loop

The typical pattern:

User sends a message
Model decides to call a function
Your code executes the function
Send the result back to the model
Model generates a final response

User: "What's the weather in Tokyo?"
↓
Model: {function: "get_weather", arguments: {location: "Tokyo, Japan"}}
↓
Your code: fetch weather API → {temp: 22, condition: "sunny"}
↓
Model: "It's currently 22°C and sunny in Tokyo."