Function Calling and Tool Use in LLM APIs

Function calling (or tool use) lets a language model request that your code run a function, such as looking up an order or checking the weather.

How It Works

1. You describe available tools with names, descriptions and JSON Schema parameters.
2. The model decides whether a tool would help and returns a structured tool call with arguments.
3. Your code executes the function.
4. You send the result back, and the model uses it to continue.

Best Practices

• Write clear tool descriptions, including when to use each tool.
• Use precise parameter schemas with enums and descriptions.
• Validate arguments before executing.
• Return concise, informative results and errors.
• Require confirmation before actions with side effects.

Controlling Tool Use

APIs often let you let the model decide, force a specific tool, or disable tools.

Parallel Calls

Models may request several tool calls at once; execute them concurrently when independent.

Structured Output

Function calling is also a reliable way to get structured data, even if the "tool" just records the output.

Security

Treat tool arguments as untrusted input, and limit what tools can do.