Function Calling

What is function calling?

In a plain chat, a model can only talk. Ask it the weather and it'll cobble together something from stale training data. Function calling gives it hands: when it decides it needs live info or has to perform an action, it says, in a structured format, "call this function with these arguments". Your code actually runs it, then feeds the result back.

What problem does it solve?

Turns "can talk" into "can do"
Chat alone can't book a ticket, place an order, or query a database. Function calling lets the model actually trigger real actions.
Fills the live-data gap
A model's training data has a cutoff date. Weather, stock prices, inventory — all change by the minute. Function calling lets it look things up instead of guessing.

How it works, roughly

Step one: declare the tools
You tell the model "here are the functions I have", with each one's purpose and parameters spelled out.
Step two: the model decides to call
After the user's message, the model decides "this needs a tool" and returns a structured request — a function name and arguments, not prose.
Step three: your code runs it
Your program executes the function and sends the result back. The model then writes the final answer from that result.

Why it's the foundation of agents

Without function calling, an AI is all talk. With it, the model finally reaches the real world — reading data, writing data, triggering workflows all start here.

Bottom line: function calling gives a model hands, turning it from something that only talks into something that can actually do things.