Function Schema

Why Function Schema Is Needed

Letting an Agent call a tool isn't about "writing code for the model," but about "telling the model what this tool does and how to use it."

The model doesn't read your code directly — it relies on the description you provide to determine:

• When this tool should be used
• What parameters to pass
• What format parameters should be in

Function Schema is this "user manual" — its quality directly determines whether Tool Calling is correct.

What Is Function Schema

One-line definition: Function Schema is a structured description of a tool's capabilities, serving as the "parameter contract" when the model calls a tool.

A typical Schema looks like this:

{
  "name": "get_weather",
  "description": "Query current weather for a specified city",
  "parameters": {
    "type": "object",
    "properties": {
      "city": {
        "type": "string",
        "description": "City name, e.g., Beijing, Shanghai"
      },
      "unit": {
        "type": "string",
        "enum": ["celsius", "fahrenheit"],
        "description": "Temperature unit"
      }
    },
    "required": ["city"]
  }
}

When the model sees this Schema, it knows:

• The tool is called get_weather and is for checking weather
• The required parameter is city
• unit can only be celsius or fahrenheit

How to Do It: How to Write Good Schema

Descriptions should clarify use cases: Don't just write "fetch data" — explain clearly when to use it.

Parameter names should be specific: city is better than input, invoice_id is better than id.

Only set truly necessary fields as required: otherwise the model is more likely to fail due to missing parameters.

Use enums when possible: clearly constrain finite options with enum.

Don't nest too deeply: deeper nested structures mean more errors from the model.

Remember this: Function Schema is primarily written for the model to read, not the program — its core value is helping the model accurately understand "when to use it, and how."

Related terms: Tool Calling