Tool declaration: what ADK reads off your function

Hand that function to ADK and it writes a description of it. That description is the only thing a model ever sees, so it is worth looking at once.

from google.adk.tools import FunctionTool


def lookup_order(order_id: str) -> str:
    """Look up the status of an order by its id."""
    return f"Order {order_id} shipped on 3 March."


declaration = FunctionTool(func=lookup_order)._get_declaration()

print("name:       ", declaration.name)
print("description:", declaration.description)

name:        lookup_order
description: Look up the status of an order by its id.

Your function name became the name. Your docstring became the description. ADK added nothing beyond what your function provides.

And the arguments

schema = declaration.parameters_json_schema

print("parameters:", list(schema["properties"]))
print("types:     ", {k: v["type"] for k, v in schema["properties"].items()})
print("required:  ", schema["required"])

parameters: ['order_id']
types:      {'order_id': 'string'}
required:   ['order_id']

The type hint became the type. The parameter is required because it has no default value, and giving it one would make it optional.

Running this also prints a warning saying the JSON schema feature is experimental. It is ADK saying the shape of that one field may change, and it does not affect anything you are about to build.

Why the declaration matters

A model never sees your code. It sees this: a name, a sentence, and a list of arguments. When an agent calls the wrong tool, this is the thing to print first, because it is exactly what the model was working from.