[DZH166]

Problem

FunctionTool._get_mandatory_args() decides whether a parameter has a default with ==:

if param.default == inspect.Parameter.empty:
    mandatory.append(param.name)

== runs the default's own __eq__. A default with elementwise equality - a numpy.ndarray, a pandas.Series, the natural default of a numeric tool parameter - returns a non-boolean:

• a truthy result marks the optional parameter as mandatory, so run_async() answers every call that omits it with "mandatory input parameters are not present" and the default is never used;
• an ambiguous result (an array with more than one element) makes _get_mandatory_args() raise ValueError: The truth value of an array with more than one element is ambiguous on every call.

_get_mandatory_args() runs on every FunctionTool.run_async() path. The comment directly above the code already states the test in identity form.

BEFORE  mandatory parameters reported: ValueError: The truth value of an array with more than one element is ambiguous.
        (scalar-eq shape: ['text', 'weights'] -> run_async fails with "mandatory input parameters are not present: weights")
AFTER   mandatory parameters reported: ['text']
        run_async result: {'text': 'hello', 'weights_is_default': True}

Fix

if param.default is inspect.Parameter.empty:

Tests

tests/unittests/tools/test_function_tool.py::test_optional_param_with_elementwise_equality_default_is_optional uses a dependency-free stand-in whose __eq__ returns an object with ambiguous truthiness.

pytest tests/unittests/tools/test_function_tool.py -q   # 121 passed (1 failed with the test only, on the unpatched code)

Related suites (test_function_tool*.py, test_authenticated_function_tool.py, utils/test_callable_utils.py): 232 passed, 1 skipped.

Note: the look-alike param.default != inspect.Parameter.empty in tools/_automatic_function_calling_util.py is intentionally not touched here - that path is the legacy function-declaration builder, while _get_mandatory_args() runs on every tool call.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[mangeshraut712]

The identity check in _get_mandatory_args is the right fix for elementwise-equality defaults.

Same hazard remains in src/google/adk/tools/_automatic_function_calling_util.py (~line 84) when building the pydantic field default: param.default != inspect.Parameter.empty still calls the default's __eq__/__ne__. For an elementwise numpy/pandas-style default that returns a non-bool, that ternary can raise the same ambiguous-truth error during schema generation. Please switch it to is / is not, matching this change and the other identity checks in _function_parameter_parse_util.py.

[mangeshraut712]

Using is here is correct: with a stand-in default whose __eq__ returns a non-bool (as in the new test), == inspect.Parameter.empty raises on truth-testing, while is keeps the parameter optional.

Please apply the same identity check in _automatic_function_calling_util.py where param.default != inspect.Parameter.empty is still used for the pydantic Field default.