[Yufeng He (he-yufeng)]

Fixes #9843

Problem

_matches_json_schema_type rejects 3.0 for "integer" schema types. JSON Schema "integer" means "a number without a fraction", so whole-number floats like 3.0 are valid. A FunctionTool declared with a JSON schema rejects provider arguments containing such values with TypeError: Invalid type for 'count'.

Fix

Accept float values whose is_integer() is true alongside int, while still rejecting bool and non-whole floats like 3.5.

Testing

New test test_tool_json_schema_integer_accepts_whole_number_float:

• 3.0 invokes without error (valid JSON Schema integer)
• 3.5 raises TypeError (not a valid integer)

python/packages/core/tests/core/test_tools.py  152 passed

AI Assistance

• No material AI assistance was used.
• This is an AI-assisted contribution. I reviewed, understood, and verified all submitted content and accept responsibility for it.

AI assistance details: implementation, regression tests, and this PR description. The author reviewed the final diff and ran the targeted core tool tests.

[Copilot]

🟡 Changes recommended

Accepted whole-number floats remain unconverted in integer-typed tool arguments, causing TypeError in callables that require actual integers.

1 open issue

• Normalize accepted whole-number floats to integers · New

What changed in this PR

This PR updates Python core tool validation to accept whole-number floats for JSON Schema integer parameters, addressing the validation failure reported in #9843.

Changes:

• Accepts floats such as 3.0 while continuing to reject booleans and fractional floats.
• Adds a regression test covering acceptance of 3.0 and rejection of 3.5.

File	Description
python/​packages/​core/​tests/​core/​test_tools.py	Adds coverage for whole-number and fractional float arguments.
python/​packages/​core/​agent_framework/​_tools.py	Extends the integer type check to accept whole-number floats.

_{Review effort: Balanced}

[Yufeng He (he-yufeng)]

The failing check here is test_invoke_mcp_tool_executor.py on the Windows shards (assert handler.call_count == 0, getting 1 == 0). That test lives in the declarative package and exercises MCP HTTP Authorization header binding. This PR only touches python/packages/core/agent_framework/_tools.py and python/packages/core/tests/core/test_tools.py (the JSON Schema integer / whole-number-float check), so the two are unrelated.

The same test passes on the ubuntu 3.11 through 3.14 shards and only fails on the Windows shards plus 3.15, which is consistent with an environment specific or flaky test rather than this change. A pure Python type check in the core package would fail uniformly across platforms if it were the cause. Could a maintainer re-run the failed jobs when convenient?