fix: stop plain langchain Tools from asking the model for config - #8117

Link to Issue or Description of Change

1. Link to an existing issue (if applicable):

• Closes: LangchainTool asks the model for config when it wraps a plain langchain Tool #8035
• Related: fix(integrations): use the wrapped function of a plain langchain Tool #8036 (earlier attempt, closed because its author couldn't sign the CLA; this PR is an independent implementation that also covers the review feedback there)

Problem:
LangchainTool.__init__ only takes tool.func / tool.coroutine for a StructuredTool. Tool and StructuredTool are siblings (both subclass BaseTool), so a plain langchain_core.tools.Tool skips that branch and falls back to Tool._run(self, *args, config: RunnableConfig, run_manager=None, **kwargs). FunctionTool builds the declaration from that signature: run_manager is ignored and *args / **kwargs are dropped, leaving config as the only required parameter. The model is asked for config, never for the wrapped function's real arguments, and every call fails the mandatory-argument check.

Solution:
Give a plain Tool the same resolution the StructuredTool branch already uses:

1. use tool.func;
2. otherwise use tool.coroutine (async-only tools);
3. otherwise, if the subclass overrides _run, keep using _run (unchanged behavior);
4. otherwise raise ValueError, as StructuredTool already does.

The change is the isinstance check plus the "_run not overridden" check, which now compares against both StructuredTool._run and Tool._run.

One intentional behavior change: Tool(func=None) with no coroutine and no _run override used to construct but could never run (it declared config and every call failed). It now raises at construction, matching StructuredTool.

Testing Plan

Unit Tests:

• I have added or updated unit tests for my change.
• All unit tests pass locally.

New tests in tests/unittests/integrations/langchain/test_langchain_tool.py. Following the review on #8036, they assert the declared parameters the model sees, not only the run_async result:

Without the fix (src change stashed), the new tests fail as expected and the guard passes:

FAILED test_plain_tool_declares_func_params_not_config - assert ['config'] == ['text']
FAILED test_plain_tool_run_async_calls_func - mandatory input parameters are not present: config
FAILED test_plain_tool_with_only_coroutine_works - assert ['config'] == ['text']
FAILED test_plain_tool_without_func_or_coroutine_raises - DID NOT RAISE ValueError
4 failed, 11 passed

With the fix:

$ pytest tests/unittests/integrations/langchain -q
15 passed

$ pytest tests/unittests/integrations tests/unittests/tools -q
5485 passed, 40 subtests passed

pre-commit run on both files passes (ruff, isort, pyink, addlicense, codespell, ADK compliance checks).

Manual End-to-End (E2E) Tests:

The issue's repro script (no model or API key needed), Python 3.11.18, langchain-core 1.8.0:

Before:

Tool func: Tool._run
Tool declares: ['config'] required: ['config']
Tool run_async: {'error': 'Invoking `shout()` failed as the following mandatory input parameters are not present:\nconfig ...'}
StructuredTool declares: ['text'] required: ['text']
StructuredTool run_async: HI

After:

Tool func: shout
Tool declares: ['text'] required: ['text']
Tool run_async: HI
StructuredTool declares: ['text'] required: ['text']
StructuredTool run_async: HI

Checklist

• I have read the CONTRIBUTING.md document.
• I have performed a self-review of my own code.
• I have commented my code, particularly in hard-to-understand areas.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.
• I have manually tested my changes end-to-end.
• Any dependent changes have been merged and published in downstream modules.