fix(langchain-openai): fix operator precedence in reasoning initialization

[wookingwoo]

Fixes #9663

Description

This PR fixes a bug in the ChatOpenAI constructor where the reasoning parameter was incorrectly overridden due to an operator precedence issue.

Problem

The original initialization logic was:

this.reasoning =
    fields?.reasoning ?? fields?.reasoningEffort
        ? { effort: fields.reasoningEffort }
        : undefined;

When fields.reasoning is provided, the expression fields?.reasoning ?? fields?.reasoningEffort evaluates to the reasoning object (truthy). Consequently, the ternary operator executes the true branch: { effort: fields.reasoningEffort }.
However, since the user provided reasoning, fields.reasoningEffort is typically undefined. This results in the reasoning field being set to { effort: undefined }, effectively discarding the user's input.

Solution

I have added parentheses to correctly group the fallback logic. The code now only checks reasoningEffort if reasoning is null or undefined.

this.reasoning=
fields?.reasoning??
  (fields?.reasoningEffort
? {effort:fields.reasoningEffort }
:undefined);

Related Issues

• Fixes [v0.3] ChatOpenAI constructor overrides 'reasoning' option due to incorrect operator precedence #9663
• Related to OpenAI Responses API documentation: https://platform.openai.com/docs/guides/reasoning

Additional Context

This is a one-line fix in the v0.3 branch that restores critical functionality for GPT-5 and o-series models. The fix maintains full backward compatibility - existing code using reasoningEffort continues to work as expected.

• Contributor profile: https://www.linkedin.com/in/ronny-woo/

[changeset-bot]

⚠️ No Changeset found

Latest commit: 490bf47

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[christian-bromann]

Can we add a test for this so we don't regress?

[wookingwoo]

Can we add a test for this so we don't regress?

Great suggestion! I've added tests for this - see the "reasoning field initialization" test suite.

We can run just these tests with:

cd libs/langchain-openai
yarn test chat_models.test.ts -t "reasoning field initialization"

[wookingwoo]

@christian-bromann Tests are added and the fix is ready.
I’d appreciate a final review/approval when you get a chance. Thanks!