fix: honor explicit thinking off on OpenAI-compatible providers

[RealKai42]

Related Issue

No linked issue. Follow-up hardening on top of the #1616 auto-injection workaround; relates to the reasoning-effort audit that also produced #1765.

Problem

On OpenAI-compatible (chat completions) providers, an explicit thinking "off" collapsed to the same internal state as "never configured": withThinking('off') stored undefined, exactly like a provider that never saw withThinking. The history-based auto reasoning_effort injection (introduced for #1616) could not tell the two apart, so:

• With think parts in history, a session whose thinking was explicitly turned Off still sent reasoning_effort: 'medium' — the UI said Off while the model kept reasoning (and billing for it).
• Switching from a reasoning model to one that rejects the field (e.g. plain gpt-4o) leaked reasoning_effort into the next request, producing a 400.
• thinkingEffort reported null for 'on', 'off', and unset alike, so request records fell back to 'off' and mislabeled an active 'on' as 'off'.

What changed

• OpenAILegacyChatProvider (both the kosong copy and the vendored agent-core-v2 copy) now stores the requested effort verbatim in _thinkingEffort; the wire encoding is derived per request: 'off'/'on' send no field, concrete efforts pass through unchanged.
• The feat(agent-core-v2): read and write the v1 session index file #1616 auto-injection now skips an explicit 'off' (still fires for 'on' and for "never configured", preserving the original workaround where it belongs).
• thinkingEffort returns the actual effort, fixing records/telemetry that previously logged 'off' for 'on'.
• Deliberately kept: no reasoning_effort: 'none' is sent — OpenAI's support for that value is model-dependent, and a blind 'none' would replace one incompatibility with another.
• Tests: 6 new cases in the existing kosong suite; a new 9-case suite for the vendored agent-core-v2 copy (previously untested); updated two registry assertions that reflected on the removed private field.

Checklist

• I have read the CONTRIBUTING document.
• I have linked a related issue, or explained the problem above.
• I have added tests that prove my feature works.
• Ran gen-changesets skill, or this PR needs no changeset.
• Ran gen-docs skill, or this PR needs no doc update.

[changeset-bot]

🦋 Changeset detected

Latest commit: f379528

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@moonshot-ai/kimi-code	Patch

Not sure what this means? Click here to learn what changesets are.

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

[pkg-pr-new]

pnpm dlx https://pkg.pr.new/@moonshot-ai/kimi-code@f379528

npx https://pkg.pr.new/@moonshot-ai/kimi-code@f379528

commit: f379528