fix(models): correct kimi-k3 reasoning efforts on moonshot

[AmineAce]

Summary

Corrects an existing wrong value: moonshot/kimi-k3's reasoningEfforts was declared as ["max"] only. Moonshot's own docs confirm three values are supported.

Fix

["max"] → ["low", "high", "max"] (default: max)

Sources (3 independent, all Moonshot-owned):

• Kimi K3 quickstart — "supports low, high, and max (default max); K3 always has thinking mode enabled"
• MoonshotAI/Kimi-K3 GitHub — same enum, same default
• Moonshot reasoning_effort guide — same enum, plus migration note that K2.x thinking config should be removed when moving to K3

Also updated a stale in-code comment that said the field "currently accepts only max."

Note for reviewers

kimi-k3 mappings on other providers (Fireworks, Canopywave) declare different enums for the same underlying model — flagging for awareness, not touched in this PR since this fix is scoped to the native moonshot mapping only.

Testing

• pnpm format — only moonshot.ts changed, 2 lines.

Summary by CodeRabbit

• New Features
  • Moonshot’s Kimi K3 provider now supports low, high, and max reasoning effort levels.
  • Updated guidance reflects the provider’s native reasoning effort behavior.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 7c19165c-6ecc-49e2-9f1f-0114b97056d1

📥 Commits

Reviewing files that changed from the base of the PR and between b6835cf and 1e1b98e.

📒 Files selected for processing (1)

• packages/models/src/models/moonshot.ts

🚧 Files skipped from review as they are similar to previous changes (1)

• packages/models/src/models/moonshot.ts

---

Walkthrough

Moonshot Kimi K3 now advertises low, high, and max native reasoning effort levels. Its configuration comment describes the accepted top-level reasoning_effort values.

Changes

Moonshot Kimi K3 configuration

Layer / File(s)	Summary
Native reasoning effort metadata packages/models/src/models/moonshot.ts	Kimi K3 changes its supported reasoning efforts from max only to low, high, and max. The description now reflects native top-level reasoning_effort handling.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Possibly related PRs

• theopenco/llmgateway#3092: Modifies the same Moonshot model configuration and reasoning-effort levels.
• theopenco/llmgateway#3294: Modifies Kimi K3’s supported reasoning-effort levels.
• theopenco/llmgateway#3380: Modifies Moonshot model reasoning-effort declarations.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the correction to Kimi K3 reasoning effort support for Moonshot.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

packages/models/src/models/moonshot.ts (1)

657-660: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick win

Cover low and high in the request-shape tests.

The supplied packages/actions/src/prepare-request-body.spec.ts:1000-1007 test covers only max. Add or confirm cases for low and high that assert native reasoning_effort forwarding and thinking remains undefined.

Based on the supplied downstream test context.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@packages/models/src/models/moonshot.ts` around lines 657 - 660, Extend the
request-shape tests in prepare-request-body.spec.ts for the K3 reasoningEfforts
configuration to cover both “low” and “high” alongside “max”. Assert each value
is forwarded through the native reasoning_effort field and that the legacy
thinking field remains undefined.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@packages/models/src/models/moonshot.ts`:
- Around line 657-660: Update the Kimi K3 comment in prepare-request-body.ts to
state that native reasoning_effort accepts “low”, “high”, and “max”, while
preserving the accurate default of “max”. Do not change the request-building
behavior, since non-disable efforts are already forwarded unchanged.

---

Nitpick comments:
In `@packages/models/src/models/moonshot.ts`:
- Around line 657-660: Extend the request-shape tests in
prepare-request-body.spec.ts for the K3 reasoningEfforts configuration to cover
both “low” and “high” alongside “max”. Assert each value is forwarded through
the native reasoning_effort field and that the legacy thinking field remains
undefined.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 5d1787d8-adde-4480-b5d2-f8b0bd8a6de6

📥 Commits

Reviewing files that changed from the base of the PR and between 1b5a083 and cbc740d.

📒 Files selected for processing (1)

• packages/models/src/models/moonshot.ts