fix(ai-gateway): drop reasoning_effort when auto model sets reasoning config

[chrarnoldus]

Summary

When an auto model (e.g. kilo-auto/frontier or a variant-resolved kilo-auto/efficient decision) splices a reasoning config into a chat completions request, any client-supplied reasoning_effort was left in the body. Upstream providers can then receive conflicting reasoning settings.

applyResolvedAutoModel now deletes reasoning_effort from chat completions bodies whenever it sets reasoning. Requests where the auto model resolves without a reasoning config keep the client's reasoning_effort unchanged.

Verification

• pnpm typecheck (apps/web)
• oxlint on apps/web/src/lib/ai-gateway/auto-model
• Added applyResolvedAutoModel tests in resolution.test.ts; test execution left to CI.

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Executive Summary

The change correctly deletes a client-supplied reasoning_effort from chat-completions bodies only when applyResolvedAutoModel splices in a resolved reasoning config, leaving it untouched on the no-reasoning path; the added tests cover both behaviors. No correctness, security, performance, or memory-leak issues found.

Files Reviewed (2 files)

• apps/web/src/lib/ai-gateway/auto-model/resolution.ts
• apps/web/src/lib/ai-gateway/auto-model/resolution.test.ts

---

_{Reviewed by deepseek-v4.1-flash · Input: 0 · Output: 0 · Cached: 0}

_{Review guidance: REVIEW.md from base branch main}