feat(ai-gateway): add direct Qwen3.7 Plus model

[lambertjosh]

Summary

• Register qwen/qwen3.7-plus as a public Alibaba-direct model alongside Qwen3.7 Max.
• Add multimodal and reasoning metadata plus Alibaba International tiered pricing for standard input, output, explicit cache reads, and explicit cache writes.
• Cover direct routing and discounted pricing behavior, including the 256K long-context boundary.

Verification

• Not manually tested; this is a gateway model registration and pricing-only change with no visual flow.

Visual Changes

N/A

Reviewer Notes

• Alibaba implicit-cache accounting for repeated one-shot requests is tracked separately in fix(ai-gateway): distinguish Alibaba explicit and implicit cache pricing #3654. Normal multi-message traffic receives explicit cache breakpoints from the gateway.

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Executive Summary

Incremental review confirms no new issues — the PR correctly registers qwen/qwen3.7-plus as a public Alibaba-direct model with tiered pricing (≤256K / >256K), consistent 0.65 discount math, correct tier boundary at TOKENS_256K = 262144, and adequate test coverage for all pricing paths.

Files Reviewed (3 files)

• apps/web/src/lib/ai-gateway/providers/qwen.ts — model definition, tiered pricing, alibabaDirectModels registration
• apps/web/src/lib/ai-gateway/processUsage.calculatKiloExclusiveCost.test.ts — pricing unit tests for ≤256K, >256K, cache, and boundary conditions
• apps/web/src/lib/ai-gateway/models.test.ts — routing and isAlibabaDirectModel tests

Fix these issues in Kilo Cloud

---

_{Reviewed by claude-sonnet-4.6 · 561,074 tokens}

_{Review guidance: REVIEW.md from base branch feat/auto-balanced-qwen37-max}

[lambertjosh]