[chrarnoldus]

Summary

Brings back the model formerly served as the stealth/cat-jam stealth model (removed in #7771) under its public name, reflection/beam-free. It is still free and still served through Vercel AI Gateway.

• Adds the reflection/beam-free Kilo-exclusive model (Reflection: Beam 1.0 (free)). Its metadata comes from the Vercel AI Gateway models endpoint: 1,048,576-token context window, 131,072 max output tokens, text-only input, low to max reasoning efforts, zdr: none, no_training: none.
• Not stealth: the model name is not redacted in usage records or error responses.
• Keeps requires-data-collection, because Vercel reports that this endpoint may train on prompts and has no ZDR.
• Restricts routing to the reflection inference provider and adds reflection to the inference provider ID registry. Vercel lists reflection as the only endpoint provider for this model. Without a provider identity, a non-stealth Vercel exclusive has no getInferenceProvider slug, so it would be left out of the providers snapshot. Enterprise orgs could then never use it: they require snapshot membership and match against provider allow lists. With the restriction, orgs can allow it through the reflection provider.
• Adds fallback reasoning efforts for reflection/beam*.

Verification

• pnpm --filter @kilocode/web-shared --filter web --filter ai-gateway typecheck
• oxlint -c .oxlintrc.json packages/web-shared/src/lib/ai-gateway
• pnpm --filter web test -- packages/web-shared/src/lib/ai-gateway/ packages/web-shared/src/lib/model-allow packages/web-shared/src/lib/organizations/ src/lib/providers/openrouter: 120 suites, 2240 tests passed (2 skipped)

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Executive Summary

The incremental change since the previous review adds only two explanatory comments to the OpenRouter inference-provider registry; no code, behavior, or type changes were introduced, and no new issues were found.

Overview

The only file changed since the last reviewed commit (194c3ea2) is packages/web-shared/src/lib/ai-gateway/providers/openrouter/inference-provider-id.ts, where two comment lines were added above the existing 'reflection' enum entry to document that reflection is a Vercel-only provider. The 'reflection' entry itself was already present and reviewed previously, and this documentation addition introduces no runtime, type, security, or logic impact.

Files Reviewed (1 file)

• packages/web-shared/src/lib/ai-gateway/providers/openrouter/inference-provider-id.ts - 0 issues

Previous Review Summary (commit 194c3ea)

Current summary above is authoritative. Previous snapshots are kept for context only.

Previous review (commit 194c3ea)

Status: No Issues Found | Recommendation: Merge

Files Reviewed (7 files)

• packages/web-shared/src/lib/ai-gateway/kilo-exclusive-models.ts
• packages/web-shared/src/lib/ai-gateway/kilo-exclusive-models.test.ts
• packages/web-shared/src/lib/ai-gateway/models.test.ts
• packages/web-shared/src/lib/ai-gateway/providers/openrouter/inference-provider-id.ts
• packages/web-shared/src/lib/ai-gateway/providers/reasoning-options.ts
• packages/web-shared/src/lib/ai-gateway/providers/reasoning-options.test.ts
• packages/web-shared/src/lib/ai-gateway/providers/vercel/mapModelIdToVercel.test.ts

---

_{Reviewed by deepseek-v4.1-flash · Input: 35.1K · Output: 4.2K · Cached: 265.3K}

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