[chrarnoldus]

Summary

Follow-up to #8018. Jira: AISAGE-706.

This PR logs a warning whenever the gateway adds Reflection reasoning placeholders to a request. That shows how often clients send Reflection history without reasoning, which happens after a model switch, compaction, or with clients that drop reasoning.

[addReflectionReasoningPlaceholders] mutated request with reasoning placeholders
{ model, requestKind, placeholderCount }

• The log fires once per request, and only when at least one placeholder was added.
• It matches the existing request-mutation logs ([removeNonSensicalMaxTokens], [applyProviderSpecificLogic] Removed unsupported request-level service tier): console.warn with a structured payload.
• The payload has only the model, the request kind and a count. Message content is never logged.

The per-API placeholder helpers now return how many placeholders they added. addReflectionReasoningPlaceholders takes the requested model so the log can include it.

Verification

• No manual testing. The change only adds a log line and a counter. Placeholder insertion is unchanged and was verified against Vercel AI Gateway in [AISAGE-706] fix(ai-gateway): add reasoning placeholders for Reflection models #8018.

Visual Changes

N/A

Reviewer Notes

The log uses console.warn, like the other request-mutation logs. console.info would also be reasonable, since placeholders are expected during normal operation.

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Executive Summary

The change threads the requested model through addReflectionReasoningPlaceholders, returns per-API placeholder counts, and logs a structured warning only when at least one placeholder was inserted; the counting is consistent with the existing mutation logic across all three request kinds.

Files Reviewed (3 files)

• packages/web-shared/src/lib/ai-gateway/providers/apply-provider-specific-logic.ts
• packages/web-shared/src/lib/ai-gateway/providers/reflection.ts
• packages/web-shared/src/lib/ai-gateway/providers/reflection.test.ts

---

_{Reviewed by deepseek-v4.1-flash-1017 · Input: 42K · Output: 5.2K · Cached: 225.5K}

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