fix(fim): set has_error on microdollar_usage for upstream errors

[kilo-code-bot]

Summary

FIM (fill-in-the-middle) requests to Mistral/Inception were never recording has_error = true in microdollar_usage, even when the upstream provider returned a 4xx/5xx HTTP response.

Root cause: parseMistralFimUsageFromString and parseMistralFimUsageFromStream had no access to the HTTP status code. The regular chat-completion parsers already mirror this pattern: the streaming parser initialises reportedError = statusCode >= 400, and the non-streaming parser includes statusCode >= 400 in the hasError expression.

Changes:

• Added statusCode: number parameter to both FIM parse functions
• Non-streaming: hasError: !json.model || statusCode >= 400 (mirrors parseMicrodollarUsageFromString)
• Streaming: let reportedError = statusCode >= 400 (mirrors parseMicrodollarUsageFromStream)
• countAndStoreFimUsage extracts usageContext.status_code ?? 0 and passes it to both parsers

Verification

• Reviewed diff against equivalent logic in processUsage.ts (lines 647 and 760)
• TypeScript: change is type-safe (new number param, boolean expression unchanged)

Visual Changes

N/A

Reviewer Notes

status_code on MicrodollarUsageContext is number | null; the ?? 0 fallback means a missing status code does not falsely set has_error. In practice status_code is always set before countAndStoreFimUsage is called (route.ts:249).

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 files)

• apps/web/src/lib/ai-gateway/llm-proxy-helpers.ts

---

_{Reviewed by gpt-5.4-20260305 · 245,776 tokens}