[diegosouzapw]

Closes #3572

Problem

A Claude Code max_tokens: 1 connectivity ping legitimately returns HTTP 200 with stop_reason: "max_tokens" and content: []. isEmptyContentResponse() flagged any empty-content payload regardless of stop_reason/finish_reason, so OmniRoute rewrote the valid 200 into a synthetic 502 "Provider returned empty content" and triggered a needless family fallback (16 false 502s in one morning on a prod proxy, per the report).

Fix (open-sse/services/errorClassifier.ts)

Treat a terminal truncation/tool signal as a legitimate completion even when content is empty:

• Claude branch: empty content[] is only flagged when stop_reason ∉ {max_tokens, tool_use}.
• OpenAI branch: empty choice is not flagged when finish_reason ∈ {length, tool_calls}.

Genuinely empty responses (no terminal reason, or stop/end_turn with empty content) are still caught — the fake-success guard for web/no-auth providers is preserved.

Regression test (Rule #18 — TDD, RED→GREEN)

tests/unit/empty-content-stopreason-3572.test.ts (8 cases): max_tokens/length pass-through, no-stop_reason still flagged, end_turn/stop still flagged. Failed on unfixed code (3 RED), passes after. Existing services-branch-hardening empty-content assertions stay green.

[gemini-code-assist]

Code Review

This pull request addresses an issue where valid, truncated upstream responses (such as those with a max_tokens or tool_use stop reason) were incorrectly flagged as empty content and rewritten into synthetic 502 errors. The isEmptyContentResponse utility in open-sse/services/errorClassifier.ts was updated to recognize legitimate terminal stop signals for both Claude and OpenAI providers, and comprehensive unit tests were added under tests/unit/ to verify this behavior. No review comments were provided, so there is no feedback to address.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

This PR correctly addresses the issue where valid max_tokens-truncated responses were misclassified as empty content. The isEmptyContentResponse() function now recognizes terminal stop signals (max_tokens/tool_use for Claude, length/tool_calls for OpenAI) as legitimate completions even with empty content.

The fix preserves the original fake-success guard for genuinely empty responses while allowing valid truncated/tool-use responses to pass through. The test coverage (8 cases) is comprehensive and validates all intended behaviors.

Files Reviewed (3 files)

• open-sse/services/errorClassifier.ts - fix correctly implemented
• tests/unit/empty-content-stopreason-3572.test.ts - comprehensive test coverage
• CHANGELOG.md - appropriate changelog entry

---

_{Reviewed by laguna-m.1-20260312:free · 2,170,947 tokens}