[Klaud Cold][DO NOT MERGE] Smoke-test codeowner-signoff-verify after #2156 / 合并 #2156 后 codeowner-signoff-verify 冒烟测试

[functionstackx]

Summary

Throwaway test PR per the validation flow in AGENTS.md: after merging #2156 (verifier prompt moved out of the workflow YAML to fix the 21000-char expression cap), a sign-off comment will be posted here to trigger codeowner-signoff-verify end to end — gate → render step (envsubst of .github/codeowner-signoff-verify-prompt.md) → Claude verifier → verdict comment + commit status.

The verdict itself is expected to be REJECTED (no sweep ran on this PR) — the test is that the pipeline runs at all, which was impossible before #2156 (Invalid workflow file: Exceeded max expression length 21000).

This PR will be closed without merging once the verifier has responded.

中文说明

按 AGENTS.md 的验证流程创建的临时测试 PR：#2156 合并后（验证器提示词移出 workflow YAML，修复 21000 字符表达式上限），将在此发布签署评论以端到端触发 codeowner-signoff-verify —— gate → 渲染步骤（对 .github/codeowner-signoff-verify-prompt.md 执行 envsubst）→ Claude 验证器 → 结论评论 + commit status。

结论本身预期为 REJECTED（本 PR 未运行 sweep）——测试目的是验证流水线能够运行，这在 #2156 之前是不可能的（Invalid workflow file: Exceeded max expression length 21000）。

验证器响应后，本 PR 将不合并直接关闭。

🤖 Generated with Claude Code

[functionstackx]

As a PR reviewer and CODEOWNER, I have reviewed this and have:

• Verified that as of the moment of typing this, this is the latest version of PR_REVIEW_CHECKLIST.md
• Verified that the general code quality meets the InferenceX standard and does not make the code quality any worse.
• Verified that this PR has passed PR validation. Please link to GitHub Action workflow that shows this.
• Verified that this PR passes evals. Please link to GitHub Action workflow that shows this.
• Verified that speculative decoding PRs uses chat templates to align the AL distribution to real world
• Verified that the model architecture isn't changed with benchmark hacks like using --hf-overrides to skipping indexer for every x layers on models that don't natively support this. As a general rule, we won't accept optimizations that reduces the number of model architecture FLOPs. Anything that makes that same computation run faster is fair game; FLOPs at lower precisions is fine, given that the config passes private evals. As an general north star princple, we should only use optimizations which is used in production by customers that care about accuracy
• If an company claims that they support vLLM/SGLang as first class LLM inference engines on their hardware, I have verified that the respective vLLM submission made using upstream https://hub.docker.com/u/vllm docker repo, upstream SGLang https://hub.docker.com/u/lmsysorg docker repo. The only exceptions are for new hardware, such as MI455X UALoE72, Vera Rubin NVL72, Rubin NVL8, etc., and for new model architectures where there is an actual reason why vLLM/SGLang does not fundamentally support them yet as supported by vLLM/SGLang community maintainers
• If an company claims that they support vLLM/SGLang as first class upstream in-tree LLM inference engines on their hardware, I have have verified that the respective vLLM/SGLang submission has been made before additional frameworks (TRT-LLM, ATOM, etc.). The only exceptions are for new hardware, such as MI455X UALoE72, Vera Rubin NVL72, Rubin NVL8, etc., and for new model architectures where there is an actual reason why vLLM/SGLang does not fundamentally support them yet.
• Verified that the single-node recipes are similar to the official vLLM recipes and/or theSGLang cookbook:
  • If they are not, I have verified that a PR has been opened in vLLM recipe repo or SGLang repo and linked it below in the additional detail section:
• Verified that this PR does not patch the inference engine or serving stack — the pinned image must run as shipped. This covers .patch files / git apply / patch, inline patches embedded in benchmark scripts (e.g. a python3/sed heredoc that rewrites installed engine sources before serving), in-place edits of site-packages, monkey-patching, overwriting container files, and installing forked/rebuilt engine wheels on top of the pinned image. The only exception is a patch covered by a filled-out waiver at docs/waiver/<PR_NUMBER>.md — named after the PR that introduces the patch and filed in that same PR, stating what is patched, why the unmodified upstream image cannot run this benchmark, the pinned upstream PR/issue link, and the removal plan — which I have linked below in the additional detail section.
• If any of the above criteria cannot reasonably be satisfied, I have provided additional reasoning below.

Additional detail section:

• This is a deliberate SMOKE-TEST sign-off, per the verifier-validation flow in AGENTS.md, to exercise codeowner-signoff-verify end to end after [Klaud Cold] Fix signoff-verify parse error: move verifier prompt out of workflow YAML / 修复 signoff-verify 解析错误：将验证器提示词移出 workflow YAML #2156 moved the verifier prompt to .github/codeowner-signoff-verify-prompt.md. This [DO NOT MERGE] PR only adds a throwaway marker file; no sweep/evals ran, so a REJECTED verdict is the expected outcome. The PR will be closed after the verifier responds.

Signed: functionstackx

[Klaud-Cold]

❌❌❌ REJECTED ❌❌❌

@functionstackx — blocking: no sweep/evals ran on any commit in this PR and no authorized /reuse-sweep-run is on record. (Sign-off declares this a deliberate smoke test; REJECTED is the expected outcome for this [DO NOT MERGE] PR.)

✅ Check 0 (CODEOWNER): PASS — the only changed path (.github/signoff-verify-smoke-test.md) is owned solely by the catch-all * @InferenceX/core; no specific-owner mismatch, signer is an org MEMBER.
❌ Check 1 (passing sweep + evals on in-PR commit): FAIL — No passing sweep/eval was found on any commit in this PR. The sole commit f66bccd carries only CodeQL/analysis check-runs; no single-node */ or eval / jobs ever ran.
❌ Check 2 (evals pass): FAIL — no eval run exists to audit (follows from Check 1).
➖ Check 3 (recipe link): N/A — PR changes no single-node benchmark files or master-config entries (throwaway marker file only).
❌ Check 4 (reuse-sweep-run posted): FAIL — No authorized /reuse-sweep-run command has been posted on this PR; an authorized maintainer must comment /reuse-sweep-run before this PR could merge via reuse.
✅ Check 5 (latest checklist template): PASS — all current-template items present; the two unchecked items (validation, evals) are explained in the additional detail section.
➖ Check 6 (upstream images / engine-first): N/A — no master-config entries added or changed.
➖ Check 7 (no architecture hacks): N/A — no server args touched.
➖ Check 8 (spec-decode chat templates): N/A — no speculative-decoding changes.
➖ Check 9 (no engine patches): N/A — no benchmark scripts, images, or configs touched.

[functionstackx]

Smoke test complete: the verifier ran end to end on main's fixed workflow (run 29122688183) — gate ✅, prompt render ✅ (21,318 bytes), Claude verification ✅, verdict comment + commit status ✅. REJECTED verdict was the expected outcome for this throwaway PR. Closing without merging. / 冒烟测试完成：验证器已在 main 的修复版 workflow 上端到端运行，按预期返回 REJECTED。不合并，直接关闭。