Repository navigation
test(codex): 补 GPT-6 Ultra 双 client 路由与名称归一化断言 - #1915
Merged
Merged
Conversation
#1827 已运行时验证并落地 GPT-6 模型目录与推理档位（三个 sol/astra 六档含 ultra、luna 五档到 max），但其测试对两个行为存在盲区，本提交仅补断言： - ultra 门控需在 codex 与 codex-app 两个 client 上同时成立 （cliModelSupportsReasoningEffort 走 isCodexReasoningCliId 路由） - 模型名经 trim + 小写归一化后命中能力表（' GPT-6-Astra ' 仍可 ultra）， 并钉死 gpt-6-luna 在两个 client 上均可 max、不可 ultra 纯测试改动，src 零变更。变异验证：去掉名称归一化 2 测红、从 client 路由 移除 codex-app 1 测红；还原后 15/15 绿，build 0 错，目录相关 66 测绿。 源自 #1869（与 #1827 同窗口平行提交，源码部分已被逐字包含），按评审结论 剥离为 #1827 的 follow-up 测试。 Co-Authored-By: Claude Code <noreply@anthropic.com>
deepcoldy
requested review from
47seek,
LucasIcarus,
xiaoxueSunn and
xiongz-c
as code owners
October 14, 2026 09:26
|
🚀 Released in v3.44.0 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
改动
纯测试 PR，仅
test/codex-reasoning-effort.test.ts+13 行，src零变更。为 #1827 已落地的 GPT-6 模型目录补两条行为断言：ultra档位门控必须在codex与codex-app两个 client 上同时成立（cliModelSupportsReasoningEffort经isCodexReasoningCliId路由）；gpt-6-luna在两个 client 上均允许max、拒绝ultra。trim().toLowerCase()后命中能力表，' GPT-6-Astra '仍应获得含ultra的六档集合。背景
源自外部 PR #1869（作者 @Zzzen）。该 PR 与 #1827 在同一时间窗口平行提交：其两处源码改动（GPT-6 静态候选 + 六档/五档映射）在 #1827 合入时已被逐字包含，且 #1827 刻意把
gpt-5.5/gpt-5.2移出静态候选（live 探测成功时以运行时目录为权威，静态列表仅作兜底），故 #1869 按原内容无法干净合入。双审确认其唯一净增量是上述两组测试断言（主干现有套件的真实盲区），经作者沟通后剥离为本 follow-up，署名保留原作者。影响面
只新增测试断言，不改任何生产代码；覆盖 codex / codex-app 共享的推理强度映射（PTY 与 Tmux 会话共用）。
验证
bun run build：0 错。bun x vitest run test/codex-reasoning-effort.test.ts：15/15 通过；连同模型目录相关 4 个文件 66/66 通过。model?.trim().toLowerCase()归一化 → 2 个测试失败；isCodexReasoningCliId移除codex-app→ 1 个测试失败；🤖 Generated with Claude Code