Skip to content

fix(translation): share Retry-After cooldown and preserve request isolation - #991

Merged
Bistutu merged 27 commits into
mainfrom
codex/pr906-cooldown-integration-20261015
Oct 14, 2026
Merged

Bistutu merged 27 commits into
mainfrom
codex/pr906-cooldown-integration-20261015

Conversation

@Bistutu

@Bistutu Bistutu commented Oct 14, 2026

Copy link
Copy Markdown
Member

整合 #906,使 OpenAI 兼容 SDK 线路在收到 429/带有效 Retry-After 的 503 后按实际凭据和可信隐私来源共享冷却。逻辑等待不占并发槽,真实 HTTP 请求持续占槽直到响应体结束;更换 Key、模型或请求策略时不会误共享旧的 pending 请求。

保留 @ArietidsZ 的原提交历史;此 PR 是包含修复的 #906 集成替代。与当前主干整合时同时保留 AI 超时设置、组合服务、图片结果策略及页面上下文恢复,并登记新增审计测试。

请求调度增量使标准油猴包超过原有体积门禁,因此让标准版复用已有 franc 语言模型纯数据无损压缩机制。完整 66 个模型、检测算法、许可、固定资源版本与体积上限保持原值。实际标准产物为 2,058,335 字节,原有 2,095,100 字节上限及 verifier 均通过。

验证:最终 26 个定向文件、1,049 个请求调度及隐私处理用例通过,broker/snapshot/scheduler/输入框入口/图片上下文五个模块四项覆盖率均为 100%;另有 205 个构建及解压用例通过。测试审计、类型检查、Chrome/Firefox/标准 userscript 构建、manifest 和 userscript verifier、文档构建均通过。独立代码复核无阻断项;未执行全量回归或真实供应商在线验收。

ArietidsZ and others added 27 commits October 9, 2026 19:17
…vacy

Keep native-batch configuration, recovery and shared real-attempt quota contracts. Capture input sender privacy before awaiting readiness and attach control to the same request.

Preserve the original standard userscript budget with CommonJS determinism and exclusion of unavailable image/area settings; relay the existing video-state query protocol. Validate 606 distinct targeted cases, compile, audit, standard verifier, Chrome/Firefox and manifests. Standalone builds but the original 3,600,000-byte gate fails at 3,816,337 bytes.
Bind the 606 targeted cases, original standard verifier, extension builds and current size ablation to the verified source. Record five distinct gzip hypotheses, unchanged decoded bytes and limits. Retain the original standalone budget failure and historical report bytes.
…nges

Rebase tree on current main while retaining branch history. Include 24 targeted audit cases. Final affected rerun: 93 regression plus 2 fixed image assertions pass; aggregate unique evidence 598/600, with two reproduced baseline stats failures. Narrow types retains two baseline diagnostics, no added diagnostics. No browser/model acceptance or merge.
@Bistutu
Bistutu merged commit b412591 into main Oct 14, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants