jlcodes99/cockpit-tools v1.3.73 on GitHub

更新日志（中文）

API 服务 Rust 引擎取消文本请求的 5 分钟总时长限制、完整响应的 32 MiB 上限、输出项的 8 MiB 累计上限，以及 Chat 转换的 32 MiB／4096 项上限；Codex 流事件限制由 8 MiB 调整为单行 50 MiB。
图片 API 响应上限由 32 MiB 提高至 96 MiB，图片上传的 64 MiB 限制由单次请求累计调整为单文件计算；独立图片 API 保留 5 分钟总时长限制。
取消 WebSocket 等待客户端首条消息的 30 秒限制，Chat 供应商流限制由 8 MiB 调整为单行 4 MiB，账号并发等待队列新增最多 100 个请求的上限。
恢复 API 服务绑定账号的可选 OAuth 额度预留，包括 5 小时与周阈值、安全垫、保护状态及两种服务引擎的请求拦截。新绑定默认不勾选“保留 OAuth 额度”，已有绑定保留已保存的选择；若旧版本已清除预留设置，需要重新开启。
API 服务客户端底部与概览共用当前会话的平均输出速度和按输入 Token 加权的平均缓存命中率，停止生成后保留；请求记录区分请求与返回的速度档位。
加快近期 API 请求记录的打开与刷新，新请求采用带索引的摘要并即时更新弹框；保留已有请求历史，不做批量重建。

Windows 与 macOS 的 Codex API 服务会话改为在当前页面启用额度兼容，无需重载窗口或等待后续页面加载；修正订阅额度耗尽对发送按钮的误拦截，隐藏对应耗尽提示，停用增强时恢复本地额度缓存。
底部与概览的输出速度、缓存命中率统一保留一位小数，直接舍弃后续位数，不四舍五入；99.99% 显示为 99.9%。
减少 Codex 切号后的重复账号刷新与统计，后台刷新期间保留已确认数据；限制启动前会话检查的耗时，并在启动弹框内明确提示超时或恢复失败。
后台刷新失败或切换弹框页签时保留已确认的请求记录，区分无记录、未关联会话与服务不可用，并拒绝旧会话或旧服务绑定的迟到结果。
数据库短暂读取失败时保留已核实的会话平均速度与缓存测量，缓存验证过期后及时失效，切换服务绑定或会话时不沿用旧测量。

Remove the Rust API Service engine's five-minute text request limit, 32 MiB full-response limit, 8 MiB accumulated output limit, and Chat conversion limits of 32 MiB and 4,096 items. Change the Codex stream event limit from 8 MiB to a 50 MiB per-line limit.
Increase image API response limits from 32 MiB to 96 MiB and change the 64 MiB image upload limit from a combined request limit to a per-file limit. Keep the independent image API's five-minute total timeout.
Remove the 30-second wait limit for the first client WebSocket message, change Chat provider stream limits from 8 MiB to 4 MiB per line, and cap the account concurrency wait queue at 100 requests.
Restore optional OAuth quota reserves for API Service bindings, including 5-hour and weekly thresholds, safety margins, protection status and admission checks in both gateway engines. Quota reservation is unchecked by default for new bindings. Existing bindings keep their saved choice; enable protection again if an earlier version removed its settings.
API Service clients use the same current-chat average output speed and input-token-weighted cache hit rate in the footer and overview, retaining the measurements after generation stops. Request records distinguish requested and returned speed tiers.
Open and refresh recent API request records faster, with indexed summaries for new requests and immediate dialog updates. Existing request history is retained without a bulk rebuild.

Activate quota compatibility in the current Codex API Service page on Windows and macOS without reloading the window or waiting for a later page load. Remove incorrect send blocking caused by exhausted subscription quota, hide the corresponding notice, and restore the local quota cache when enhancements stop.
Show output speed and cache hit rates in the footer and overview with exactly one decimal place, discarding remaining digits without rounding; 99.99% displays as 99.9%.
Reduce repeated account refreshes and statistics work after Codex switches, retaining confirmed counters during background refresh. Limit launch-time session checks and report timeouts or recovery failures inside the launch dialog.
Keep verified request records visible during background refresh failures and when switching dialog tabs. Distinguish empty results, unlinked chats and unavailable services, and reject replies from previous chats or service bindings.
Preserve verified current-chat average speed and cache measurements across brief database read failures, expire caches when their verification lapses, and clear measurements when the service binding or chat changes.