@feiyang666/dsh-usage-plugin 1.14.0 → 1.14.1

package/CHANGELOG.md

@@ -8,9 +8,27 @@
 
 ---
 
-## v1.14.0 (2026-08-27)
+## v1.14.1 (2026-08-29)
+
+### 修复 / Fixed
+
+- **中断调用兜底记录，使「调用次数」与 DeepSeek 官方后台对齐**（`lib/index.js`）：harness 在流被中断（aborted / error / timeout / 用户停止生成）时不会产出 `usage` chunk（usage 只在收到 `[DONE]` 哨兵后才由 adapter yield），导致这类调用此前完全不被插件记录——而官方后台仍会把该次请求计入「API 请求次数」并按实际 token 计费，于是插件统计的调用次数长期低于官方（如 8 月 27 日 985 vs 1033）。现在无 usage 但确为真实模型调用（通过 `isRealCall`，排除内部 `dsh2shell-*`/`fake`）的流，会在 `observe()` 的 `finally` 里补记一条 **0-token 的「中断调用」**（新增 `interrupted: true` 字段，`finishReason` 如实标记 `aborted`/`error`/`timeout`）。token 与费用均为 0，不会虚增消耗；调用次数与官方口径一致。
+- **面板透明展示中断调用**（`lib/client.js`）：
+  - 概览「调用次数」卡片新增 `· 中断 N（未计费）` 提示；
+  - 缓存命中列表与用量日历当日明细中，中断调用行在模型名旁显示红色「中断」徽标、结束原因红色标注、消耗列显示 `—`；
+  - PNG 导出报告同样在模型名旁标注「中断」、消耗列显示 `—`；
+  - 结束原因新增 `aborted → 已中断 / Interrupted`、`timeout → 超时 / Timeout` 映射（中英双语）。
+  - **用量面板顶部新增「本地统计与官方后台的差异」提示横幅**：说明本面板统计的是插件本地捕获的调用（官方价格 + 峰谷时段），与官方后台（platform.deepseek.com 用量页）相比金额可能更低——① 中断/出错/超时的调用官方仍按实际 token 计费而插件按 0 记录；② 账号下其他 API Key（其它应用/脚本）的调用不经过 DeepSeek Harness，官方包含而插件不包含；③ 精确对账可导出官方月度账单 CSV 对比。检测到中断调用时额外显示「当前记录中有 N 次中断调用（未计费）」红字提示。
+- **持久化与 API 透传 `interrupted`**（`lib/index.js`）：`normalizeRecord` / `projectRecord` 均保留该字段，重启恢复、`/usage/api` list 输出、JSON 导出一致。
+
+### 测试 / Tests
 
-> 📄 本版本英文发布说明（English release notes）：[RELEASE_NOTES.md](./RELEASE_NOTES.md)
+- 新增 `test/interrupt.test.js`：验证中断兜底仅记录真实模型调用（DeepSeek 官方 / 第三方真实 provider），内部 `fake`/`dsh*` 占位调用永不兜底记录。
+- 全部 30/30 通过。
+
+---
+
+## v1.14.0 (2026-08-27)
 
 ### 新增 / New
 
@@ -277,6 +295,3 @@
 - CSV / JSON / PNG 长图导出（最新在前，最多 2000 条，超出提示），支持自定义导出目录（原生目录选择器）与「打开所在目录」。
 - JSON / CSV 导入合并，按时间去重。
 - Windows / macOS / Linux 跨平台路径与原生操作适配；启动诊断日志 `dsh-usage-boot.log`。
----
-
-📄 **英文发布说明入口（English release notes）**：[RELEASE_NOTES.md](./RELEASE_NOTES.md)

package/README.md

@@ -35,6 +35,8 @@ dsh-usage-plugin is a **usage & cost tracker** plugin in the DeepSeek Harness ec
 - **Usage & Cost**: records each model call's token usage and cache hits (input miss / cache hit / cache write / output / reasoning / finish reason), and computes cost using DeepSeek's peak/valley or base pricing (peak hours on weekdays are automatically priced by Beijing time 09:00–12:00 and 14:00–18:00; since 2026-08-23 weekends are billed entirely at the off-peak rate). Model names come from the actual request parameters, so non-DeepSeek models are shown truthfully instead of "unknown model"; models without an official price are counted as 0. The overview shows a by-model table plus a by-API-provider × model drill-down (each provider grouped with every model's calls and peak/off-peak cost split) and a grand total row. The overview also supports **date filtering** (Today / Last 7 days / Last 30 days / All, plus a custom start–end range), so the aggregate stats can be scoped to any single day or date range.
 - **Usage Calendar**: a monthly daily-usage heatmap (colored by cost or call count), hover for details including the peak/off-peak cost split, click a day for its call list and peak/off-peak totals, plus a per-day statistics table with peak cost / off-peak cost / total columns and monthly rollups.
 - **Cache Hit List**: newest-first, fully scrollable, with quick filters (Today / 7 days / 30 days / All) and custom date ranges; the summary line and footer total split peak vs off-peak consumption with a grand cost total. The list is paginated (100 rows per page), so it stays smooth even with large data volumes.
+- **Interrupted calls shown truthfully**: calls that were aborted / errored / timed out (e.g. manually stopped generation, stream interruption) are shown with a red **"Interrupted"** badge, the finish reason (Interrupted / Error / Timeout) and a `—` cost. The official console still counts these as API requests and bills their actual tokens, but the harness does not report their usage to the plugin — so the plugin records them at 0 tokens, keeping the **call count aligned with the official console** while costs are unaffected. The Overview's "Calls" card adds a `· interrupted N (not billed)` hint.
+- **Local stats vs official console**: a fixed notice banner at the top of the panel explains that this panel reflects calls captured locally by the plugin (official prices + peak/off-peak hours), and that the official console (platform.deepseek.com usage page) may show a higher amount because: ① interrupted/failed/timed-out calls are still billed by the console while the plugin records them as 0; ② calls from other API keys on your account (other apps/scripts) do not pass through DeepSeek Harness — the console includes them, the plugin does not; ③ for exact reconciliation, export the official monthly billing CSV and compare. When interrupted calls are detected, the banner also shows a red "Current records include N interrupted call(s) (not billed)" line.
 - **Price Table**: the official DeepSeek API price table (covering `deepseek-v4-flash` / `deepseek-v4-flash-vision-exp` / `deepseek-v4-pro`) — base and peak/valley unit prices shown side by side (peak vs off-peak), editable in-panel and persisted to `pricing.json`, with a reset-to-default option.
 - **Balance Query**: queries your DeepSeek account balance using the configured `DEEPSEEK_API_KEY`.
 - **Export**: CSV / JSON / **PNG long image** (newest-first, up to the latest 2000 records, warns if exceeded; the PNG report includes peak/off-peak cost columns), to any directory (native picker), auto-opens the folder after export.
@@ -309,6 +311,10 @@ pnpm dsh web
 - **[@mumuer1024](https://github.com/mumuer1024)**: reported and diagnosed the persistence-path drift across workspaces (history "disappearing" / counted as 0) and proposed storing data in a fixed, dedicated directory ([#4](https://github.com/feiyang-dev/dsh-usage-plugin/issues/4)).
 - **[@liu3734](https://github.com/liu3734)**: reported and diagnosed the Windows-only path handling / spawn issues on macOS (POSIX) and proposed the cross-platform fix ([#1](https://github.com/feiyang-dev/dsh-usage-plugin/issues/1)).
 
+## Changelog
+
+- [CHANGELOG.md](./CHANGELOG.md) — detailed notes for every release.
+
 ## License
 
 MIT © dsh-usage-plugin

package/README.zh.md

@@ -6,7 +6,6 @@
 
 [GitHub](https://github.com/feiyang-dev/dsh-usage-plugin) · [npm](https://www.npmjs.com/package/@feiyang666/dsh-usage-plugin) · MIT License
 
-> 📄 **本版本英文发布说明（What’s New in v1.14.0）**：[RELEASE_NOTES.md](./RELEASE_NOTES.md)
 
 **由开发者制作的 DeepSeek Harness 插件** —— 记录每一次模型调用的 token 用量与消耗，支持峰谷计费、余额查询、日历热力图与 CSV / JSON / PNG 导出。
 
@@ -37,6 +36,8 @@ dsh-usage-plugin 是 DeepSeek Harness 生态的**用量与消耗统计插件**
 - **用量与消耗**：记录每次模型调用的 token 用量与缓存命中（输入·未命中 / 缓存命中 / 缓存写入 / 输出 / 推理 / 结束原因），按 DeepSeek 峰谷/基础价格计算消耗（工作日高峰时段按北京时间 9:00–12:00、14:00–18:00 计价；自 2026-08-23 起周末全天按空闲价计费）。模型名以请求参数为准如实显示（非 DeepSeek 模型不再显示为「未知模型」，无官方价格的模型消耗按 0 统计）。概览含「按模型」表与「按 API 服务商 × 模型」明细表（每个服务商一组，组内列出各模型的调用与高峰/空闲分列消耗），底部有总费用合计。概览还支持**日期筛选**（今天 / 近 7 天 / 近 30 天 / 全部，外加自定义起止区间），可将聚合统计限定到任意一天或日期范围。
 - **用量日历**：按月查看每日用量热力图（按消耗或调用数着色），悬停查看详情（含高峰 / 空闲消耗拆分）、点击某天查看当日调用明细与高峰/空闲消耗统计，附本月每日统计表（高峰消耗 / 空闲消耗 / 总消耗分列）与月度汇总。
 - **缓存命中列表**：最新记录排在最前，支持 今天 / 近7天 / 近30天 / 全部 快捷筛选与自定义起止日期区间，汇总行与表尾合计区分高峰消耗 / 空闲消耗 / 总费用合计；列表分页渲染（每页 100 条），记录量大也不卡顿。
+- **中断调用如实呈现**：被中止 / 出错 / 超时（如手动停止生成、流中断）的调用在面板中标红显示「中断」徽标、结束原因（已中断 / 错误 / 超时）与消耗 `—`。这类调用官方后台仍会计入「API 请求次数」并按实际 token 计费，但 harness 不向插件上报 usage——插件按 0 token 记录它们，**调用次数与官方口径一致，消耗不受影响**；概览「调用次数」卡片会提示 `中断 N（未计费）`。
+- **本地统计 vs 官方后台的差异提示**：用量面板顶部固定显示说明横幅——本面板统计的是插件本地捕获的调用（官方价格 + 峰谷时段），与官方后台（platform.deepseek.com 用量页）相比金额可能更低：① 中断/出错/超时的调用官方仍按实际 token 计费而插件按 0 记录；② 账号下其他 API Key（其它应用/脚本）的调用不经过 DeepSeek Harness，官方包含而插件不包含；③ 精确对账可导出官方月度账单 CSV 对比。检测到中断调用时，横幅额外以红字显示「当前记录中有 N 次中断调用（未计费）」。
 - **价格表**：**DeepSeek 官方 API 价格表**（覆盖官方模型 `deepseek-v4-flash` / `deepseek-v4-flash-vision-exp` / `deepseek-v4-pro`），展示基础价与峰谷价（高峰/空闲）单价表，高峰价与空闲价分列展示，支持在面板内直接编辑价格并持久化（数据目录 `pricing.json`），也可一键恢复默认。
 - **剩余余额查询**：用当前配置的 `DEEPSEEK_API_KEY` 查询 DeepSeek 账户余额；并支持 **百炼 Token Plan 配额查询**（复用 `bl auth login --console` 的控制台 OAuth token，无需阿里云 AccessKey，展示本周配额已用百分比与重置时间）。
 - **导出**：CSV / JSON / **PNG 长图**（按最新在前展示，最多含最近 2000 条，超出会提示；PNG 报告含高峰 / 空闲消耗分列统计），可导出到任意目录（原生目录选择器），导出后自动打开所在目录。
@@ -311,6 +312,10 @@ pnpm dsh web
 - **[@mumuer1024](https://github.com/mumuer1024)**：报告并定位了持久化路径随会话工作区漂移导致历史数据"消失/统计为 0"的问题，提出把数据写到固定专用目录的方案（[#4](https://github.com/feiyang-dev/dsh-usage-plugin/issues/4)）。
 - **[@liu3734](https://github.com/liu3734)**：报告并定位 macOS（POSIX）下路径处理与 spawn 的 Windows 专用问题，提出跨平台修复方案（[#1](https://github.com/feiyang-dev/dsh-usage-plugin/issues/1)）。
 
+## 更新日志
+
+- [CHANGELOG.md](./CHANGELOG.md) —— 每次发布的详细更新内容。
+
 ## 许可
 
 MIT © dsh-usage-plugin

package/lib/client.js

@@ -21,7 +21,7 @@ window.__ModuleLoader__.load({
     } catch (e3) {}
 
     // ── i18n: zh keys → en dictionary (community addition) ──
-    var __T_EN = {"本期无消耗（所有用量为 0）":"No usage this turn (all zeros)","按模型":"By model","…及另外 ":"…plus "," 个模型（完整列表见「用量与消耗」页签）":" more models (full list in the Usage & Cost tab)","对话累计":"Conversation total","本轮明细":"This turn","本轮 token":"Turn tokens","本轮消耗":"Turn cost","本轮缓存命中率":"Turn hit rate","本轮耗时":"Turn duration","本轮无记录":"No records this turn","工具调用（内部）":"Tool calls (internal)"," 条内部调用":" internal calls","展开":"Expand","收起":"Collapse","详细记录请在「用量与消耗」页签查看":"Full details are in the Usage & Cost tab","获取中…":"Loading…","总 token":"Total tokens","Token 明细":"Token details","本次输出":"This output","总消耗":"Total cost","耗时":"Run time","缓存命中率":"Cache hit rate","关闭":"Close","无 Token 数据":"No token data","当前 · ":"Current · ","工作日高峰时段":"Weekday peak hours","工作日空闲时段":"Weekday off-peak hours","周末高峰时段":"Weekend peak hours","周末空闲时段":"Weekend off-peak hours","周一至周五 9:00–12:00、14:00–18:00":"Mon–Fri 9:00–12:00 & 14:00–18:00","工作日其余时间":"Other weekday hours","周末（周六、周日）全天按空闲价计费":"All weekend (Sat & Sun) billed at the off-peak rate","2026-08-23 前仍按原规则分峰谷":"Before 2026-08-23 the old peak/valley split still applies","高峰 / 空闲时段说明":"Peak / off-peak hours","工作日高峰":"Weekday peak","空闲时段":"Off-peak hours"," 周一至周五 9:00–12:00、14:00–18:00":" Mon–Fri 9:00–12:00 & 14:00–18:00"," 工作日其余时间，以及周末（周六、周日）全天（自 2026-08-23 起；此前仍按原规则分峰谷）":" Weekday off-peak hours, plus all weekend (Sat & Sun) since 2026-08-23 (before that, the old peak/valley split still applies)","工作日高峰 9:00–12:00、14:00–18:00 · 周末全天空闲价":"Weekday peak 9:00–12:00 & 14:00–18:00 · all-weekend off-peak rate","\n调用 ":"\ncalls ","）\n输入·未命中 ":")\ninput·miss ","\n高峰消耗 ":"\npeak cost ","\n无记录":"\nno records","自动模式：按调用时间自动选择计费档位——新价格表生效前的调用按基础价（旧价格表），生效后的调用按峰谷价（工作日高峰时段 9:00–12:00、14:00–18:00 用高峰价，其余时间——含周末全天——用空闲价）。当前显示：":"Auto mode: the billing tier is chosen automatically by call time — calls made before the new price schedule took effect are billed at base (old) prices; calls after it use peak/valley prices (weekday peak hours 9:00–12:00 & 14:00–18:00 at the peak price, all other times — including all weekend — at the off-peak price). Currently shown:","单位：元 / 百万 tokens。工作日高峰时段（北京时间 9:00–12:00、14:00–18:00）用高峰价，其余时间用空闲价（空闲价 = 高峰价的一半）；自 2026-08-23 起周末（周六、周日）全天按空闲价计费，新规则生效前仍按原规则。概览、用量日历、缓存命中列表中的消耗均已按高峰 / 空闲分列统计。":"Unit: CNY per million tokens. Weekday peak hours (Beijing time 9:00–12:00 & 14:00–18:00) use the peak price; all other times use the off-peak price (off-peak = half of peak). Since 2026-08-23, weekends (Sat & Sun) are billed at the off-peak rate all day; calls before the new rule still follow the old split. Costs in Overview, Usage Calendar and the Cache Hit List are split into peak / off-peak accordingly.","未知模型":"Unknown model","未知服务商":"Unknown provider","完成":"Done","工具调用":"Tool call","超长":"Overlong","错误":"Error","自动（生效前基础价 · 生效后峰谷价）":"Auto (base price before effective date · peak/valley after)","基础价格":"Base price","峰谷价格":"Peak/valley price","自动":"Auto","canvas 2d 不可用":"canvas 2D unavailable","用量与消耗报告":"Usage & Cost Report","生成时间 ":"Generated ","（北京） · 共 ":" (Beijing) · total "," 条调用 · 缓存命中率 ":" calls · cache-hit rate "," · 高峰价消耗 ":" · peak-rate cost "," · 空闲价消耗 ":" · off-peak cost "," · 合计 ":" · total "," · 报告仅含最近 ":" · report limited to the latest "," 条":" records","调用次数":"Calls","输入 · 未命中":"Input · miss","输入 · 缓存命中":"Input · cache hit","输出":"Output","高峰消耗":"Peak cost","空闲消耗":"Off-peak cost","总消耗 (峰谷价)":"Total cost (peak/valley)","缓存命中列表（共 ":"Cache hit list ("," 条）":" records)","时间(北京)":"Time (Beijing)","模型":"Model","输入·未命中":"Input·miss","缓存命中":"Cache hit","缓存写入":"Cache write","推理":"Reasoning","命中率":"Hit rate","时段":"Period","结束":"End","消耗(峰谷)":"Cost (peak/valley)","峰":"Peak","谷":"Valley","消耗表（按模型 · 峰谷价 · 高峰/空闲分列）":"Cost table (by model · peak/valley prices)","调用":"Calls","总消耗":"Total cost","合计":"Total","消耗表（按 API 服务商 × 模型 · 峰谷价）":"Cost table (by API provider × model · peak/valley)","API 服务商 / 模型":"API provider / model"," 个模型）":" models)","总费用合计":"Total cost overall","计价按请求的 API 服务商分别应用已核验价格；DigitalOcean 美元价按 USD/CNY 汇率折算；无法可靠映射的第三方价格按 ¥0。":"Billing applies verified prices per requesting API provider; DigitalOcean USD prices are converted at the USD/CNY rate; third-party prices that cannot be mapped reliably are counted as ¥0."," 个服务商":" providers","今天":"Today","近7天":"Last 7 days","近30天":"Last 30 days","全部":"All","自定义区间：":"Custom range:","至":"to","清除筛选":"Clear filters","当前范围：共 ":"Current range: ","（全部记录）":"(all records)"," · 总消耗 ":" · total cost ","高峰 ":"Peak "," · 空闲 ":" · off-peak ","命中率 ":"hit rate ","高峰时段 9:00–12:00、14:00–18:00":"Peak hours 9:00–12:00 & 14:00–18:00","其余空闲时段":"other hours are off-peak","高峰 + 空闲 · ":"Peak + off-peak · ","消耗表（按模型）":"Cost table (by model)","API 服务商":"API provider","高峰/空闲":"Peak/Off-peak","消耗明细（按 API 服务商 × 模型）":"Cost breakdown (by API provider × model)","DeepSeek 官方请求按官方峰谷价；SiliconFlow 按自身人民币公开价；DigitalOcean 按美元公开价 × USD/CNY 汇率折算人民币；千问暂不计费；AMD GPU Cloud DeepSeek V4 Flash 免费按 ¥0。模型名与 API 服务商均以请求参数为准。":"DeepSeek official requests use official peak/valley prices; SiliconFlow uses its own public CNY prices; DigitalOcean uses public USD prices converted to CNY at the USD/CNY rate; Qwen is not billed yet; AMD GPU Cloud DeepSeek V4 Flash is free at ¥0. Model names and API providers follow the actual request parameters."," 次（高峰 ":" calls (peak "," / 空闲 ":" / off-peak "," · 缓存命中 ":" · cache hit "," · 输出 ":" · output "," · 推理 ":" · reasoning "," · 空闲消耗 ":" · off-peak cost ","‹ 上月":"‹ Prev month","下月 ›":"Next month ›","按消耗":"By cost","按调用数":"By calls","热力：":"Heatmap: ","低 → 高（":"low → high (","当日消耗":"Day cost","当日调用数":"Day calls","· 悬停查看详情，点击某天查看当日调用":"· hover for details, click a day for that day's calls","本月调用":"Calls this month","本月消耗":"Cost this month","本月暂无记录。":"No records this month.","日期":"Date","总消耗(自动)":"Total cost (auto)"," 调用明细":" call details"," 条 · ":" records · ","　｜　空闲 ":"　|　off-peak ","　｜　合计 ":"　|　total ","该日无记录。":"No records for this day.","消耗":"Cost","清除区间":"Clear range"," 条 · 输入·未命中 ":" records · input·miss "," · 缓存写入 ":" · cache write "," · 高峰消耗 ":" · peak cost ","该时间范围内没有记录。":"No records in this range.","‹ 上一页":"‹ Prev page","第 ":"Page "," 页 · 每页 ":" page · per page "," 条 · 共 ":" records · total ","下一页 ›":"Next page ›","新价格表生效时间未知":"New price schedule effective date unknown","新价格表（峰谷价）将于 ":"The new peak/valley price schedule takes effect ","（北京时间）生效，当前按旧价格表（基础价）计费":" (Beijing time); currently billing on the old base-price schedule","新价格表（峰谷价）已生效（自 ":"New peak/valley schedule in effect (since "," 起）":")","今日 $ / ¥ 汇率":"Today's $/¥ rate","暂不可用":"Temporarily unavailable","汇率日期 ":"Rate date ","等待获取":"Pending fetch","汇率状态":"Rate status","缓存汇率":"Cached rate","最新可用":"Latest available","获取失败":"Fetch failed","刷新汇率":"Refresh rate","美元计价的 DigitalOcean 消耗按该 USD/CNY 汇率换算成人民币；人民币定价不受汇率影响。":"USD-denominated DigitalOcean costs are converted to CNY at this rate; CNY pricing is unaffected by it.","DeepSeek 官方 API 价格表":"DeepSeek Official API Price Table","本表仅用于 DeepSeek 官方 Provider；第三方 Provider 按下方已核验的自身价格计费，无法可靠匹配价格的模型按 0 统计。价格按官方公布固定，不可编辑。":"This table applies only to the official DeepSeek provider; third-party providers bill at their own verified prices below, and models without a reliable price match are counted as 0. Prices are fixed as officially published and are not editable.","峰谷价":"Peak/valley","基础价":"Base","自动模式：按调用时间自动选择计费档位——新价格表生效前的调用按基础价（旧价格表），生效后的调用按峰谷价（高峰时段 9:00–12:00、14:00–18:00 用高峰价，其余空闲时段用空闲价）。当前显示：":"Auto mode: the billing tier is chosen automatically by call time — calls made before the new schedule took effect use base (old) prices, calls after it use peak/valley prices (peak hours 9:00–12:00 & 14:00–18:00 at peak price, other hours at off-peak). Currently shown:","基础价表（旧价格表，新价格生效前）":"Base price table (old schedule, before new prices take effect)","峰谷价表（新价格表）":"Peak/valley table (new schedule)","。价格按官方公布固定，不可编辑。":". Prices are fixed as officially published and are not editable.","缓存命中（输入）":"Cache hit (input)","空闲 · 缓存命中":"Off-peak · cache hit","空闲 · 输入":"Off-peak · input","空闲 · 输出":"Off-peak · output","高峰 · 缓存命中":"Peak · cache hit","高峰 · 输入":"Peak · input","高峰 · 输出":"Peak · output","单位：元 / 百万 tokens。高峰时段（北京时间 9:00–12:00、14:00–18:00）用高峰价，其余空闲时段用空闲价（空闲价 = 高峰价的一半）。概览、用量日历、缓存命中列表中的消耗均已按高峰 / 空闲分列统计。":"Unit: CNY per million tokens. Peak hours (Beijing 9:00–12:00 & 14:00–18:00) use peak prices; other hours use off-peak (= half of peak). Costs in Overview, Usage Calendar and Cache Hit List are split into peak / off-peak accordingly.","第三方平台价格与覆盖状态（2026-08-19 核验）":"Third-party platform prices & coverage (verified 2026-08-19)","只把能与记录中的 provider/model 可靠匹配的价格用于自动计费；美元价格先按 USD 计算，再乘 USD/CNY 汇率统一折算成人民币。":"Only prices that reliably match a record's provider/model are used for automatic billing; USD prices are computed in USD first, then converted to CNY at the USD/CNY rate.","模型 / 状态":"Model / status","输入":"Input","说明":"Notes","已用于自动计费":"Used for auto billing","缓存命中价按 2026-08-03 起生效公告":"Cache-hit price per the notice effective 2026-08-03","按输入价":"Billed at input price","公告未单列缓存命中价":"The notice lists no separate cache-hit price","自动按 USD/CNY 汇率换算人民币并计费":"Auto-converted to CNY at the USD/CNY rate and billed","Alibaba / 千问":"Alibaba / Qwen","暂不计费":"Not billed yet","调用与 token 正常统计；费用暂按 ¥0":"Calls and tokens tracked normally; cost counted as ¥0 for now","免费；费用固定按 ¥0":"Free; cost fixed at ¥0","DigitalOcean 使用美元官方价并按最新可用 USD/CNY 汇率折算人民币；千问暂不计费；AMD GPU Cloud 的 DeepSeek V4 Flash 免费按 ¥0。":"DigitalOcean bills at official USD prices converted via the latest available USD/CNY rate; Qwen is not billed yet; AMD GPU Cloud DeepSeek V4 Flash is free at ¥0.","概览":"Overview","用量日历":"Usage Calendar","缓存命中列表":"Cache Hit List","价格表":"Price Table","已导出：":"Exported: ","导出失败：":"Export failed: ","未知错误":"unknown error","导出中…":"Exporting…","生成图片中…":"Rendering image…","画布不可用":"Canvas unavailable","生成图片失败：":"Image rendering failed: ","打开目录选择…":"Open directory picker…","导出目标：":"Export destination:","选择目录失败：":"Failed to pick directory: ","打开文件夹失败：":"Failed to open folder: ","读取文件…":"Reading file…","导入成功：新增 ":"Import OK: added "," 条，跳过重复 ":", skipped duplicates "," 条，忽略无效 ":", ignored invalid "," 条，现有共 ":", total now ","导入失败：":"Import failed: ","读取文件失败":"Failed to read file","用量与消耗":"Usage & Cost","记录插件激活后的每一次模型调用":"Logs every model call since the plugin was activated","当前 · 高峰时段":"Current · peak hours","当前 · 空闲时段":"Current · off-peak hours","新价格已生效":"New prices in effect","新价格未生效":"New prices not yet in effect","刷新":"Refresh","清空":"Clear","计价说明：DeepSeek 官方与 SiliconFlow 按人民币价格；DigitalOcean 按美元官方价 × USD/CNY 汇率折算人民币；千问暂不计费；AMD GPU Cloud DeepSeek V4 Flash 免费按 ¥0。":"Billing notes: DeepSeek official and SiliconFlow bill in CNY; DigitalOcean bills at official USD prices × USD/CNY rate; Qwen is not billed yet; AMD GPU Cloud DeepSeek V4 Flash is free at ¥0.","导出 CSV":"Export CSV","导出 JSON":"Export JSON","导出图片 (PNG)":"Export image (PNG)","打开目录":"Open folder","导出目标目录（留空 = 默认数据目录）":"Export target directory (empty = default data dir)","选择目录…":"Choose directory…","已恢复默认数据目录":"Restored default data directory","重置":"Reset","选择文件导入":"Choose a file to import","数据持久化：":"Data persistence:","（每次调用实时落盘，插件重启后自动恢复，最多保留 100000 条）":"(every call is written to disk live, restored automatically after restart, up to 100000 records)","持久化未启用：":"Persistence disabled:","未知原因":"unknown reason","DEEPSEEK_API_KEY（推理 Key）":"DEEPSEEK_API_KEY (inference key)","模型设置中 Provider ID 或显示名为 siliconflow 的提供商 API Key":"API key of the provider whose Provider ID or display name is siliconflow in model settings","打开 SiliconFlow API 密钥":"Open SiliconFlow API keys","账户级 dop_v1_ Personal Access Token；Read Only（api:read）或 billing:read":"Account-level dop_v1_ Personal Access Token; Read Only (api:read) or billing:read","创建 DigitalOcean Account API Token":"Create a DigitalOcean account API token","暂无公开余额 API，仅支持控制台查看":"No public balance API yet — console only","打开 AMD Developer Cloud":"Open AMD Developer Cloud","查询失败":"Query failed","无法读取凭据状态":"Cannot read credential status","请输入以 dop_v1_ 开头的 DigitalOcean 账户 PAT。":"Enter a DigitalOcean account PAT starting with dop_v1_.","保存失败":"Save failed","使用凭据 ":"Using credential ","来源 ":"source ","模型提供商 ":"Model provider ","端点 ":"Endpoint "," · 账户余额":" · account balance","USD · 预付款/信用余额":"USD · prepaid/credit balance","USD · 待结算金额":"USD · pending amount","USD · 当前无余额":"USD · no balance"," · 账户可用":" · account available"," · 账户不可用":" · account unavailable"," · 查询成功":" · query succeeded","剩余余额查询":"Remaining balance query","查询 DeepSeek、SiliconFlow 与 DigitalOcean 账户余额":"Query DeepSeek, SiliconFlow and DigitalOcean account balances","打开控制台":"Open console","查询中…":"Querying…","查询账单":"Query billing","查询余额":"Query balance","所需凭据：":"Required credentials:","仅支持控制台查看":"Console only","插件不会尝试未经公开文档确认的端点，也不会把 AMD 推理 Key 当作账单凭据。":"The plugin never probes endpoints unconfirmed by public docs, and never uses AMD inference keys as billing credentials.","请确认凭据类型正确、权限包含余额/账单读取，且网络可访问服务商官方 API。":"Please verify the credential type is correct, its permissions include balance/billing read, and the provider's official API is reachable from this network.","打开服务商凭据页面":"Open provider credentials page","正在查询余额…":"Querying balance…","可用":"Available","不可用":"Unavailable","账单更新时间":"Billing updated","查询时间":"Queried at","北京时间":"Beijing time","请创建账户级 Personal Access Token。可选 Read Only（api:read，包含 billing:read），或自定义 billing:read；不要使用 Gradient AI 推理 Key。":"Please create an account-level Personal Access Token. Choose Read Only (api:read incl. billing:read) or a custom scope with billing:read; do not use a Gradient AI inference key.","正在检查已保存的 Token…":"Checking saved token…","已保存：":"Saved: ","（来源：":" (source: ","无法读取 Token 状态：":"Cannot read token status: ","尚未保存 DIGITALOCEAN_TOKEN。":"DIGITALOCEAN_TOKEN is not saved yet.","输入新的 dop_v1_ Token 可替换（当前值已隐藏）":"Enter a new dop_v1_ token to replace it (current value hidden)","保存中…":"Saving…","保存并查询":"Save & query","当前 Token 来自只读环境变量；请在原来源中修改，页面不会覆盖它。":"The current token comes from a read-only environment variable; edit it at its source — this panel will not overwrite it.","Token 已安全保存；页面和 API 响应不会回传明文。":"Token stored securely; neither the page nor API responses echo it back.","SiliconFlow 查询说明":"SiliconFlow query notes","公开 API 已成功返回 ¥0.00。SiliconFlow 的 /v1/user/info 当前不返回代金券或历史用量；控制台可用总额可能非零。这里忠实展示 API 原始余额字段，不把它等同于控制台完整额度。":"The public API returned ¥0.00 successfully. SiliconFlow's /v1/user/info currently returns no vouchers or historical usage; the console total may still be non-zero. The raw balance field from the API is shown faithfully here and should not be read as the full console credit.","余额卡片展示 SiliconFlow /v1/user/info 公开 API 返回的原始余额字段。":"The balance card shows the raw balance field returned by SiliconFlow's public /v1/user/info API.","返回字段":"Returned fields","凭据读取规则":"Credential lookup rules","插件只检查“设置 → 模型”中 Provider ID 或显示名为 siliconflow 的提供商，并读取其 apiKeyEnv 对应的已保存 API Key。":"The plugin only checks providers in Settings → Models whose Provider ID or display name is siliconflow, and reads the saved API key referenced by their apiKeyEnv.","如果未找到提供商、未填写 API Key 或凭据引用失效，插件会停止查询并说明需要修复的配置；不会回退到其他服务商的 Key。":"If no such provider exists, the key is empty, or the credential reference is broken, the plugin stops querying and explains what needs fixing; it never falls back to another provider's key.","DigitalOcean 查询的是主账户 Billing API；DigitalOcean AI 推理 Key 与 AMD GPU Cloud 推理 Key 均不能直接查询账单。":"DigitalOcean queries the main account Billing API; DigitalOcean AI inference keys and AMD GPU Cloud inference keys cannot query billing directly.","\n调用 ":"\ncalls ","）\n输入·未命中 ":")\ninput·miss ","\n高峰消耗 ":"\npeak cost ","\n无记录":"\nno records","查询 DeepSeek、SiliconFlow、DigitalOcean 与百炼 Token Plan":"Query DeepSeek, SiliconFlow, DigitalOcean & Bailian Token Plan balances","百炼 Token Plan · 1 周配额":"Bailian Token Plan · 1-week quota","百炼 Token Plan 配额用量":"Bailian Token Plan quota usage","百炼 Token Plan":"Bailian Token Plan","本周配额":"This week's quota","开始":"Start","结束":"End","已用 ":"Used ","百炼 Token Plan 查询说明":"Bailian Token Plan query notes","本查询复用百炼 CLI（bl）的“控制台登录”OAuth token，不需要阿里云 AccessKey，权限面很小。":"This query reuses the console-login OAuth token from the Bailian CLI (bl); no Alibaba Cloud AccessKey is needed and the permission scope is minimal.","如尚未登录，请先在终端执行：bl auth login --console，然后回到本页点“查询余额”。":"If you haven't logged in yet, run: bl auth login --console, then come back and click 'Query balance'.","打开百炼 CLI 登录说明":"Open Bailian CLI login docs","百炼控制台 OAuth 登录（~/.bailian/config.json 的 access_token，用 bl auth login --console 生成）":"Bailian console OAuth login (access_token in ~/.bailian/config.json, created by bl auth login --console)","数据来自阿里云百炼控制台内部门户网关（zeldaHttp.apikeyMgr./tokenplan/personal/api/v2/usage），展示的是 1 周配额已用进度与周期。":"Data from Alibaba Cloud Bailian console portal gateway (zeldaHttp.apikeyMgr./tokenplan/personal/api/v2/usage), showing 1-week quota usage progress and period."};
+    var __T_EN = {"本期无消耗（所有用量为 0）":"No usage this turn (all zeros)","按模型":"By model","…及另外 ":"…plus "," 个模型（完整列表见「用量与消耗」页签）":" more models (full list in the Usage & Cost tab)","对话累计":"Conversation total","本轮明细":"This turn","本轮 token":"Turn tokens","本轮消耗":"Turn cost","本轮缓存命中率":"Turn hit rate","本轮耗时":"Turn duration","本轮无记录":"No records this turn","工具调用（内部）":"Tool calls (internal)"," 条内部调用":" internal calls","展开":"Expand","收起":"Collapse","详细记录请在「用量与消耗」页签查看":"Full details are in the Usage & Cost tab","获取中…":"Loading…","总 token":"Total tokens","Token 明细":"Token details","本次输出":"This output","总消耗":"Total cost","耗时":"Run time","缓存命中率":"Cache hit rate","关闭":"Close","无 Token 数据":"No token data","当前 · ":"Current · ","工作日高峰时段":"Weekday peak hours","工作日空闲时段":"Weekday off-peak hours","周末高峰时段":"Weekend peak hours","周末空闲时段":"Weekend off-peak hours","周一至周五 9:00–12:00、14:00–18:00":"Mon–Fri 9:00–12:00 & 14:00–18:00","工作日其余时间":"Other weekday hours","周末（周六、周日）全天按空闲价计费":"All weekend (Sat & Sun) billed at the off-peak rate","2026-08-23 前仍按原规则分峰谷":"Before 2026-08-23 the old peak/valley split still applies","高峰 / 空闲时段说明":"Peak / off-peak hours","工作日高峰":"Weekday peak","空闲时段":"Off-peak hours"," 周一至周五 9:00–12:00、14:00–18:00":" Mon–Fri 9:00–12:00 & 14:00–18:00"," 工作日其余时间，以及周末（周六、周日）全天（自 2026-08-23 起；此前仍按原规则分峰谷）":" Weekday off-peak hours, plus all weekend (Sat & Sun) since 2026-08-23 (before that, the old peak/valley split still applies)","工作日高峰 9:00–12:00、14:00–18:00 · 周末全天空闲价":"Weekday peak 9:00–12:00 & 14:00–18:00 · all-weekend off-peak rate","\n调用 ":"\ncalls ","）\n输入·未命中 ":")\ninput·miss ","\n高峰消耗 ":"\npeak cost ","\n无记录":"\nno records","自动模式：按调用时间自动选择计费档位——新价格表生效前的调用按基础价（旧价格表），生效后的调用按峰谷价（工作日高峰时段 9:00–12:00、14:00–18:00 用高峰价，其余时间——含周末全天——用空闲价）。当前显示：":"Auto mode: the billing tier is chosen automatically by call time — calls made before the new price schedule took effect are billed at base (old) prices; calls after it use peak/valley prices (weekday peak hours 9:00–12:00 & 14:00–18:00 at the peak price, all other times — including all weekend — at the off-peak price). Currently shown:","单位：元 / 百万 tokens。工作日高峰时段（北京时间 9:00–12:00、14:00–18:00）用高峰价，其余时间用空闲价（空闲价 = 高峰价的一半）；自 2026-08-23 起周末（周六、周日）全天按空闲价计费，新规则生效前仍按原规则。概览、用量日历、缓存命中列表中的消耗均已按高峰 / 空闲分列统计。":"Unit: CNY per million tokens. Weekday peak hours (Beijing time 9:00–12:00 & 14:00–18:00) use the peak price; all other times use the off-peak price (off-peak = half of peak). Since 2026-08-23, weekends (Sat & Sun) are billed at the off-peak rate all day; calls before the new rule still follow the old split. Costs in Overview, Usage Calendar and the Cache Hit List are split into peak / off-peak accordingly.","未知模型":"Unknown model","未知服务商":"Unknown provider","完成":"Done","工具调用":"Tool call","超长":"Overlong","错误":"Error","已中断":"Interrupted","超时":"Timeout","中断":"Interrupted"," · 中断 ":" · interrupted ","（未计费）":" (not billed)","本地统计与官方后台的差异":"Local stats vs official console","本面板统计的是插件在本地捕获的模型调用（按官方价格与峰谷时段计费）。与官方后台（platform.deepseek.com 用量页）相比，金额可能更低，常见原因：":"This panel shows calls captured locally by the plugin (billed at official prices with peak/off-peak rates). Compared with the official console (platform.deepseek.com usage page), the amount may be lower — common reasons:","① 中断/出错/超时的调用：官方仍按实际 token 计费，插件无法获取其用量，按 0 记录（明细中标「中断」）；":"① Interrupted/failed/timed-out calls: the console still bills their actual tokens, but the plugin cannot obtain their usage, so they are recorded as 0 (marked \"Interrupted\" in details);","② 账号下其他 API Key（如其它应用/脚本）的调用不经过 DeepSeek Harness，官方统计包含它们，本插件不包含；":"② Calls from other API keys on your account (e.g. other apps/scripts) do not pass through DeepSeek Harness — the console includes them, this plugin does not;","③ 如需精确对账，可在官方用量页导出月度账单 CSV 与本插件对比。":"③ For exact reconciliation, export the monthly billing CSV from the official usage page and compare it with this plugin.","当前记录中有 ":"Current records include "," 次中断调用（未计费）——对应官方后台已计费的部分。":" interrupted call(s) (not billed) — corresponding to the part already billed by the official console.","自动（生效前基础价 · 生效后峰谷价）":"Auto (base price before effective date · peak/valley after)","基础价格":"Base price","峰谷价格":"Peak/valley price","自动":"Auto","canvas 2d 不可用":"canvas 2D unavailable","用量与消耗报告":"Usage & Cost Report","生成时间 ":"Generated ","（北京） · 共 ":" (Beijing) · total "," 条调用 · 缓存命中率 ":" calls · cache-hit rate "," · 高峰价消耗 ":" · peak-rate cost "," · 空闲价消耗 ":" · off-peak cost "," · 合计 ":" · total "," · 报告仅含最近 ":" · report limited to the latest "," 条":" records","调用次数":"Calls","输入 · 未命中":"Input · miss","输入 · 缓存命中":"Input · cache hit","输出":"Output","高峰消耗":"Peak cost","空闲消耗":"Off-peak cost","总消耗 (峰谷价)":"Total cost (peak/valley)","缓存命中列表（共 ":"Cache hit list ("," 条）":" records)","时间(北京)":"Time (Beijing)","模型":"Model","输入·未命中":"Input·miss","缓存命中":"Cache hit","缓存写入":"Cache write","推理":"Reasoning","命中率":"Hit rate","时段":"Period","结束":"End","消耗(峰谷)":"Cost (peak/valley)","峰":"Peak","谷":"Valley","消耗表（按模型 · 峰谷价 · 高峰/空闲分列）":"Cost table (by model · peak/valley prices)","调用":"Calls","总消耗":"Total cost","合计":"Total","消耗表（按 API 服务商 × 模型 · 峰谷价）":"Cost table (by API provider × model · peak/valley)","API 服务商 / 模型":"API provider / model"," 个模型）":" models)","总费用合计":"Total cost overall","计价按请求的 API 服务商分别应用已核验价格；DigitalOcean 美元价按 USD/CNY 汇率折算；无法可靠映射的第三方价格按 ¥0。":"Billing applies verified prices per requesting API provider; DigitalOcean USD prices are converted at the USD/CNY rate; third-party prices that cannot be mapped reliably are counted as ¥0."," 个服务商":" providers","今天":"Today","近7天":"Last 7 days","近30天":"Last 30 days","全部":"All","自定义区间：":"Custom range:","至":"to","清除筛选":"Clear filters","当前范围：共 ":"Current range: ","（全部记录）":"(all records)"," · 总消耗 ":" · total cost ","高峰 ":"Peak "," · 空闲 ":" · off-peak ","命中率 ":"hit rate ","高峰时段 9:00–12:00、14:00–18:00":"Peak hours 9:00–12:00 & 14:00–18:00","其余空闲时段":"other hours are off-peak","高峰 + 空闲 · ":"Peak + off-peak · ","消耗表（按模型）":"Cost table (by model)","API 服务商":"API provider","高峰/空闲":"Peak/Off-peak","消耗明细（按 API 服务商 × 模型）":"Cost breakdown (by API provider × model)","DeepSeek 官方请求按官方峰谷价；SiliconFlow 按自身人民币公开价；DigitalOcean 按美元公开价 × USD/CNY 汇率折算人民币；千问暂不计费；AMD GPU Cloud DeepSeek V4 Flash 免费按 ¥0。模型名与 API 服务商均以请求参数为准。":"DeepSeek official requests use official peak/valley prices; SiliconFlow uses its own public CNY prices; DigitalOcean uses public USD prices converted to CNY at the USD/CNY rate; Qwen is not billed yet; AMD GPU Cloud DeepSeek V4 Flash is free at ¥0. Model names and API providers follow the actual request parameters."," 次（高峰 ":" calls (peak "," / 空闲 ":" / off-peak "," · 缓存命中 ":" · cache hit "," · 输出 ":" · output "," · 推理 ":" · reasoning "," · 空闲消耗 ":" · off-peak cost ","‹ 上月":"‹ Prev month","下月 ›":"Next month ›","按消耗":"By cost","按调用数":"By calls","热力：":"Heatmap: ","低 → 高（":"low → high (","当日消耗":"Day cost","当日调用数":"Day calls","· 悬停查看详情，点击某天查看当日调用":"· hover for details, click a day for that day's calls","本月调用":"Calls this month","本月消耗":"Cost this month","本月暂无记录。":"No records this month.","日期":"Date","总消耗(自动)":"Total cost (auto)"," 调用明细":" call details"," 条 · ":" records · ","　｜　空闲 ":"　|　off-peak ","　｜　合计 ":"　|　total ","该日无记录。":"No records for this day.","消耗":"Cost","清除区间":"Clear range"," 条 · 输入·未命中 ":" records · input·miss "," · 缓存写入 ":" · cache write "," · 高峰消耗 ":" · peak cost ","该时间范围内没有记录。":"No records in this range.","‹ 上一页":"‹ Prev page","第 ":"Page "," 页 · 每页 ":" page · per page "," 条 · 共 ":" records · total ","下一页 ›":"Next page ›","新价格表生效时间未知":"New price schedule effective date unknown","新价格表（峰谷价）将于 ":"The new peak/valley price schedule takes effect ","（北京时间）生效，当前按旧价格表（基础价）计费":" (Beijing time); currently billing on the old base-price schedule","新价格表（峰谷价）已生效（自 ":"New peak/valley schedule in effect (since "," 起）":")","今日 $ / ¥ 汇率":"Today's $/¥ rate","暂不可用":"Temporarily unavailable","汇率日期 ":"Rate date ","等待获取":"Pending fetch","汇率状态":"Rate status","缓存汇率":"Cached rate","最新可用":"Latest available","获取失败":"Fetch failed","刷新汇率":"Refresh rate","美元计价的 DigitalOcean 消耗按该 USD/CNY 汇率换算成人民币；人民币定价不受汇率影响。":"USD-denominated DigitalOcean costs are converted to CNY at this rate; CNY pricing is unaffected by it.","DeepSeek 官方 API 价格表":"DeepSeek Official API Price Table","本表仅用于 DeepSeek 官方 Provider；第三方 Provider 按下方已核验的自身价格计费，无法可靠匹配价格的模型按 0 统计。价格按官方公布固定，不可编辑。":"This table applies only to the official DeepSeek provider; third-party providers bill at their own verified prices below, and models without a reliable price match are counted as 0. Prices are fixed as officially published and are not editable.","峰谷价":"Peak/valley","基础价":"Base","自动模式：按调用时间自动选择计费档位——新价格表生效前的调用按基础价（旧价格表），生效后的调用按峰谷价（高峰时段 9:00–12:00、14:00–18:00 用高峰价，其余空闲时段用空闲价）。当前显示：":"Auto mode: the billing tier is chosen automatically by call time — calls made before the new schedule took effect use base (old) prices, calls after it use peak/valley prices (peak hours 9:00–12:00 & 14:00–18:00 at peak price, other hours at off-peak). Currently shown:","基础价表（旧价格表，新价格生效前）":"Base price table (old schedule, before new prices take effect)","峰谷价表（新价格表）":"Peak/valley table (new schedule)","。价格按官方公布固定，不可编辑。":". Prices are fixed as officially published and are not editable.","缓存命中（输入）":"Cache hit (input)","空闲 · 缓存命中":"Off-peak · cache hit","空闲 · 输入":"Off-peak · input","空闲 · 输出":"Off-peak · output","高峰 · 缓存命中":"Peak · cache hit","高峰 · 输入":"Peak · input","高峰 · 输出":"Peak · output","单位：元 / 百万 tokens。高峰时段（北京时间 9:00–12:00、14:00–18:00）用高峰价，其余空闲时段用空闲价（空闲价 = 高峰价的一半）。概览、用量日历、缓存命中列表中的消耗均已按高峰 / 空闲分列统计。":"Unit: CNY per million tokens. Peak hours (Beijing 9:00–12:00 & 14:00–18:00) use peak prices; other hours use off-peak (= half of peak). Costs in Overview, Usage Calendar and Cache Hit List are split into peak / off-peak accordingly.","第三方平台价格与覆盖状态（2026-08-19 核验）":"Third-party platform prices & coverage (verified 2026-08-19)","只把能与记录中的 provider/model 可靠匹配的价格用于自动计费；美元价格先按 USD 计算，再乘 USD/CNY 汇率统一折算成人民币。":"Only prices that reliably match a record's provider/model are used for automatic billing; USD prices are computed in USD first, then converted to CNY at the USD/CNY rate.","模型 / 状态":"Model / status","输入":"Input","说明":"Notes","已用于自动计费":"Used for auto billing","缓存命中价按 2026-08-03 起生效公告":"Cache-hit price per the notice effective 2026-08-03","按输入价":"Billed at input price","公告未单列缓存命中价":"The notice lists no separate cache-hit price","自动按 USD/CNY 汇率换算人民币并计费":"Auto-converted to CNY at the USD/CNY rate and billed","Alibaba / 千问":"Alibaba / Qwen","暂不计费":"Not billed yet","调用与 token 正常统计；费用暂按 ¥0":"Calls and tokens tracked normally; cost counted as ¥0 for now","免费；费用固定按 ¥0":"Free; cost fixed at ¥0","DigitalOcean 使用美元官方价并按最新可用 USD/CNY 汇率折算人民币；千问暂不计费；AMD GPU Cloud 的 DeepSeek V4 Flash 免费按 ¥0。":"DigitalOcean bills at official USD prices converted via the latest available USD/CNY rate; Qwen is not billed yet; AMD GPU Cloud DeepSeek V4 Flash is free at ¥0.","概览":"Overview","用量日历":"Usage Calendar","缓存命中列表":"Cache Hit List","价格表":"Price Table","已导出：":"Exported: ","导出失败：":"Export failed: ","未知错误":"unknown error","导出中…":"Exporting…","生成图片中…":"Rendering image…","画布不可用":"Canvas unavailable","生成图片失败：":"Image rendering failed: ","打开目录选择…":"Open directory picker…","导出目标：":"Export destination:","选择目录失败：":"Failed to pick directory: ","打开文件夹失败：":"Failed to open folder: ","读取文件…":"Reading file…","导入成功：新增 ":"Import OK: added "," 条，跳过重复 ":", skipped duplicates "," 条，忽略无效 ":", ignored invalid "," 条，现有共 ":", total now ","导入失败：":"Import failed: ","读取文件失败":"Failed to read file","用量与消耗":"Usage & Cost","记录插件激活后的每一次模型调用":"Logs every model call since the plugin was activated","当前 · 高峰时段":"Current · peak hours","当前 · 空闲时段":"Current · off-peak hours","新价格已生效":"New prices in effect","新价格未生效":"New prices not yet in effect","刷新":"Refresh","清空":"Clear","计价说明：DeepSeek 官方与 SiliconFlow 按人民币价格；DigitalOcean 按美元官方价 × USD/CNY 汇率折算人民币；千问暂不计费；AMD GPU Cloud DeepSeek V4 Flash 免费按 ¥0。":"Billing notes: DeepSeek official and SiliconFlow bill in CNY; DigitalOcean bills at official USD prices × USD/CNY rate; Qwen is not billed yet; AMD GPU Cloud DeepSeek V4 Flash is free at ¥0.","导出 CSV":"Export CSV","导出 JSON":"Export JSON","导出图片 (PNG)":"Export image (PNG)","打开目录":"Open folder","导出目标目录（留空 = 默认数据目录）":"Export target directory (empty = default data dir)","选择目录…":"Choose directory…","已恢复默认数据目录":"Restored default data directory","重置":"Reset","选择文件导入":"Choose a file to import","数据持久化：":"Data persistence:","（每次调用实时落盘，插件重启后自动恢复，最多保留 100000 条）":"(every call is written to disk live, restored automatically after restart, up to 100000 records)","持久化未启用：":"Persistence disabled:","未知原因":"unknown reason","DEEPSEEK_API_KEY（推理 Key）":"DEEPSEEK_API_KEY (inference key)","模型设置中 Provider ID 或显示名为 siliconflow 的提供商 API Key":"API key of the provider whose Provider ID or display name is siliconflow in model settings","打开 SiliconFlow API 密钥":"Open SiliconFlow API keys","账户级 dop_v1_ Personal Access Token；Read Only（api:read）或 billing:read":"Account-level dop_v1_ Personal Access Token; Read Only (api:read) or billing:read","创建 DigitalOcean Account API Token":"Create a DigitalOcean account API token","暂无公开余额 API，仅支持控制台查看":"No public balance API yet — console only","打开 AMD Developer Cloud":"Open AMD Developer Cloud","查询失败":"Query failed","无法读取凭据状态":"Cannot read credential status","请输入以 dop_v1_ 开头的 DigitalOcean 账户 PAT。":"Enter a DigitalOcean account PAT starting with dop_v1_.","保存失败":"Save failed","使用凭据 ":"Using credential ","来源 ":"source ","模型提供商 ":"Model provider ","端点 ":"Endpoint "," · 账户余额":" · account balance","USD · 预付款/信用余额":"USD · prepaid/credit balance","USD · 待结算金额":"USD · pending amount","USD · 当前无余额":"USD · no balance"," · 账户可用":" · account available"," · 账户不可用":" · account unavailable"," · 查询成功":" · query succeeded","剩余余额查询":"Remaining balance query","查询 DeepSeek、SiliconFlow 与 DigitalOcean 账户余额":"Query DeepSeek, SiliconFlow and DigitalOcean account balances","打开控制台":"Open console","查询中…":"Querying…","查询账单":"Query billing","查询余额":"Query balance","所需凭据：":"Required credentials:","仅支持控制台查看":"Console only","插件不会尝试未经公开文档确认的端点，也不会把 AMD 推理 Key 当作账单凭据。":"The plugin never probes endpoints unconfirmed by public docs, and never uses AMD inference keys as billing credentials.","请确认凭据类型正确、权限包含余额/账单读取，且网络可访问服务商官方 API。":"Please verify the credential type is correct, its permissions include balance/billing read, and the provider's official API is reachable from this network.","打开服务商凭据页面":"Open provider credentials page","正在查询余额…":"Querying balance…","可用":"Available","不可用":"Unavailable","账单更新时间":"Billing updated","查询时间":"Queried at","北京时间":"Beijing time","请创建账户级 Personal Access Token。可选 Read Only（api:read，包含 billing:read），或自定义 billing:read；不要使用 Gradient AI 推理 Key。":"Please create an account-level Personal Access Token. Choose Read Only (api:read incl. billing:read) or a custom scope with billing:read; do not use a Gradient AI inference key.","正在检查已保存的 Token…":"Checking saved token…","已保存：":"Saved: ","（来源：":" (source: ","无法读取 Token 状态：":"Cannot read token status: ","尚未保存 DIGITALOCEAN_TOKEN。":"DIGITALOCEAN_TOKEN is not saved yet.","输入新的 dop_v1_ Token 可替换（当前值已隐藏）":"Enter a new dop_v1_ token to replace it (current value hidden)","保存中…":"Saving…","保存并查询":"Save & query","当前 Token 来自只读环境变量；请在原来源中修改，页面不会覆盖它。":"The current token comes from a read-only environment variable; edit it at its source — this panel will not overwrite it.","Token 已安全保存；页面和 API 响应不会回传明文。":"Token stored securely; neither the page nor API responses echo it back.","SiliconFlow 查询说明":"SiliconFlow query notes","公开 API 已成功返回 ¥0.00。SiliconFlow 的 /v1/user/info 当前不返回代金券或历史用量；控制台可用总额可能非零。这里忠实展示 API 原始余额字段，不把它等同于控制台完整额度。":"The public API returned ¥0.00 successfully. SiliconFlow's /v1/user/info currently returns no vouchers or historical usage; the console total may still be non-zero. The raw balance field from the API is shown faithfully here and should not be read as the full console credit.","余额卡片展示 SiliconFlow /v1/user/info 公开 API 返回的原始余额字段。":"The balance card shows the raw balance field returned by SiliconFlow's public /v1/user/info API.","返回字段":"Returned fields","凭据读取规则":"Credential lookup rules","插件只检查“设置 → 模型”中 Provider ID 或显示名为 siliconflow 的提供商，并读取其 apiKeyEnv 对应的已保存 API Key。":"The plugin only checks providers in Settings → Models whose Provider ID or display name is siliconflow, and reads the saved API key referenced by their apiKeyEnv.","如果未找到提供商、未填写 API Key 或凭据引用失效，插件会停止查询并说明需要修复的配置；不会回退到其他服务商的 Key。":"If no such provider exists, the key is empty, or the credential reference is broken, the plugin stops querying and explains what needs fixing; it never falls back to another provider's key.","DigitalOcean 查询的是主账户 Billing API；DigitalOcean AI 推理 Key 与 AMD GPU Cloud 推理 Key 均不能直接查询账单。":"DigitalOcean queries the main account Billing API; DigitalOcean AI inference keys and AMD GPU Cloud inference keys cannot query billing directly.","\n调用 ":"\ncalls ","）\n输入·未命中 ":")\ninput·miss ","\n高峰消耗 ":"\npeak cost ","\n无记录":"\nno records","查询 DeepSeek、SiliconFlow、DigitalOcean 与百炼 Token Plan":"Query DeepSeek, SiliconFlow, DigitalOcean & Bailian Token Plan balances","百炼 Token Plan · 1 周配额":"Bailian Token Plan · 1-week quota","百炼 Token Plan 配额用量":"Bailian Token Plan quota usage","百炼 Token Plan":"Bailian Token Plan","本周配额":"This week's quota","开始":"Start","结束":"End","已用 ":"Used ","百炼 Token Plan 查询说明":"Bailian Token Plan query notes","本查询复用百炼 CLI（bl）的“控制台登录”OAuth token，不需要阿里云 AccessKey，权限面很小。":"This query reuses the console-login OAuth token from the Bailian CLI (bl); no Alibaba Cloud AccessKey is needed and the permission scope is minimal.","如尚未登录，请先在终端执行：bl auth login --console，然后回到本页点“查询余额”。":"If you haven't logged in yet, run: bl auth login --console, then come back and click 'Query balance'.","打开百炼 CLI 登录说明":"Open Bailian CLI login docs","百炼控制台 OAuth 登录（~/.bailian/config.json 的 access_token，用 bl auth login --console 生成）":"Bailian console OAuth login (access_token in ~/.bailian/config.json, created by bl auth login --console)","数据来自阿里云百炼控制台内部门户网关（zeldaHttp.apikeyMgr./tokenplan/personal/api/v2/usage），展示的是 1 周配额已用进度与周期。":"Data from Alibaba Cloud Bailian console portal gateway (zeldaHttp.apikeyMgr./tokenplan/personal/api/v2/usage), showing 1-week quota usage progress and period."};
 
     function __detectLang() {
       // 语言完全跟随宿主设置（通用设置 → 语言）：不再读 localStorage，避免旧选择覆盖系统语言
@@ -186,6 +186,7 @@ window.__ModuleLoader__.load({
       badgeHit: { display: "inline-block", padding: "2px 8px", borderRadius: 999, fontSize: fs(11), fontWeight: 500, background: "rgba(46,204,113,.18)", color: "#2ecc71", whiteSpace: "nowrap" },
       badgePeak: { display: "inline-block", padding: "2px 8px", borderRadius: 999, fontSize: fs(11), fontWeight: 500, background: "rgba(255,152,0,.18)", color: "#ff9800", whiteSpace: "nowrap" },
       badgeValley: { display: "inline-block", padding: "2px 8px", borderRadius: 999, fontSize: fs(11), fontWeight: 500, background: "rgba(90,140,255,.18)", color: "#5a8cff", whiteSpace: "nowrap" },
+      badgeWarn: { display: "inline-block", padding: "2px 8px", borderRadius: 999, fontSize: fs(11), fontWeight: 500, background: "rgba(231,76,60,.16)", color: "#e74c3c", whiteSpace: "nowrap" },
       costPeak: { color: "#e08700", fontWeight: 600, whiteSpace: "nowrap" },
       costOff: { color: "#3d6bd6", fontWeight: 600, whiteSpace: "nowrap" },
       err: { color: "#ff6b6b", fontSize: fs(12) },
@@ -193,6 +194,10 @@ window.__ModuleLoader__.load({
       note: { fontSize: fs(11), opacity: 0.55 },
       errbox: { border: "1px solid rgba(255,107,107,.4)", background: "rgba(255,107,107,.08)", borderRadius: 8, padding: 12, fontSize: fs(12) },
       errboxTitle: { fontWeight: 600, marginBottom: 4, color: "#ff6b6b" },
+      diffbox: { border: "1px solid rgba(255,152,0,.4)", background: "rgba(255,152,0,.07)", borderRadius: 8, padding: "8px 12px", marginTop: 10, fontSize: fs(11), lineHeight: 1.6, color: "inherit" },
+      diffboxTitle: { fontWeight: 600, marginBottom: 4, color: "#e08700" },
+      diffboxText: { opacity: 0.85, marginBottom: 2 },
+      diffboxWarn: { marginTop: 6, color: "#e74c3c", fontWeight: 600 },
       infobox: { border: "1px solid rgba(90,140,255,.35)", background: "rgba(90,140,255,.08)", borderRadius: 8, padding: 12, marginTop: 18, fontSize: fs(12) },
       infoboxTitle: { fontWeight: 600, marginBottom: 4, color: "#5a8cff" },
       actionLink: { display: "inline-block", marginTop: 8, color: "#3d6bd6", textDecoration: "underline", fontSize: fs(12) },
@@ -293,7 +298,7 @@ window.__ModuleLoader__.load({
       var s = String(p == null ? "" : p).trim();
       return s ? s.toLowerCase() : "unknown";
     };
-    var finishLabel = function (f) { if (f === "stop") return t("完成"); if (f === "tool-calls") return t("工具调用"); if (f === "max-tokens") return t("超长"); if (f === "error") return t("错误"); return f || "—"; };
+    var finishLabel = function (f) { if (f === "stop") return t("完成"); if (f === "tool-calls") return t("工具调用"); if (f === "max-tokens") return t("超长"); if (f === "error") return t("错误"); if (f === "aborted") return t("已中断"); if (f === "timeout") return t("超时"); return f || "—"; };
     var pct = function (a, b) { return b > 0 ? (a / b * 100).toFixed(1) + "%" : "—"; };
     // 周末统一按空闲价规则生效时间：北京时间 2026-08-23 00:00（周日）。此前（含周末）仍按原峰谷分段。
     var WEEKEND_FLAT_AT = Date.parse("2026-08-23T00:00:00+08:00");
@@ -542,10 +547,12 @@ var api = function (payload) {
       for (var r3 = 0; r3 < count; r3++) {
         var rec = records[r3];
         var hit = rec.cacheReadTokens || 0, miss = rec.inputTokens || 0, cost = rec.peakValleyCost || 0;
+        var intR = !!rec.interrupted;
         ctx.fillStyle = r3 % 2 === 1 ? "#fafbfc" : "#ffffff"; ctx.fillRect(P, y, W - 2 * P, rowH);
         var ty = y + Math.round(rowH * 0.68);
         ctx.textAlign = "left"; ctx.fillStyle = "#1c2733"; font(fBase, "400");
         ctx.fillText(fmtTime(rec.time), c1.time, ty); ctx.fillText(modelName(rec), c1.model, ty);
+        if (intR) { ctx.fillStyle = "#e74c3c"; ctx.fillText(t("中断"), c1.model + ctx.measureText(modelName(rec)).width + 12, ty); ctx.fillStyle = "#1c2733"; }
         ctx.textAlign = "right";
         ctx.fillText(fmtInt(miss), c1.missR, ty);
         ctx.fillStyle = "#22a45d"; ctx.fillText(fmtInt(hit), c1.hitR, ty);
@@ -557,8 +564,8 @@ var api = function (payload) {
         ctx.fillText(rec.reasoningTokens ? fmtInt(rec.reasoningTokens) : "—", c1.reasonR, ty);
         ctx.fillStyle = "#55617a"; ctx.fillText(pct(hit, hit + miss), c1.rateR, ty);
         ctx.fillStyle = rec.peak ? "#e08700" : "#3d6bd6"; ctx.fillText(rec.peak ? t("峰") : t("谷"), c1.periodR, ty);
-        ctx.fillStyle = "#55617a"; ctx.fillText(finishLabel(rec.finishReason), c1.endR, ty);
-        ctx.fillStyle = rec.peak ? "#e08700" : "#3d6bd6"; ctx.fillText(fmtMoney(cost), c1.costR, ty);
+        ctx.fillStyle = intR ? "#e74c3c" : "#55617a"; ctx.fillText(finishLabel(rec.finishReason), c1.endR, ty);
+        ctx.fillStyle = rec.peak ? "#e08700" : "#3d6bd6"; ctx.fillText(intR ? "—" : fmtMoney(cost), c1.costR, ty);
         y += rowH;
       }
 
@@ -679,10 +686,12 @@ var api = function (payload) {
         records.push(fr);
       }
       var totalCalls = records.length;
+      var totalInterrupted = 0;
       var totalHit = 0, totalMiss = 0, totalWrite = 0, totalOut = 0, totalReason = 0, totalCost = 0, peakTotal = 0, offTotal = 0, peakCalls = 0, offCalls = 0;
       var byModel = {}, byProvider = {}, byProviderModel = {};
       for (var i = 0; i < records.length; i++) {
         var r = records[i];
+        if (r.interrupted) totalInterrupted += 1;
         totalHit += r.cacheReadTokens || 0; totalMiss += r.inputTokens || 0; totalWrite += r.cacheWriteTokens || 0;
         totalOut += r.outputTokens || 0; totalReason += r.reasoningTokens || 0;
         var cost = costOf(r, regime);
@@ -820,7 +829,7 @@ var api = function (payload) {
           t("当前范围：共 ") + fmtInt(records.length) + t(" 条") + (preset === "all" && !fromDate && !toDate ? t("（全部记录）") : "") + t(" · 总消耗 ") + fmtMoney(totalCost) + "（" + costHint(regime) + "）"
         ),
         el("div", { style: st.cards },
-          el(Card, { label: t("调用次数"), value: fmtInt(totalCalls), hint: t("高峰 ") + fmtInt(peakCalls) + t(" · 空闲 ") + fmtInt(offCalls) }),
+          el(Card, { label: t("调用次数"), value: fmtInt(totalCalls), hint: t("高峰 ") + fmtInt(peakCalls) + t(" · 空闲 ") + fmtInt(offCalls) + (totalInterrupted > 0 ? t(" · 中断 ") + fmtInt(totalInterrupted) + t("（未计费）") : "") }),
           el(Card, { label: t("输入 · 未命中"), value: fmtInt(totalMiss), hint: "token" }),
           el(Card, { label: t("输入 · 缓存命中"), value: fmtInt(totalHit), hint: t("命中率 ") + hitRate }),
           el(Card, { label: t("输出"), value: fmtInt(totalOut), hint: "token" }),
@@ -1080,10 +1089,11 @@ var api = function (payload) {
                       el("tbody", null, selRecords.map(function (rr) {
                         var hit2 = rr.cacheReadTokens || 0, miss2 = rr.inputTokens || 0;
                         var cost2 = costOf(rr, regime);
+                        var int2 = !!rr.interrupted;
                         return el("tr", { key: rr.time },
                           el("td", { style: st.tdFirst }, fmtTime(rr.time)),
                           el("td", { style: st.tdWrap },
-                            el("div", null, modelName(rr)),
+                            el("div", null, modelName(rr), int2 ? el("span", { style: st.badgeWarn, marginLeft: 6 }, t("中断")) : null),
                             el("div", { style: st.sub }, providerName(rr.provider))
                           ),
                           el("td", { style: st.td }, fmtInt(miss2)),
@@ -1091,8 +1101,8 @@ var api = function (payload) {
                           el("td", { style: st.td }, fmtInt(rr.outputTokens || 0)),
                           el("td", { style: st.td }, rr.reasoningTokens ? fmtInt(rr.reasoningTokens) : "—"),
                           el("td", { style: st.td }, rr.peak ? el("span", { style: st.badgePeak }, t("峰")) : el("span", { style: st.badgeValley }, t("谷"))),
-                          el("td", { style: st.td }, finishLabel(rr.finishReason)),
-                          el("td", { style: st.td }, el("span", { style: rr.peak ? st.costPeak : st.costOff }, fmtMoney(cost2)))
+                          el("td", { style: st.td }, int2 ? el("span", { style: st.badgeWarn }, finishLabel(rr.finishReason)) : finishLabel(rr.finishReason)),
+                          el("td", { style: st.td }, int2 ? el("span", { style: st.costOff }, "—") : el("span", { style: rr.peak ? st.costPeak : st.costOff }, fmtMoney(cost2)))
                         );
                       }))
                     )
@@ -1193,10 +1203,11 @@ var api = function (payload) {
                     pageRows.map(function (rr3, idx) {
                       var hit3 = rr3.cacheReadTokens || 0, miss3 = rr3.inputTokens || 0;
                       var cost3 = costOf(rr3, regime);
+                      var int3 = !!rr3.interrupted;
                       return el("tr", { key: rr3.time + "-" + idx },
                         el("td", { style: st.tdFirst }, fmtTime(rr3.time)),
                         el("td", { style: st.tdWrap },
-                          el("div", null, modelName(rr3)),
+                          el("div", null, modelName(rr3), int3 ? el("span", { style: st.badgeWarn, marginLeft: 6 }, t("中断")) : null),
                           el("div", { style: st.sub }, (rr3.provider || "") + (rr3.purpose ? " · " + rr3.purpose : ""))
                         ),
                         el("td", { style: st.td }, fmtInt(miss3)),
@@ -1206,8 +1217,8 @@ var api = function (payload) {
                         el("td", { style: st.td }, rr3.reasoningTokens ? fmtInt(rr3.reasoningTokens) : "—"),
                         el("td", { style: st.td }, pct(hit3, hit3 + miss3)),
                         el("td", { style: st.td }, rr3.peak ? el("span", { style: st.badgePeak }, t("峰")) : el("span", { style: st.badgeValley }, t("谷"))),
-                        el("td", { style: st.td }, finishLabel(rr3.finishReason)),
-                        el("td", { style: st.td }, el("span", { style: rr3.peak ? st.costPeak : st.costOff }, fmtMoney(cost3)))
+                        el("td", { style: st.td }, int3 ? el("span", { style: st.badgeWarn }, finishLabel(rr3.finishReason)) : finishLabel(rr3.finishReason)),
+                        el("td", { style: st.td }, int3 ? el("span", { style: st.costOff }, "—") : el("span", { style: rr3.peak ? st.costPeak : st.costOff }, fmtMoney(cost3)))
                       );
                     }),
                     el("tr", null,
@@ -1473,6 +1484,8 @@ var api = function (payload) {
       }
 
       var records = data.records || [];
+      var interruptedTotal = 0;
+      for (var ixi = 0; ixi < records.length; ixi++) { if (records[ixi].interrupted) interruptedTotal += 1; }
       var dataPath = data.dataPath || "";
       var nowPeriod = periodNow(Date.now());
       var effectiveAt = data.effectiveAt || 0;
@@ -1502,6 +1515,16 @@ var api = function (payload) {
             }, t(sub.z));
           })
         ),
+        el("div", { style: st.diffbox },
+          el("div", { style: st.diffboxTitle }, t("本地统计与官方后台的差异")),
+          el("div", { style: st.diffboxText }, t("本面板统计的是插件在本地捕获的模型调用（按官方价格与峰谷时段计费）。与官方后台（platform.deepseek.com 用量页）相比，金额可能更低，常见原因：")),
+          el("div", { style: st.diffboxText }, t("① 中断/出错/超时的调用：官方仍按实际 token 计费，插件无法获取其用量，按 0 记录（明细中标「中断」）；")),
+          el("div", { style: st.diffboxText }, t("② 账号下其他 API Key（如其它应用/脚本）的调用不经过 DeepSeek Harness，官方统计包含它们，本插件不包含；")),
+          el("div", { style: st.diffboxText }, t("③ 如需精确对账，可在官方用量页导出月度账单 CSV 与本插件对比。")),
+          interruptedTotal > 0
+            ? el("div", { style: st.diffboxWarn }, t("当前记录中有 ") + fmtInt(interruptedTotal) + t(" 次中断调用（未计费）——对应官方后台已计费的部分。"))
+            : null
+        ),
         el(BillingHint, null),
         el("div", { style: st.note, marginTop: 6 },
           t("计价说明：DeepSeek 官方与 SiliconFlow 按人民币价格；DigitalOcean 按美元官方价 × USD/CNY 汇率折算人民币；千问暂不计费；AMD GPU Cloud DeepSeek V4 Flash 免费按 ¥0。")

package/lib/index.js

@@ -186,6 +186,7 @@ export default {
           inputTokens: r.inputTokens, outputTokens: r.outputTokens,
           cacheReadTokens: r.cacheReadTokens, cacheWriteTokens: r.cacheWriteTokens,
           reasoningTokens: r.reasoningTokens, finishReason: r.finishReason,
+          interrupted: !!r.interrupted,
           usdCnyRate: r.usdCnyRate || 0, fxDate: r.fxDate || '',
           modelKey: modelKey(r.model),
           baseCost: costFor(r, 'base'), peakValleyCost: costFor(r, 'peakValley'), autoCost: costFor(r, 'auto'),
@@ -254,7 +255,7 @@ export default {
       function buildDays() {
         const map = {}
         for (const r of records) {
-          if (!isRealCall(r)) continue;
+          if (!isRealCall(r)) continue;
           const key = bjKey(r.time)
           let d = map[key]
           if (!d) {
@@ -400,6 +401,7 @@ export default {
           cacheWriteTokens: toNum(raw.cacheWriteTokens),
           reasoningTokens: toNum(raw.reasoningTokens),
           finishReason: String(raw.finishReason || ''),
+          interrupted: !!raw.interrupted,
           usdCnyRate: toNum(raw.usdCnyRate),
           fxDate: String(raw.fxDate || '')
         }
@@ -559,13 +561,14 @@ export default {
                 yield chunk
               }
             } finally {
+              if (!sessionId) {
+                try { const ag = currentAgent(); if (ag && ag.session && ag.session.id) sessionId = String(ag.session.id) } catch (e) {}
+              }
+              if (provider === 'digital-ocean' || provider === 'digitalocean') {
+                try { await refreshFxRate(false) } catch (e) {}
+              }
+              const isDO = provider === 'digital-ocean' || provider === 'digitalocean'
               if (usage) {
-                if (!sessionId) {
-                  try { const ag = currentAgent(); if (ag && ag.session && ag.session.id) sessionId = String(ag.session.id) } catch (e) {}
-                }
-                if (provider === 'digital-ocean' || provider === 'digitalocean') {
-                  try { await refreshFxRate(false) } catch (e) {}
-                }
                 // 兼容性：DeepSeek 系 provider 的 usage 不单独上报 cacheWriteTokens
                 // （缓存写入发生在未命中时，即 cacheWriteTokens == inputTokens）。
                 // 当上游未提供该字段时，用未命中 token 数（inputTokens）兜底，避免
@@ -583,20 +586,42 @@ export default {
                   cacheWriteTokens: cacheWrite,
                   reasoningTokens: usage.reasoningTokens || 0,
                   finishReason,
-                  usdCnyRate: (provider === 'digital-ocean' || provider === 'digitalocean') ? (FX.rate || 0) : 0,
-                  fxDate: (provider === 'digital-ocean' || provider === 'digitalocean') ? (FX.date || '') : ''
+                  usdCnyRate: isDO ? (FX.rate || 0) : 0,
+                  fxDate: isDO ? (FX.date || '') : ''
+                })
+              } else if (isRealCall({ model, provider, time: startedAt })) {
+                // 兜底：harness 在流被中断（aborted / error / timeout）时不会产出
+                // usage chunk（usage 只在收到 [DONE] 哨兵后才由 adapter yield），
+                // 但官方后台仍会按该次请求的实际 token 计费。这里补记一条 0-token
+                // 的「中断调用」，使调用次数与官方口径对齐；token 与费用均为 0，
+                // 不会虚增消耗。interrupted 标记供客户端 UI 区分展示。
+                records.push({
+                  time: startedAt,
+                  model,
+                  provider,
+                  purpose,
+                  sessionId,
+                  inputTokens: 0,
+                  outputTokens: 0,
+                  cacheReadTokens: 0,
+                  cacheWriteTokens: 0,
+                  reasoningTokens: 0,
+                  finishReason: finishReason || 'aborted',
+                  interrupted: true,
+                  usdCnyRate: 0,
+                  fxDate: ''
                 })
-                if (records.length > MAX_RECORDS) records.splice(0, records.length - MAX_RECORDS)
-                try {
-                  const agent = currentAgent()
-                  const sp = ctx.get('sandboxPolicy')
-                  if (sp && typeof sp.resolve === 'function' && agent && agent.session) {
-                    const policy = sp.resolve({ session: agent.session })
-                    if (policy && policy.workspaceRoot) cachedPolicy = policy
-                  }
-                } catch (e) {}
-                ensureSessionRoot().then(persistNow).catch(() => {})
               }
+              if (records.length > MAX_RECORDS) records.splice(0, records.length - MAX_RECORDS)
+              try {
+                const agent = currentAgent()
+                const sp = ctx.get('sandboxPolicy')
+                if (sp && typeof sp.resolve === 'function' && agent && agent.session) {
+                  const policy = sp.resolve({ session: agent.session })
+                  if (policy && policy.workspaceRoot) cachedPolicy = policy
+                }
+              } catch (e) {}
+              ensureSessionRoot().then(persistNow).catch(() => {})
             }
           }
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@feiyang666/dsh-usage-plugin",
-  "version": "1.14.0",
+  "version": "1.14.1",
   "description": "DeepSeek Harness usage & cost tracker plugin: per-call token/cache-hit stats, peak/off-peak billing, DeepSeek balance query, CSV/JSON/PNG export with custom destination, and persistent local storage. Ships a host half plus a web client half in one npm package; installs into a DSH profile as a dsh.bundle with one command (dsh plugin --profile web add @feiyang666/dsh-usage-plugin).",
   "keywords": [
     "deepseek",
@@ -71,8 +71,7 @@
     "cordis.patch.yml",
     "README.md",
     "README.zh.md",
-    "CHANGELOG.md",
-    "RELEASE_NOTES.md"
+    "CHANGELOG.md"
   ],
   "scripts": {
     "wire": "node scripts/wire.js",

package/RELEASE_NOTES.md

@@ -1,46 +0,0 @@
-# Release Notes (English)
-
-The authoritative **English** release notes for `@feiyang666/dsh-usage-plugin`.
-
-Each new version's notes are added at the **top** of this file; the **[Version index](#version-index)** at the **bottom** is the default entry point for jumping to any release.
-
----
-
-## v1.14.0 (2026-08-27)
-
-This release focuses on **per-message token insights**, cleaner cost breakdowns, **system-level language sync**, and **mobile adaptation**.
-
-### Per-message "Turn tokens" popup
-
-After a turn finishes, the assistant message's footer action row gains a **"Turn tokens"** button. Clicking it opens a `Token Details` popup with two layers:
-
-- **Conversation total** — the whole session's total tokens, total cost, run time and cache-hit rate;
-- **This turn** — this output's tokens, turn tokens, turn cost, turn duration, turn cache-hit rate, a cache-hit bar, and per-model cards (each model's input·miss / cache hit / output (+ reasoning) plus its peak/off-peak cost split).
-
-Durations are shown as `Xm Ys`. Every usage record is tagged with the conversation (`sessionId`), so per-turn and per-conversation stats are computed independently.
-
-### Internal/tool calls grouped separately
-
-Internal/placeholder calls (e.g. `dsh2shell-*` with model `fake`) no longer clutter the model cost tables; they are collected in a collapsible **"Tool calls (internal)"** group below the cost breakdown, so only real model calls count in the cost tables.
-
-### Language follows the harness setting
-
-The in-panel language toggle is removed; the plugin now follows the harness's own language setting (General Settings → Language) instantly, with no `localStorage` override — this also fixes the "still English after switching back" bug.
-
-### Mobile adaptation
-
-Tables fit the screen width on ≤900px (no horizontal scrollbar), wide tables collapse the middle columns, and the popup adapts to the viewport.
-
-### Also fixed
-
-- The popup previously rendered off-screen inside transformed ancestors — now portaled to `document.body`;
-- Per-message windowing now reads the message's real step boundaries, so "this turn" and "conversation total" stats are truly independent;
-- Wide tables no longer wrap numbers into unreadable lines.
-
----
-
-## Version index
-
-| Version | Date | English notes |
-| --- | --- | --- |
-| v1.14.0 | 2026-08-27 | [What's New in v1.14.0](#v1140-2026-08-27) |