dsh-connect-trae 4.0.1 → 4.2.0

package/CHANGELOG.md

@@ -1,5 +1,114 @@
 # Changelog
 
+## 4.2.0 (2026-10-21)
+
+> **模型行显示活动优惠徽标：只在上游此刻真在打折时出现。**
+>
+> Trae 把促销放在模型目录的同一次响应里——remote 目录的 `features.activity_discount`、wire 名单的
+> `display_contact_config.activity_discount`（都是 JSON 字符串）。上游发的是**规则**（折扣比例、原价、
+> 闲时时段与命名时区），不是文案。插件解析规则、在卡片**每次渲染时**按当前时钟判定，在模型名旁边画一枚
+> 胶囊：`5 折` / `闲时 5 折` / `限时 5 折`，悬停给出原价→现价与时段。请求数、请求身份、积分语义全部不变。
+>
+> 判定放在渲染时刻而不是解析时刻，是因为 `time_windows` 是「Asia/Shanghai 周一到周五 00:00–09:00、
+> 12:00–14:00、18:00–24:00，周末全天」这种墙钟时段：若在解析时挑定结论，夜里读到的「闲时 5 折」会在
+> 第二天早上继续挂着。实测同一份规则，时钟平移 +0/6/12/24/36h 分别画出 3/5/5/3/5 枚（2026-10-21 18:16）。
+
+### 新增
+
+- **模型行显示活动优惠徽标**（`feat(model)`）。解析 Trae 目录里的 `activity_discount` 规则，按渲染时刻判定后
+  在模型名旁边画一枚胶囊：`5 折` / `闲时 5 折` / `限时 5 折`，悬停给出原价→现价、时段与时区。
+  - 解析层只留规则、不挑徽标；规则随 `lastCatalog` 持久化，判定在卡片每次渲染时做，胶囊随窗口自动出现和消失。
+  - **先信上游的 `current`**（服务端算好的「此刻状态」）：`discount_type !== 'none'` 且折后价真比折前低才显示；
+    只有它报「没打折」时才用 `off_peak.time_windows` 按模型声明的时区自己算（跨零点不推断、`start` 含 `end` 不含）。
+  - 文案只认上游明说的：只有 `limited_discount` 写「限时」，其余按「补贴 / 闲时」；颜色是插件自己的选择。
+  - **不从倍率反推折扣**：那一行本来就显示 `x0.40`，再挂「5 折」是同一句话说两遍。
+  - 纯提示：不新增请求、不改计费语义；国际区（订阅制）目录里没有这份数据，显示路径同一条、上游补齐即生效。
+  - 详见 `docs/PROMO_BADGE.md`；复算脚本 `scripts/probe-promo.mjs`（只读，不消耗积分）。
+
+### 修复
+
+- **跨 group 去重不再丢促销规则**（`fix(solo-remote)`）。同一次目录响应里，同一个模型在部分 group 带
+  `activity_discount`、在部分 group 完全没有，而 group 顺序每次调用都可能不同（issue #23 的同一现象）。
+  `preferStrongerRow` 原先只按容量选行（等容量时先到先得），会把带规则的那一行丢掉，导致同一份上游数据
+  相邻两次调用一次 1 枚徽标、一次 0 枚。现在选行仍按容量，但把另一行携带的规则**并到赢家身上**；
+  没有改成「选带规则的行」，因为那会丢掉 1M Max 档——徽标不能以用户失去上下文预算为代价。
+- **活动徽标的两处边界收紧**（`fix(promo)`）。
+  - 两条规则评分并列（都没有 `current`、各 1 个窗口）时，原先留先到的那条，今晚有没有徽标就取决于网关的
+    分组顺序；改按规则内容的规范化 JSON（键名排序）取定，任意分组顺序结果一致。
+  - 上游 `time_zone` 写坏时，判定退回账号时区但悬停仍报那个读不出来的名字；改为报**实际用于判定的**时区。
+
+### 文档
+
+- 新增 `docs/PROMO_BADGE.md`（两处字段位置、逐字抓包、判定规则、与 WorkBuddy #91 的能力对照、实机复核输出）。
+- README / README.en 功能列表各加一条「活动优惠徽标」。
+- `docs/PROMO_BADGE.md` 按 2026-10-21 17:14 复测校正：wire 侧实测 214 行 / 26 行带规则 / 9 个模型
+  （旧记 54 行 / 8 行 / 3 行在打折）；补记判不了的窗口直接丢弃（不推断「全天」或跨零点）、
+  `limited` 分支的 `end_at` 未被读取（实测限时活动 `current` 报 `none`，所以「限时」暂不会真的出现）。
+
+### 测试
+
+- 新增 `tests/promo.spec.ts`（20 例）：`enable: false` 丢弃、JSON 字符串形态、两段价格字段（`consumption_rate` /
+  `after_consumption_rate`）、窗口过滤与丢弃（缺周几、空区间、越界）、时区判定、坏时区退回账号时区、
+  跨零点不外推、`current` 优先、非 `limited` 不写「限时」、无数字不画、以及「同一份规则在框内/框外判定不同」。
+- 新增 `tests/card-promo.spec.tsx`（6 例）：胶囊渲染在模型名旁边、三种类型的颜色类、悬停数字、无规则不渲染、
+  窗口外不渲染。
+- `tests/solo-remote.spec.ts` 补跨 group 去重 4 例：容量与规则并存、评分排序、并列按内容取定、
+  同一份数据换 group 顺序结果不变。
+- 全仓 **78 文件 951 例**通过；`tsc` 两个工程干净；`build` 通过（CJS 344.00 kB / ESM 426.33 kB）；
+  CI（ubuntu + windows）全绿。
+
+## 4.1.0 (2026-10-21)
+
+> **受限网络（企业网只放行少数域名）下「刷新模型 500 / 聊天 4001」的完整修复。**
+>
+> 链路是三段式的：`solo.trae.cn` 的 CONNECT 被代理拒绝（403）→ 目录 `fetch failed`（刷新 500）→
+> 模型丢失 `wireFunction` → 聊天按默认 `solo_work_lite` 发请求 → 模型不在该 function 的 config 名单里就
+> 4001 `config_name rejected`。本版对三段各有一处改动：目录 host 回退到聊天网关的同路径（带冷却重试）、
+> 目录分组的 function 兜底 `wireFunction`（只认可聊分组）、bridge 对「只带 function」的目录行保留
+> resolver 的 `config_name` 大小写（issue #38 的坑）。
+>
+> 合并 PR #43 前做了对抗性复现，修掉了四处会伤到真实用户的回归——**最关键的一处**：原实现把功能名
+> 属于 `solo_agent`/`builder_v3` 的目录行整行丢弃，而受限网络下那正是 9 个只在 `solo_agent` 分组里
+> 出现的模型（`gpt-5.5`、`gpt-5.6-*`、`gpt-6-astra`、`glm-5.2`、`kimi-k2.7-code`、`Dola-Seed-2.0-Code`、
+> `deepseek-v4-flash-0731`），它们正是 issue #19 修复要暴露的那批、且实测在默认 chat function 下
+> 都能正常应答。现在只把非可聊 function **降级清空**，行保留。
+
+### 修复
+
+- **受限网络下模型目录走聊天网关备用 host，目录分组 function 兜底 `wireFunction`**（`fix(cn)`）。修复分三层：
+  - **目录 host 回退**：官方 remote host 网络错误 / 超时 / 5xx 时，按序尝试聊天网关的同一路径
+    （CN 为 `trae-api-cn.mchost.guru/api/remote/v1`，实测同路径同响应结构）；401/403/404 等语义性拒绝**不回退**。
+    进程内记住最近一次**给出了非空目录**的 host。官方 host 失败后进入 5 分钟冷却：冷却期内先走记住的备用
+    host、官方 host 排在链尾（备用也挂了时仍有一次机会），冷却到期后官方 host 重新排第一 —— 只把官方 host
+    「排在第二」并不能让它被重试，因为循环在第一个成功处返回，所以必须显式冷却重试，恢复的网络才会在
+    一个冷却窗口内重新接管。空目录（200-but-empty）既不算成功（不记忆）也不终止链路（继续试下一个 base），
+    否则一个还没同步的镜像会永久堵死官方 host。回退只作用于 remote/v1 目录这一个 base；
+    `get_detail_param`、`llm_utils_chat` 照旧走各自原有 host。
+  - **目录分组 function 归因**：`fetchModels` 不再丢弃 `data.list[].function`——解析时把**可聊**分组的 function
+    附到每个模型（同一模型出现于多个分组时按 `TRAE_DIRECTORY_FUNCTIONS` 的实测优先级取，而非请求参数
+    `functions=` 的顺序；后者是历史列表，与 wire roster 自己的收集顺序不同，混用会让同一模型在「roster 有没有
+    返回」两种网络下拿到不同 function，`solo_agent_remote` 在 Remote 付费门后，优先它会多出 1005）。
+    `mergeTraeModelSources` 的 `wireFunction` 优先取 `get_detail_param` 的结论，目录分组兜底（roster 知道
+    账号的付费/可用状态，目录不知道）。同一 function 名重复出现时按名分桶保留全部，不再只留最后一组。
+  - **非可聊 function 只降级、不丢行**：`solo_agent`、`builder_v3` 是 IDE/Remote-agent roster，聊天通道实测
+    4011/4001，所以**不把这两个 function 写进 `wireFunction`**；模型本身**保留**，改由默认 chat function
+    调用。把它们整行丢掉等于在本次修复针对的受限网络里重新藏起 issue #19 的成果（详见上方导语）。
+    仅带 `wireFunction`（无 `wireConfigName`）的目录行，聊天请求会带上该 function，但 `config_name` 仍由
+    resolver 提供（目录 id 只是展示名，网关 `config_name` 大小写敏感，见 issue #38）。默认 function 与
+    fallback 模型常量未动。
+  - 已知取舍：模型被目录放在 `chat_v3` 分组下、但实际不可聊时，插件仅凭目录数据无法区分，仍会展示，
+    聊天失败走现有 4001 提示翻译（诊断入口见 `docs/ISSUE19_DIAGNOSIS.md`）。
+
+### 测试
+
+- 回归用例：host 回退（官方网络错误 → 备用 host 同路径同查询；401 不回退；5xx 回退；**空目录继续尝试下一个
+  base 且不被记住**；记住最近成功 host；**冷却到期后官方 host 重新排第一**；显式 baseUrl 钉住单 host）、
+  分组 function 归因（可聊分组优先；**同一 function 名重复出现时不丢模型**；容量来自后分组时保留先到先得的
+  function）、合并兜底与 `withChatCallableFunctions` 的**降级而非丢弃**、bridge 对 wireFunction-only 目录行
+  保留 resolver 的 `config_name`（大小写）、以及受限网络端到端（wire roster 为空时 9 个 `solo_agent`-only
+  模型仍然在目录里）。
+- 全仓 **921** 例通过，0 skipped；`tsc` 两个工程干净；`build` 通过；CI（ubuntu + windows）全绿。
+
 ## 4.0.1 (2026-10-19)
 
 > CN 区模型卡「最大输出」整列空白（issue #42）修复：上游**是**发布输出上限的——`get_detail_param` 的

package/README.en.md

@@ -37,6 +37,7 @@ A [DeepSeek Harness (DSH)](https://github.com/deepseek-ai/deepseek-harness) bund
 - **Any panel can be switched off on its own** — each tab carries its own checkbox (CN and international checked by default, enterprise not): unchecking **withdraws that provider from the DSH model picker entirely** (not merely hiding the tab) and stops all of its requests; checking it again restores it with the account, directory, picks, and context budgets intact. If you never use the international side, just uncheck it.
 - **Tabbed settings card** — the card's top carries a "Domestic / Enterprise / Global" tab bar; each tab holds its own account picker, usage overview, and model management. Account, directory, selection, and unsaved drafts are fully isolated per tab — switching accounts or refreshing models on one tab never touches the other side's runtime catalog or sessions. CN and enterprise share one CN credential pool but keep **separate account bindings and model picks**: the two rosters differ (`DeepSeek-V4.1-Flash` is listed only on the enterprise pool), so picks never overwrite each other.
 - **Multiplier in the model name** — model names show the credit multiplier in Trae's own menu format (e.g. `GLM-5.2 · x0.79`), updated with each directory refresh.
+- **Promotion badge** — Trae publishes its discount rules in the model directory (`activity_discount`: pay ratio, list price, off-peak windows with their time zone). The plugin judges them against **the clock at render time** and paints one capsule beside the model name (`5 折` / `off-peak 5 折` / `limited 5 折`), with the old price, the new price and the window on hover. Judging at render time is what keeps a night-time discount from still hanging there the next morning; when upstream says nothing is discounted right now no badge is painted, the multiplier is never used to guess a promotion, and no campaign without an end time is labelled "limited". It is information only: no request and no billing changes, and the international region (subscription) has no such data and therefore shows no badge. See [docs/PROMO_BADGE.md](docs/PROMO_BADGE.md).
 - **DSH local tool loop** — gets pending structured `tool_calls` from Trae `llm_utils_chat`, lets DSH execute its own local tools, then returns tool results to the model.
 - **Dual-region detection (CN / international)** — detects local sign-ins across Trae CN, TRAE SOLO CN, Trae, and TRAE SOLO installs. The region is derived from the credential's own `userRegion` claim (host suffix and edition label as fallbacks), with no manual switch.
 - **Region-isolated directories and selections** — CN and international each keep their own model directory, enabled picks, image opt-ins, and context budgets, and each provider reads only its own slot.

package/README.md

@@ -41,6 +41,7 @@
   > 池是**显式选择**：默认关闭。它不会替你在没打开池的情况下换账号——「换账号等于换人计费」，所以这件事只能由你点头。池打开后卡片会明确写出**当前是谁在付费**（Host 报告的是「下一个真正发送请求的账号」，不是保存的那个绑定）。
 
 - **倍率内嵌模型名** —— 模型名称按 Trae 自身菜单的格式显示积分倍率（如 `GLM-5.2 · x0.79`），倍率随目录刷新更新。
+- **活动优惠徽标** —— Trae 把促销规则发在目录里（`activity_discount`：折扣比例、原价、闲时时段与时区），插件按**当前时钟**判定后，在模型名旁边画一枚胶囊：`5 折` / `闲时 5 折` / `限时 5 折`，悬停写明原价、现价与时段。判定放在**渲染时刻**而不是解析时刻，所以夜间折扣不会在第二天早上继续挂着；上游说「此刻没打折」就不画，**不从倍率反推**，也不给没有结束时间的活动编「限时」。这是纯提示：不改请求、不改计费，国际区（订阅制）没有这项数据因此不显示。详见 [docs/PROMO_BADGE.md](docs/PROMO_BADGE.md)。
 - **DSH 本地工具循环** —— 通过 Trae `llm_utils_chat` 获取待执行的结构化 `tool_calls`，交由 DSH 自带的本地工具执行，再将工具结果回传模型。
 - **国内国际双区域自动识别** —— 自动发现 Trae CN / TRAE SOLO CN / Trae / TRAE SOLO 四个本地安装的登录账号；区域由凭证自带的 `userRegion` 声明自动判定（host 后缀与 edition 标签兜底），无需手动指定。
 - **目录与勾选按面板隔离** —— 三个面板各一套模型目录、勾选、图片开关与上下文预算，各自读各自的槽位。名录从**该面板所钉的那条网关**读取——原本钉在企业池的客户端去读消费池名录，会永远发现不了自己池的独占模型。

package/lib/client.js

@@ -374,6 +374,125 @@ window.__ModuleLoader__.load({
 			return tier !== void 0 && tier !== "" && !/^free$/i.test(tier);
 		}
 		//#endregion
+		//#region src/promo.ts
+		/** Zone used when upstream omits `off_peak.time_zone` (CN accounts). */
+		const TRAE_PROMO_DEFAULT_TIME_ZONE = "Asia/Shanghai";
+		const WEEKDAY_INDEX = {
+			Mon: 1,
+			Tue: 2,
+			Wed: 3,
+			Thu: 4,
+			Fri: 5,
+			Sat: 6,
+			Sun: 7
+		};
+		/** Wall clock in one named zone: ISO weekday (1–7) plus minutes since midnight. */
+		function zonedClock(now, timeZone) {
+			let parts;
+			try {
+				parts = new Intl.DateTimeFormat("en-US", {
+					timeZone,
+					weekday: "short",
+					hour: "2-digit",
+					minute: "2-digit",
+					hourCycle: "h23"
+				}).formatToParts(now);
+			} catch {
+				return;
+			}
+			let isoWeekday;
+			let hour;
+			let minute;
+			for (const part of parts) if (part.type === "weekday") isoWeekday = WEEKDAY_INDEX[part.value];
+			else if (part.type === "hour") hour = Number.parseInt(part.value, 10);
+			else if (part.type === "minute") minute = Number.parseInt(part.value, 10);
+			if (isoWeekday === void 0 || hour === void 0 || minute === void 0) return void 0;
+			if (!Number.isFinite(hour) || !Number.isFinite(minute)) return void 0;
+			return {
+				isoWeekday,
+				minutes: hour % 24 * 60 + minute
+			};
+		}
+		function clockOf(minutes) {
+			const hour = Math.floor(minutes / 60);
+			const minute = minutes % 60;
+			return `${String(hour).padStart(2, "0")}:${String(minute).padStart(2, "0")}`;
+		}
+		function runningWindow(discount, now, defaultTimeZone) {
+			const offPeak = discount.offPeak;
+			if (offPeak === void 0 || offPeak.windows.length === 0) return void 0;
+			const declaredZone = offPeak.timeZone ?? defaultTimeZone;
+			let usedZone = declaredZone;
+			let clock = zonedClock(now, declaredZone);
+			if (clock === void 0 && declaredZone !== defaultTimeZone) {
+				usedZone = defaultTimeZone;
+				clock = zonedClock(now, defaultTimeZone);
+			}
+			if (clock === void 0) return void 0;
+			const match = offPeak.windows.find((window) => window.weekdays.includes(clock.isoWeekday) && clock.minutes >= window.startMinute && clock.minutes < window.endMinute);
+			return match === void 0 ? void 0 : {
+				...match,
+				timeZone: usedZone
+			};
+		}
+		/** `subKey` is upstream's machine name; it is the only honest source for "限时". */
+		function kindOf(discount) {
+			const key = discount.subKey ?? "";
+			if (key.includes("off_peak")) return "offPeak";
+			if (key.includes("subsidy")) return "subsidy";
+			if (key.includes("limited")) return "limited";
+			const type = discount.current?.discountType ?? "";
+			if (type.includes("off")) return "offPeak";
+			if (type.includes("limited")) return "limited";
+			return "subsidy";
+		}
+		/** Pay ratio as upstream would phrase it; falls back to the rates it published. */
+		function discountOf(price) {
+			if (price.discountPercent !== void 0) return price.discountPercent;
+			if (price.beforeRate === void 0 || price.rate === void 0) return void 0;
+			if (price.beforeRate <= 0 || price.rate >= price.beforeRate) return void 0;
+			return Math.round(price.rate / price.beforeRate * 100);
+		}
+		function badgeOf(kind, price, discount, window) {
+			const discountPercent = discountOf(price);
+			if (discountPercent === void 0) return void 0;
+			const memberDiscountPercent = discount.member === void 0 ? void 0 : discountOf(discount.member);
+			return {
+				kind,
+				discountPercent,
+				percentOff: 100 - discountPercent,
+				...price.beforeRate === void 0 ? {} : { beforeRate: price.beforeRate },
+				...price.rate === void 0 ? {} : { rate: price.rate },
+				...memberDiscountPercent === void 0 ? {} : { memberDiscountPercent },
+				...window === void 0 ? {} : {
+					fromClock: clockOf(window.startMinute),
+					untilClock: clockOf(window.endMinute),
+					timeZone: window.timeZone
+				}
+			};
+		}
+		/**
+		* Decide the one badge (if any) a model shows right now.
+		*
+		* `now` is injected so the verdict is testable and so the card, which re-renders
+		* on its own poll, always judges against the current clock. `defaultTimeZone` is
+		* the account's zone, used only when upstream omits `off_peak.time_zone`.
+		*/
+		function selectTraePromoBadge(discount, now, defaultTimeZone = TRAE_PROMO_DEFAULT_TIME_ZONE) {
+			if (discount === void 0) return void 0;
+			const current = discount.current;
+			if (current !== void 0) {
+				if ((current.discountType ?? "none") !== "none" && (current.beforeRate === void 0 || current.rate === void 0 || current.rate < current.beforeRate)) {
+					const badge = badgeOf(kindOf(discount), current, discount);
+					if (badge !== void 0) return badge;
+				}
+			}
+			const window = runningWindow(discount, now, defaultTimeZone);
+			if (window !== void 0 && discount.offPeak !== void 0) {
+				if ((discount.current?.discountType ?? "none") === "none" || discountOf(discount.offPeak) !== void 0) return badgeOf("offPeak", discount.offPeak, discount, window);
+			}
+		}
+		//#endregion
 		//#region src/client/icon.ts
 		/**
 		* Plugin icon (LD brand logo, 64px) shared across the LaoDing plugin family.
@@ -691,6 +810,15 @@ window.__ModuleLoader__.load({
 .dsm-trae-model-probe-gated,.dsm-trae-model-probe-unsupported{color:var(--dsw-alias-state-warn-primary,#f59e0b);border:1px solid color-mix(in srgb,var(--dsw-alias-state-warn-primary,#f59e0b) 40%,transparent)}
 .dsm-trae-model-probe-auth{color:var(--dsw-alias-state-error-primary,#d92d20);border:1px solid color-mix(in srgb,var(--dsw-alias-state-error-primary,#d92d20) 40%,transparent)}
 .dsm-trae-model-probe-rate-limited,.dsm-trae-model-probe-unknown{color:var(--dsw-alias-label-tertiary,#999);border:1px solid color-mix(in srgb,var(--dsw-alias-label-tertiary,#999) 35%,transparent)}
+/* Promotion capsule (src/promo.ts). Unlike WorkBuddy's badge there is NO
+   upstream colour to honour — Trae publishes a pay ratio and nothing else — so
+   these three colours are the plugin's own choice, picked only so the kinds stay
+   distinguishable at a glance: red = subsidy price, orange = explicitly
+   "limited", blue = off-peak hours. */
+.dsm-trae-model-promo{flex:none;border-radius:999px;padding:1px 7px;font-size:10px;font-weight:500;line-height:15px;white-space:nowrap;border:1px solid transparent}
+.dsm-trae-model-promo-subsidy{color:var(--dsw-alias-state-error-primary,#d92d20);border-color:color-mix(in srgb,var(--dsw-alias-state-error-primary,#d92d20) 40%,transparent);background:color-mix(in srgb,var(--dsw-alias-state-error-primary,#d92d20) 10%,transparent)}
+.dsm-trae-model-promo-limited{color:var(--dsw-alias-state-warn-primary,#f59e0b);border-color:color-mix(in srgb,var(--dsw-alias-state-warn-primary,#f59e0b) 45%,transparent);background:color-mix(in srgb,var(--dsw-alias-state-warn-primary,#f59e0b) 12%,transparent)}
+.dsm-trae-model-promo-offPeak{color:var(--dsw-alias-brand-primary,#5686fe);border-color:color-mix(in srgb,var(--dsw-alias-brand-primary,#5686fe) 40%,transparent);background:color-mix(in srgb,var(--dsw-alias-brand-primary,#5686fe) 10%,transparent)}
 .dsm-trae-model-meta{display:flex;align-items:center;gap:7px 12px;flex-wrap:wrap;color:var(--dsw-alias-label-tertiary,#999);font-size:11px;line-height:16px}
 .dsm-trae-model-options{display:flex;align-items:center;justify-content:flex-end;gap:12px;flex:none}
 .dsm-trae-model-image{display:inline-flex;align-items:center;gap:4px;color:var(--dsw-alias-label-secondary,#c6c9d0);font-size:11px;line-height:16px;cursor:pointer}
@@ -2973,6 +3101,72 @@ window.__ModuleLoader__.load({
 				children: t(PROBE_LABELS[probe.verdict])
 			});
 		}
+		/**
+		* How each promotion kind is phrased.
+		*
+		* The wording is PLUGIN copy, not upstream's: Trae publishes a rule — a pay
+		* ratio plus optional cheap-hours windows — and never a label (contrast
+		* WorkBuddy's `badge.label`, see docs/PROMO_BADGE.md §3). Two things
+		* follow, and both are deliberate:
+		*
+		*  - only `limited_discount` says 限时 / "limited"; upstream gives no end time
+		*    for the other kinds, so claiming one would be inventing urgency;
+		*  - the number shown is upstream's own pay ratio (`discount: 50` ⇒ 五折 /
+		*    "50% off"), never a re-derived price.
+		*/
+		const PROMO_LABELS = {
+			limited: "row.promoLimited",
+			subsidy: "row.promoSubsidy",
+			offPeak: "row.promoOffPeak"
+		};
+		/** Pay ratio as the "折" figure a Chinese price tag uses: 50 → "5", 25 → "2.5". */
+		function promoZhe(discountPercent) {
+			const zhe = Math.round(discountPercent * 10) / 100;
+			return Number.isInteger(zhe) ? String(zhe) : zhe.toFixed(1);
+		}
+		/**
+		* The model row's promotion capsule, or nothing.
+		*
+		* Judged HERE, at paint time, not when the directory was parsed: the rule's
+		* windows are wall clocks in an upstream-named zone (00:00–09:00 on weekdays,
+		* …), so a verdict frozen into a saved catalog would still be shouting about
+		* last night's discount tomorrow afternoon. The card re-renders on its poll, so
+		* the capsule appears and disappears with the window.
+		*
+		* One capsule at most, and it sits BESIDE the name rather than inside it: the
+		* name is ellipsis-truncated, so a badge placed inside would be clipped away
+		* exactly on the long model names.
+		*/
+		function promoBadge(model, t, now) {
+			const badge = selectTraePromoBadge(model.activityDiscount, now);
+			if (badge === void 0) return null;
+			const zhe = promoZhe(badge.discountPercent);
+			const hints = [];
+			if (badge.beforeRate !== void 0 && badge.rate !== void 0) hints.push(t("row.promoHintRate", {
+				before: String(badge.beforeRate),
+				rate: String(badge.rate)
+			}));
+			if (badge.fromClock !== void 0 && badge.untilClock !== void 0 && badge.timeZone !== void 0) hints.push(t("row.promoHintWindow", {
+				from: badge.fromClock,
+				until: badge.untilClock,
+				zone: badge.timeZone
+			}));
+			if (badge.memberDiscountPercent !== void 0) hints.push(t("row.promoHintMember", {
+				zhe: promoZhe(badge.memberDiscountPercent),
+				percent: String(100 - badge.memberDiscountPercent)
+			}));
+			return /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+				className: `dsm-trae-model-promo dsm-trae-model-promo-${badge.kind}`,
+				title: hints.length === 0 ? t(PROMO_LABELS[badge.kind], {
+					zhe,
+					percent: String(badge.percentOff)
+				}) : hints.join(" · "),
+				children: t(PROMO_LABELS[badge.kind], {
+					zhe,
+					percent: String(badge.percentOff)
+				})
+			});
+		}
 		function dotStyle(status) {
 			return { background: status === "signed-in" ? "var(--dsw-alias-state-success-primary, #22a06b)" : status === "error" ? "var(--dsw-alias-state-error-primary, #d92d20)" : "var(--dsw-alias-label-dimmed, #9aa0a6)" };
 		}
@@ -3804,6 +3998,15 @@ window.__ModuleLoader__.load({
 			const savedImageIds = new Set(slotImageIds ?? legacyImageIds);
 			const draft = drafts[activeRegion];
 			const visibleModels = draft?.models ?? (status.status === "signed-in" ? status.models : []);
+			/**
+			* One clock reading for the whole model list.
+			*
+			* Promotion badges are judged against it (see `promoBadge`): reading the clock
+			* once per render keeps every row's verdict consistent with the others, and
+			* because the card re-renders on its poll, a badge still appears and vanishes
+			* with its window instead of sticking until the next manual refresh.
+			*/
+			const nowForPromo = /* @__PURE__ */ new Date();
 			const savedEnabledIds = status.status === "signed-in" ? new Set(status.enabledModelIds) : /* @__PURE__ */ new Set();
 			const activeEnabledIds = draft?.enabledIds ?? savedEnabledIds;
 			const activeImageIds = draft?.imageIds ?? savedImageIds;
@@ -4567,9 +4770,9 @@ window.__ModuleLoader__.load({
 																onChange: () => {
 																	toggleModel(model.id);
 																}
-															}), /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+															}), /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
 																className: "dsm-trae-model-copy",
-																children: /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
+																children: [/* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
 																	className: "dsm-trae-model-name",
 																	children: [
 																		model.name,
@@ -4579,7 +4782,7 @@ window.__ModuleLoader__.load({
 																		}),
 																		probeBadge(probeByRegion[activeRegion][model.id], t)
 																	]
-																})
+																}), promoBadge(model, t, nowForPromo)]
 															})]
 														}), /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
 															className: "dsm-trae-model-options",
@@ -5574,6 +5777,12 @@ window.__ModuleLoader__.load({
 			"row.probeRateLimited": "throttled",
 			"row.probeAuth": "sign-in rejected",
 			"row.probeUnknown": "not determined",
+			"row.promoLimited": "Limited {percent}% off",
+			"row.promoSubsidy": "{percent}% off",
+			"row.promoOffPeak": "Off-peak {percent}% off",
+			"row.promoHintRate": "was {before} credits/Mtoken, now {rate}",
+			"row.promoHintWindow": "off-peak hours {from}–{until} ({zone})",
+			"row.promoHintMember": "subscription price is a further {percent}% off",
 			"row.modelEnabled": "Enabled",
 			"row.discard": "Discard changes",
 			"row.save": "Save",
@@ -5894,6 +6103,12 @@ window.__ModuleLoader__.load({
 			"row.probeRateLimited": "被限流",
 			"row.probeAuth": "登录凭据被拒",
 			"row.probeUnknown": "本次未测出",
+			"row.promoLimited": "限时 {zhe} 折",
+			"row.promoSubsidy": "{zhe} 折",
+			"row.promoOffPeak": "闲时 {zhe} 折",
+			"row.promoHintRate": "原价 {before} 积分/百万 token，现价 {rate}",
+			"row.promoHintWindow": "闲时时段 {from}–{until}（{zone}）",
+			"row.promoHintMember": "会员价另有 {zhe} 折",
 			"row.modelEnabled": "启用模型",
 			"row.discard": "放弃修改",
 			"row.save": "保存",