@gamegeek-saikel/dsh-cost-meter 0.2.0 → 0.3.0-alpha.1

package/README.md

@@ -44,7 +44,7 @@ Chat costs in DeepSeek pricing change over time (list prices, USD→CNY exchange
 | Cost anchoring | Append-only pricebook snapshots; step cost computed once at the event's own time |
 | Price sources | Manual override > official pricing page > built-in fallback > OpenRouter (fallback only, USD→CNY) > none |
 | Official page | Parses both the 2026-08-21 combined zh/en tables (including the `deepseek-v4-flash-vision-exp` column and the English UTC schedule) and the legacy split tables; the built-in historical list still anchors `single` when the page no longer carries one |
-| Peak pricing | 2026-08-17 00:00 Beijing rollout; peak/off-peak windows parsed from each page's own schedule (the zh and en pages may differ; fallback 09:00–12:00 / 14:00–18:00 Beijing), off-peak half price |
+| Peak pricing | 2026-08-17 00:00 Beijing rollout; peak windows apply Monday–Friday only (the zh and en pages may state different timezones; fallback 09:00–12:00 / 14:00–18:00 Beijing), weekends and all other hours off-peak at half price |
 | Cost formula | Uncached input + cache reads (hit rate) + cache writes (billed at uncached input rate) + output, per 1M tokens, CNY |
 | Account balance | Official `GET /user/balance`, cached 60 s, single in-flight request, trust-fenced route |
 | Subagent support | BFS over the live agent tree; conversation totals = main session + descendants |
@@ -72,7 +72,7 @@ The pricebook (`src/pricebook.ts`) is the durable price source, persisted on the
 
 - **Priority chain** — per canonical model key (`provider/model`, bare model, or the `flash`/`pro` pricing key for DeepSeek-family models): manual override > official page > built-in fallback > OpenRouter (fallback only, USD→CNY, cache reads at the configured discount) > none.
 - **Snapshot selection** — `snapshotForTime` picks the newest snapshot with `effectiveAt <= event time` (pre-install sessions anchor to the first snapshot once).
-- **Peak/off-peak** — before the 2026-08-17 rollout all steps price at the single list price; after it, the band is chosen by the EVENT's own time against the schedule of the pricebook's own page (the zh and en pages each parse their own peak windows — the redesigned English page states UTC — falling back to Beijing 09:00–12:00 / 14:00–18:00, everything else off-peak). When the combined page has no legacy single-price column, `single` keeps anchoring to the built-in historical list; a page that still carries a separate legacy table wins. Each step's band is anchored once at fold time, so the per-reply cards and chips always show the band that round was billed at — never the band of the moment you are looking.
+- **Peak/off-peak** — before the 2026-08-17 rollout all steps price at the single list price; after it, the band is chosen by the EVENT's own time against the schedule of the pricebook's own page (the zh and en pages each parse their own peak windows — the redesigned English page states UTC — falling back to Beijing 09:00–12:00 / 14:00–18:00, everything else off-peak). The official pages restrict peak windows to Monday–Friday, so Saturdays and Sundays are off-peak all day in the schedule's own timezone. When the combined page has no legacy single-price column, `single` keeps anchoring to the built-in historical list; a page that still carries a separate legacy table wins. Each step's band is anchored once at fold time, so the per-reply cards and chips always show the band that round was billed at — never the band of the moment you are looking.
 - **Immutable ledger** — the `sessionCost` projection (`src/session-cost-projection.ts`) folds `request/header` (model) and usage-carrying events into per-step rows; a second usage sample for the same (turn, step) replaces the first (same-step finalization, not a re-price), with O(1) incremental totals.
 
 ## Project Structure

package/README.zh-CN.md

@@ -44,7 +44,7 @@ DeepSeek 的价格随时间变化（官方价目表、USD→CNY 汇率、以及
 | 成本锚定 | 只追加价格簿快照；步成本在事件自身时刻只计算一次 |
 | 价格来源 | 手动覆盖 > 官方价格页 > 内置回退 > OpenRouter（仅回退，USD→CNY）> 无 |
 | 官方页适配 | 同时解析 2026-08-21 新版中英文页面（含 `deepseek-v4-flash-vision-exp` 第三列与英文 UTC 时段）和旧版分表；无旧单价的页面继续用内置历史单价锚定 |
-| 峰值定价 | 2026-08-17 00:00 北京生效；高峰/闲时窗口按各页面自己的时段表解析（中英文页面可能不同，缺省 09:00–12:00 / 14:00–18:00 北京），闲时为半价 |
+| 峰值定价 | 2026-08-17 00:00 北京生效；高峰窗口仅周一至周五（中英文页面时区可能不同，缺省 09:00–12:00 / 14:00–18:00 北京），周六、周日全天及其余时段为半价闲时 |
 | 成本公式 | 未命中输入 + 缓存命中（命中价）+ 缓存写入（按未命中输入价计）+ 输出，每百万 tokens，CNY |
 | 账户余额 | 官方 `GET /user/balance`，缓存 60 秒，单飞请求，路由带信任围栏 |
 | 子代理支持 | 沿活跃代理树 BFS；对话总花费 = 主会话 + 全部后代 |
@@ -72,7 +72,7 @@ DeepSeek 的价格随时间变化（官方价目表、USD→CNY 汇率、以及
 
 - **优先级链**——按规范模型键（`provider/model`、裸模型名，或 DeepSeek 系模型的 `flash`/`pro` 定价键）：手动覆盖 > 官方页面 > 内置回退 > OpenRouter（仅回退，USD→CNY，缓存读按配置折扣）> 无。
 - **快照选取**——`snapshotForTime` 取 `effectiveAt <= 事件时间` 的最新快照（安装前的会话一次性锚到首个快照）。
-- **峰值/闲时**——2026-08-17 上线前所有步按单一价目计费；上线后按事件自身时刻与该价格簿页面的时段表选档（中英文页面各自解析自己的高峰窗口，英文新版为 UTC，抓取失败时回退到北京 09:00–12:00 / 14:00–18:00，其余闲时）。新版合并表没有旧单价列时，`single` 继续锚定内置历史单价；页面若有独立旧表，则优先用它。每步的档位在折叠时锚定一次，逐回复卡片与回复小标签永远显示该轮对话当时被计价的档位，而不是查看时刻的档位。
+- **峰值/闲时**——2026-08-17 上线前所有步按单一价目计费；上线后按事件自身时刻与该价格簿页面的时段表选档（中英文页面各自解析自己的高峰窗口，英文新版为 UTC，抓取失败时回退到北京 09:00–12:00 / 14:00–18:00，其余闲时）。官方页面限定高峰仅周一至周五，因此北京时间周六、周日全天为闲时。新版合并表没有旧单价列时，`single` 继续锚定内置历史单价；页面若有独立旧表，则优先用它。每步的档位在折叠时锚定一次，逐回复卡片与回复小标签永远显示该轮对话当时被计价的档位，而不是查看时刻的档位。
 - **不可变账本**——`sessionCost` 投影（`src/session-cost-projection.ts`）把 `request/header`（模型）与携带用量的事件折叠为逐步记录；同一 (turn, step) 的第二次用量样本**替换**第一条（同一步终结，而非重新计价），总计以 O(1) 增量维护。
 
 ## 项目结构

package/lib/client.js

@@ -157,11 +157,13 @@ window.__ModuleLoader__.load({
 		const PEAK_PRICING_START_MS = Date.UTC(2026, 7, 16, 16, 0, 0);
 		/**
 		* Built-in fallback zh peak schedule: the official announcement's windows in
-		* Beijing time (peak 09:00-12:00 / 14:00-18:00, everything else off-peak).
+		* Beijing time (peak 09:00-12:00 / 14:00-18:00, Monday–Friday; everything
+		* else, including all of Saturday and Sunday, off-peak).
 		*/
 		const PEAK_SCHEDULE_ZH = {
 			timezone: "Asia/Shanghai",
-			ranges: [[9, 12], [14, 18]]
+			ranges: [[9, 12], [14, 18]],
+			weekdaysOnly: true
 		};
 		/** Built-in fallback USD prices from the English pricing page. */
 		const FALLBACK_CURRENT_USD = {
@@ -180,7 +182,9 @@ window.__ModuleLoader__.load({
 		/**
 		* Whether the given instant is a peak-pricing hour under one schedule: the
 		* zh fallback is Beijing time 09:00-12:00 and 14:00-18:00; each pricebook
-		* passes the schedule parsed from its own (zh or en) pricing page.
+		* passes the schedule parsed from its own (zh or en) pricing page. When the
+		* schedule is Monday–Friday only, Saturdays and Sundays in the schedule's
+		* timezone are off-peak all day.
 		* @param now - the instant to classify.
 		* @param schedule - the peak-hour schedule to classify against.
 		* @returns true during peak hours.
@@ -189,8 +193,13 @@ window.__ModuleLoader__.load({
 			const parts = new Intl.DateTimeFormat("en-US", {
 				timeZone: schedule.timezone,
 				hour: "numeric",
-				hour12: false
+				hour12: false,
+				weekday: "short"
 			}).formatToParts(now);
+			if (schedule.weekdaysOnly === true) {
+				const weekday = parts.find((part) => part.type === "weekday")?.value;
+				if (weekday === "Sat" || weekday === "Sun") return false;
+			}
 			const hour = Number(parts.find((part) => part.type === "hour")?.value);
 			if (Number.isNaN(hour)) return false;
 			return schedule.ranges.some(([start, end]) => hour >= start && hour < end);
@@ -314,28 +323,6 @@ window.__ModuleLoader__.load({
 			if (offPeakTotal === 0) return null;
 			return (peak.cacheReadPerMillion + peak.inputPerMillion + peak.outputPerMillion) / offPeakTotal;
 		}
-		/** CJK characters (the /1.5 density class). */
-		const CJK_RE = /[\u3400-\u4dbf\u4e00-\u9fff]/g;
-		/**
-		* Heuristic character→token estimate: CJK ÷ 1.5, everything else ÷ 4
-		* (the streaming estimate's output side).
-		* @param text - the accumulated text.
-		* @returns the estimated token count.
-		*/
-		function estimateTokens(text) {
-			const cjk = (text.match(CJK_RE) ?? []).length;
-			return Math.ceil(cjk / 1.5 + (text.length - cjk) / 4);
-		}
-		/**
-		* Tokens of the streamed text/reasoning blocks of a partial assistant.
-		* @param blocks - the partial's content blocks (text/reasoning carry `text`).
-		* @returns the estimated token count.
-		*/
-		function blocksOutputTokens(blocks) {
-			let tokens = 0;
-			for (const block of blocks ?? []) if ((block.kind === "text" || block.kind === "reasoning") && typeof block.text === "string" && block.text.length > 0) tokens += estimateTokens(block.text);
-			return tokens;
-		}
 		/**
 		* The cache-read ratio of the newest completed step (input side of the
 		* estimate). Absent history conservatively prices all input as uncached
@@ -810,44 +797,28 @@ window.__ModuleLoader__.load({
 		//#endregion
 		//#region src/client/AssistantCostChip.tsx
 		/**
-		* Per-reply cost chip: the anchored cost of ONE finalized reply. Rendered
-		* through the assistant-actions slot but visually joined to the reply's
-		* timing row (01:51 · 用时 · 首 token · tok/s): a flex `order` places it
-		* after the clock text, and the type matches the clock's (label-tertiary,
-		* 14px) with a dot separator, so it reads as part of the same metadata run.
-		* The `messageId` owner prop maps to the assistant node's (turn, step),
-		* which addresses the projection's step ledger — the cost is the anchored
-		* snapshot value, never a current-price recompute. Unpriced replies render
+		* Per-reply cost at the end of the completed Turn's timing row. Rendered
+		* through the `conversation.chat.turnTail` chain slot — the selector only
+		* accepts closed Turns, so the extension appears once the reply settled and
+		* sits directly after the timing facts (用时 · 首 token · tok/s), with a dot
+		* separator and the same label-tertiary type as the clock text.
+		*
+		* DSH 0.1.5 removed the per-message assistant-actions slot and replaced the
+		* legacy chat-node window with turn/step locations; the cost is therefore
+		* addressed per Turn: the anchored step ledger (`sessionCost` projection) is
+		* filtered to the Turn's own steps and summed. Every price is the anchored
+		* snapshot value, never a current-price recompute; an unpriced Turn renders
 		* `—` (the Cost tab explains why).
 		*/
-		/**
-		* Locate a finalized assistant node by its message id and return its
-		* (turn, step) coordinates, or null when the node is out of window.
-		* @param snapshot - the conversation snapshot.
-		* @param messageId - the addressed message identity.
-		* @returns the step coordinates, or null.
-		*/
-		function stepOfMessage(snapshot, messageId) {
-			const nodes = snapshot.chat.legacy.nodes;
-			for (let i = nodes.length - 1; i >= 0; i -= 1) {
-				const node = nodes[i];
-				if (node?.kind === "assistant" && node.messageId === messageId && node.turn !== void 0) return {
-					turn: node.turn,
-					step: node.step
-				};
-			}
-			return null;
+		/** The Turn's own ledger steps, in fold order. */
+		function stepsOfTurn(steps, turn) {
+			return steps?.filter((entry) => entry.turn === turn) ?? [];
 		}
-		/** The projection step for one (turn, step) coordinate, if the ledger has it. */
-		function stepOf(steps, turn, step) {
-			return steps?.find((entry) => entry.turn === turn && entry.step === step);
-		}
-		const AssistantCostChip = (0, react.memo)(function AssistantCostChip({ messageId, useSession, useProjection, t }) {
+		const AssistantCostChip = (0, react.memo)(function AssistantCostChip({ matched, useProjection, t }) {
 			const currency = displayCurrency(t);
-			const step = useSession((snapshot) => stepOfMessage(snapshot, messageId));
 			const costCny = useProjection("sessionCost");
 			const costUsd = useProjection("sessionCostUsd");
-			const ledger = step === null ? void 0 : stepOf((currency === "USD" ? costUsd : costCny)?.steps, step.turn, step.step);
+			const cost = currency === "USD" ? costUsd : costCny;
 			const [response, setResponse] = (0, react.useState)(null);
 			(0, react.useEffect)(() => {
 				let alive = true;
@@ -859,9 +830,11 @@ window.__ModuleLoader__.load({
 					alive = false;
 				};
 			}, [currency]);
-			if (ledger === void 0 || ledger.cost === null) {
-				if (ledger === void 0) return null;
-				const reason = ledger.unpricedReason === "NO_MODEL" ? t("reason.NO_MODEL") : t("reason.NO_PRICE");
+			const steps = stepsOfTurn(cost?.steps, matched.turn);
+			const priced = steps.filter((entry) => entry.cost !== null);
+			if (steps.length === 0) return null;
+			if (priced.length === 0) {
+				const reason = steps[0]?.unpricedReason === "NO_MODEL" ? t("reason.NO_MODEL") : t("reason.NO_PRICE");
 				return /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
 					className: AssistantCostChip_module_css_default.unpriced,
 					"data-cost-chip": true,
@@ -870,9 +843,12 @@ window.__ModuleLoader__.load({
 					children: t("chip.unpriced")
 				});
 			}
-			const amount = `${currencySymbol(currency)}${formatMoney(ledger.cost)}`;
-			const band = ledger.band ?? bandForTime(ledger.time);
-			const ratio = peakOffPeakMultiplier(response?.pricebook?.current ?? null, ledger.provider, ledger.model);
+			const total = priced.reduce((sum, entry) => sum + (entry.cost ?? 0), 0);
+			const amount = `${currencySymbol(currency)}${formatMoney(total)}`;
+			const last = priced[priced.length - 1];
+			if (last === void 0) return null;
+			const band = last.band ?? bandForTime(last.time);
+			const ratio = peakOffPeakMultiplier(response?.pricebook?.current ?? null, last.provider, last.model);
 			const bandLabel = band === "peak" ? ratio === null ? t("band.peak") : t("price.peakRatio", { multiplier: formatMultiplier(ratio) }) : band === "offPeak" ? ratio === null ? t("band.offPeak") : t("price.offPeakRatio", { multiplier: formatMultiplier(1 / ratio) }) : null;
 			return /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("span", {
 				className: AssistantCostChip_module_css_default.root,
@@ -925,7 +901,7 @@ window.__ModuleLoader__.load({
 		* ratio (conservative all-uncached without history), output at the
 		* character heuristic — both at CURRENT snapshot prices.
 		*/
-		/** Read the conversation's running flag (the pill's estimate gate). */
+		/** Read the session's running flag (the pill's estimate gate). */
 		function runningOf(snapshot) {
 			return snapshot.running;
 		}
@@ -961,7 +937,7 @@ window.__ModuleLoader__.load({
 			const cost = currency === "USD" ? costUsd : costCny;
 			const pressure = useProjection("contextPressure");
 			const running = useSession(runningOf);
-			const partial = useSession((snapshot) => snapshot.partial);
+			const partial = null;
 			const [open, setOpen] = (0, react.useState)(false);
 			const [response, setResponse] = (0, react.useState)(null);
 			const wrapRef = (0, react.useRef)(null);
@@ -995,7 +971,7 @@ window.__ModuleLoader__.load({
 				if (!running || projection === void 0) return null;
 				const bucket = bucketAt(snapshot ?? void 0, model?.provider, model?.model ?? "unknown", Date.now(), schedule);
 				if (bucket === void 0) return null;
-				const stepEstimate = estimateCost(bucket, pressure?.projectedTokens ?? pressure?.pressureTokens ?? 0, cacheReadRatioOf(projection.steps), blocksOutputTokens(partial?.blocks));
+				const stepEstimate = estimateCost(bucket, pressure?.projectedTokens ?? pressure?.pressureTokens ?? 0, cacheReadRatioOf(projection.steps), 0);
 				return (combined?.cost ?? 0) + stepEstimate;
 			}, [
 				running,
@@ -1476,10 +1452,9 @@ window.__ModuleLoader__.load({
 				locale: NS,
 				label: () => t("view.cost")
 			}, CostView));
-			ctx.slots.inject("conversation.chat.assistant-actions", () => ctx.slots.register({
-				name: "conversation.chat.assistant-actions",
-				id: "cost",
-				order: 20,
+			ctx.slots.inject("conversation.chat.turnTail", () => ctx.slots.register({
+				name: "conversation.chat.turnTail",
+				select: (owner) => owner.turn.status === "closed" ? { turn: owner.turn.turn } : null,
 				locale: NS
 			}, AssistantCostChip));
 			ctx.slots.inject("conversation.session.header.utilities", () => ctx.slots.register({