@narumitw/pi-statusline 0.35.0 → 0.37.0

package/README.md

@@ -8,14 +8,15 @@ without setup and keeps useful context visible as the terminal narrows.
 A representative uncolored layout:
 
 ```text
-░▒▓ 🤖 sonnet-4 🧠 high 📁 pi-extensions 🌿 main ~2 🪟 ctx 42% 💸 $0.184
+░▒▓ 🤖 sonnet-4 🧠 high 📁 pi-extensions 🌿 main ~2 🪟 ctx 42.0%/200k 🕒 16:42
 ```
 
 ## ✨ Features
 
-- **Zero-config default:** model, thinking, workspace, Git/PR state, context use, and cost.
+- **Zero-config default:** model, thinking, workspace, Git/PR state, context use, and local time.
 - **Responsive:** removes lower-priority segments before important information gets clipped.
 - **Quiet when idle:** activity appears only while Pi is streaming or running tools.
+- **Native-aligned usage:** optional token, prompt-cache, subscription, and cost details.
 - **Easy choices:** three information levels and seven previewable color presets.
 - **Still flexible:** custom layouts, multiline rows, colors, labels, separators, and status icons stay
   available under **Advanced**.
@@ -73,11 +74,12 @@ preserved.
 | Level | Included segments |
 | --- | --- |
 | **Minimal** | `model cwd branch context` |
-| **Balanced** (default) | `model thinking cwd branch tools context cost` |
-| **Detailed** | `provider model thinking cwd branch tools context tokens cost time` |
+| **Balanced** (default) | `model thinking cwd branch tools context time` |
+| **Detailed** | `provider model thinking cwd branch tools context tokens cache cost time` |
 | **Custom** | Any other segment order, including explicit line breaks |
 
-The `tools` segment takes no space while idle.
+The `tools` segment takes no space while idle. `cache` takes no space when Pi has reported no cache
+reads or writes.
 
 ## 💬 Commands
 
@@ -100,7 +102,7 @@ lowest-priority segment, recomputes the powerline transitions, and repeats until
 Retention priority is highest to lowest:
 
 ```text
-context model branch tools cwd thinking cost provider tokens time turn brand
+context model branch tools cwd thinking cost provider cache tokens time turn brand
 ```
 
 Explicit `line_break` entries remain row boundaries. A single segment that is still too wide is
@@ -117,6 +119,20 @@ ANSI-safely truncated.
 - Context color changes to warning at 70% and error at 90%.
 - Git state is cached outside footer rendering and stale session results are ignored.
 
+### Usage and context
+
+- `context` renders one-decimal current usage and the model window, such as `2.4%/272k`. After
+  compaction it can temporarily render `?/272k` until the next valid assistant response.
+- `tokens`, `cache`, and `cost` total every usage-bearing session entry, matching Pi's native footer:
+  assistant messages, nested-LLM tool results, compactions, and branch summaries, including abandoned
+  branches retained in the session.
+- Cache tokens are `R<read>`, `W<write>`, and `CH<rate>`. `R` and `W` are cumulative; `CH` uses only
+  the latest assistant prompt: `cacheRead / (input + cacheRead + cacheWrite) * 100`.
+- Subscription-backed OAuth models and `kimi-coding` append `(sub)` to cost. The dollar value is
+  usage cost, not proof of an amount billed under a subscription.
+- Pi's public extension API does not expose the current auto-compaction toggle, so this footer cannot
+  reliably show the native `(auto)` marker.
+
 ## ⚙️ Settings
 
 The extension uses one user-level file:
@@ -156,7 +172,7 @@ A compact customization example:
   "palettePreset": "ocean",
   "density": "compact",
   "separator": "dot",
-  "segments": ["model", "thinking", "cwd", "branch", "context", "cost"],
+  "segments": ["model", "thinking", "cwd", "branch", "context", "cache", "cost"],
   "segmentText": {
     "context": { "prefix": "ctx ", "suffix": "" }
   },
@@ -218,7 +234,7 @@ Every successful change saves and applies immediately; closing the screen does n
 Available data segments:
 
 ```text
-brand provider model thinking cwd branch tools context tokens cost time turn
+brand provider model thinking cwd branch tools context tokens cache cost time turn
 ```
 
 Data segments must be unique. `line_break` may repeat when data segments separate occurrences, but
@@ -278,6 +294,7 @@ extensions/pi-statusline/
 │   ├── index.ts
 │   ├── statusline.ts
 │   ├── render.ts
+│   ├── usage.ts
 │   ├── powerline.ts
 │   ├── information-profiles.ts
 │   ├── commands.ts
@@ -299,7 +316,7 @@ extensions/pi-statusline/
 ## 🔎 Keywords
 
 Pi extension, Pi coding agent, statusline, Tokyo Night, powerline, responsive terminal footer,
-context usage, model status.
+context usage, prompt cache, cache hit rate, model status.
 
 ## 📄 License
 

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@narumitw/pi-statusline",
-	"version": "0.35.0",
+	"version": "0.37.0",
 	"description": "Pi extension that replaces the footer with an information-rich statusline.",
 	"type": "module",
 	"license": "MIT",

package/src/commands.ts

@@ -51,6 +51,7 @@ const SEGMENT_DESCRIPTIONS: Record<SegmentName, string> = {
 	tools: "Current tool and streaming activity",
 	context: "Current context-window usage",
 	tokens: "Session token totals",
+	cache: "Prompt-cache reads, writes, and latest hit rate",
 	cost: "Session cost",
 	time: "Current local time",
 	turn: "Current session turn count",

package/src/information-profiles.ts

@@ -8,7 +8,7 @@ export const INFORMATION_PROFILES: Readonly<
 	Record<InformationProfileName, readonly SegmentName[]>
 > = {
 	minimal: ["model", "cwd", "branch", "context"],
-	balanced: ["model", "thinking", "cwd", "branch", "tools", "context", "cost"],
+	balanced: ["model", "thinking", "cwd", "branch", "tools", "context", "time"],
 	detailed: [
 		"provider",
 		"model",
@@ -18,6 +18,7 @@ export const INFORMATION_PROFILES: Readonly<
 		"tools",
 		"context",
 		"tokens",
+		"cache",
 		"cost",
 		"time",
 	],

package/src/powerline.ts

@@ -49,6 +49,7 @@ const SEGMENT_RETENTION_PRIORITY: Readonly<Record<RenderSegment["name"], number>
 	thinking: 70,
 	cost: 60,
 	provider: 50,
+	cache: 45,
 	tokens: 40,
 	time: 30,
 	turn: 20,

package/src/presets/index.ts

@@ -37,6 +37,7 @@ const SEGMENT_BLOCKS: Record<SegmentName, PowerlineBlockName> = {
 	tools: "runtime",
 	context: "runtime",
 	tokens: "runtime",
+	cache: "runtime",
 	cost: "meter",
 	time: "meter",
 	turn: "meter",

package/src/render.ts

@@ -21,6 +21,7 @@ import {
 	type SegmentName,
 	type StatuslineConfig,
 } from "./types.js";
+import { type FooterUsageSummary, summarizeFooterUsage } from "./usage.js";
 
 type ThinkingLevel = ReturnType<ExtensionAPI["getThinkingLevel"]>;
 export interface RuntimeState extends ExtensionStatusRuntime {
@@ -31,11 +32,6 @@ export interface RuntimeState extends ExtensionStatusRuntime {
 	gitStatus?: GitStatusSummary;
 	requestRender?: () => void;
 }
-interface TokenTotals {
-	input: number;
-	output: number;
-	cost: number;
-}
 const GITHUB_PR_KEY = "github-pr";
 const GITHUB_PR_STATUS_KEYS = new Set([GITHUB_PR_KEY]);
 export function renderStatusline(
@@ -48,6 +44,7 @@ export function renderStatusline(
 ): string {
 	if (width <= 0) return "";
 
+	const usageSummary = summarizeFooterUsage(ctx.sessionManager.getEntries());
 	const rows: Array<{ configuredSegments: number; segments: RenderSegment[] }> = [
 		{ configuredSegments: 0, segments: [] },
 	];
@@ -60,7 +57,7 @@ export function renderStatusline(
 		const row = rows.at(-1);
 		if (!row) continue;
 		row.configuredSegments += 1;
-		const rendered = buildSegment(name, ctx, footerData, config, runtime);
+		const rendered = buildSegment(name, ctx, footerData, config, runtime, usageSummary);
 		if (rendered && rendered.text.length > 0) row.segments.push(rendered);
 	}
 
@@ -103,6 +100,7 @@ function buildSegment(
 	footerData: ReadonlyFooterDataProvider,
 	config: StatuslineConfig,
 	runtime: RuntimeState,
+	usageSummary: FooterUsageSummary,
 ): RenderSegment | undefined {
 	switch (name) {
 		case "brand":
@@ -138,25 +136,41 @@ function buildSegment(
 		}
 		case "context": {
 			const usage = ctx.getContextUsage();
-			const value =
+			const percentage =
 				usage?.percent === null || usage?.percent === undefined
 					? "?"
-					: `${usage.percent.toFixed(0)}%`;
-			return segment(name, value, config, contextColor(usage?.percent), "runtime");
+					: `${usage.percent.toFixed(1)}%`;
+			const contextWindow = usage?.contextWindow ?? ctx.model?.contextWindow ?? 0;
+			return segment(
+				name,
+				`${percentage}/${formatCount(contextWindow)}`,
+				config,
+				contextColor(usage?.percent),
+				"runtime",
+			);
 		}
 		case "tokens": {
-			const totals = getTokenTotals(ctx);
 			const value =
-				totals.input === 0 && totals.output === 0
+				usageSummary.input === 0 && usageSummary.output === 0
 					? "tok 0"
-					: `↑${formatCount(totals.input)} ↓${formatCount(totals.output)}`;
+					: `↑${formatCount(usageSummary.input)} ↓${formatCount(usageSummary.output)}`;
 			return segment(name, value, config, "accent", "runtime");
 		}
+		case "cache": {
+			if (usageSummary.cacheRead === 0 && usageSummary.cacheWrite === 0) return undefined;
+			const values: string[] = [];
+			if (usageSummary.cacheRead > 0) values.push(`R${formatCount(usageSummary.cacheRead)}`);
+			if (usageSummary.cacheWrite > 0) values.push(`W${formatCount(usageSummary.cacheWrite)}`);
+			if (usageSummary.latestCacheHitRate !== undefined) {
+				values.push(`CH${usageSummary.latestCacheHitRate.toFixed(1)}%`);
+			}
+			return segment(name, values.join(" "), config, "accent", "runtime");
+		}
 		case "cost": {
-			const totals = getTokenTotals(ctx);
+			const subscription = isSubscriptionBacked(ctx) ? " (sub)" : "";
 			return segment(
 				name,
-				totals.cost.toFixed(totals.cost >= 1 ? 2 : 3),
+				`${usageSummary.cost.toFixed(usageSummary.cost >= 1 ? 2 : 3)}${subscription}`,
 				config,
 				"accent",
 				"meter",
@@ -268,26 +282,12 @@ function compactPrState(value: string): string | undefined {
 	return undefined;
 }
 
-function getTokenTotals(ctx: ExtensionContext): TokenTotals {
-	const totals: TokenTotals = { input: 0, output: 0, cost: 0 };
-
-	for (const entry of ctx.sessionManager.getBranch()) {
-		if (entry.type !== "message" || entry.message.role !== "assistant") continue;
-
-		const usage = entry.message.usage as
-			| {
-					input?: number;
-					output?: number;
-					cost?: { total?: number };
-			  }
-			| undefined;
-
-		totals.input += usage?.input ?? 0;
-		totals.output += usage?.output ?? 0;
-		totals.cost += usage?.cost?.total ?? 0;
-	}
-
-	return totals;
+function isSubscriptionBacked(ctx: ExtensionContext): boolean {
+	const model = ctx.model;
+	return (
+		model !== undefined &&
+		(model.provider === "kimi-coding" || ctx.modelRegistry.isUsingOAuth(model))
+	);
 }
 
 export function formatCount(value: number): string {

package/src/settings.ts

@@ -67,6 +67,7 @@ export const DEFAULT_STATUSLINE_CONFIG: StatuslineConfig = {
 		tools: { prefix: "", suffix: "" },
 		context: { prefix: "🪟 ctx ", suffix: "" },
 		tokens: { prefix: "🔢 ", suffix: "" },
+		cache: { prefix: "📦 ", suffix: "" },
 		cost: { prefix: "💸 $", suffix: "" },
 		time: { prefix: "🕒 ", suffix: "" },
 		turn: { prefix: "🔁 #", suffix: "" },

package/src/types.ts

@@ -10,6 +10,7 @@ export const SEGMENT_NAMES = [
 	"tools",
 	"context",
 	"tokens",
+	"cache",
 	"cost",
 	"time",
 	"turn",

package/src/usage.ts

@@ -0,0 +1,53 @@
+import type { SessionEntry } from "@earendil-works/pi-coding-agent";
+
+interface UsageLike {
+	input?: number;
+	output?: number;
+	cacheRead?: number;
+	cacheWrite?: number;
+	cost?: { total?: number };
+}
+
+export interface FooterUsageSummary {
+	input: number;
+	output: number;
+	cacheRead: number;
+	cacheWrite: number;
+	cost: number;
+	latestCacheHitRate?: number;
+}
+
+export function summarizeFooterUsage(entries: readonly SessionEntry[]): FooterUsageSummary {
+	const totals: FooterUsageSummary = {
+		input: 0,
+		output: 0,
+		cacheRead: 0,
+		cacheWrite: 0,
+		cost: 0,
+	};
+
+	for (const entry of entries) {
+		let usage: UsageLike | undefined;
+		if (entry.type === "message" && entry.message.role === "assistant") {
+			usage = entry.message.usage;
+			const input = usage.input ?? 0;
+			const cacheRead = usage.cacheRead ?? 0;
+			const cacheWrite = usage.cacheWrite ?? 0;
+			const promptTokens = input + cacheRead + cacheWrite;
+			totals.latestCacheHitRate = promptTokens > 0 ? (cacheRead / promptTokens) * 100 : undefined;
+		} else if (entry.type === "message" && entry.message.role === "toolResult") {
+			usage = entry.message.usage;
+		} else if (entry.type === "compaction" || entry.type === "branch_summary") {
+			usage = entry.usage;
+		}
+		if (!usage) continue;
+
+		totals.input += usage.input ?? 0;
+		totals.output += usage.output ?? 0;
+		totals.cacheRead += usage.cacheRead ?? 0;
+		totals.cacheWrite += usage.cacheWrite ?? 0;
+		totals.cost += usage.cost?.total ?? 0;
+	}
+
+	return totals;
+}