@oh-my-pi/omp-stats 17.2.14 → 17.3.0

package/dist/client/styles.css

@@ -436,9 +436,6 @@
   .whitespace-nowrap {
     white-space: nowrap;
   }
-  .text-\[var\(--accent-cyan\)\] {
-    color: var(--accent-cyan);
-  }
   .text-\[var\(--text-muted\)\] {
     color: var(--text-muted);
   }
@@ -947,7 +944,7 @@
 }
 .stats-metric-primary-grid {
   display: grid;
-  grid-template-columns: repeat(4, 1fr);
+  grid-template-columns: repeat(5, minmax(0, 1fr));
   gap: 16px;
 }
 @media (max-width: 1023px) {

package/dist/types/shared-types.d.ts

@@ -24,8 +24,13 @@ export interface AggregatedStats {
     totalCacheReadTokens: number;
     /** Total cache write tokens */
     totalCacheWriteTokens: number;
-    /** Cache hit rate (0-1) */
+    /** Percentage of prompt input tokens served from cache (0-1). */
     cacheRate: number;
+    /**
+     * Prompt-input cost saved relative to billing the same tokens uncached
+     * (0-1; negative when cache writes cost more than reads save).
+     */
+    cacheSavings: number;
     /** Total cost */
     totalCost: number;
     /** Total premium requests */

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@oh-my-pi/omp-stats",
-	"version": "17.2.14",
+	"version": "17.3.0",
 	"description": "Local observability dashboard for pi AI usage statistics",
 	"homepage": "https://omp.sh",
 	"author": "Can Boluk",
@@ -39,9 +39,9 @@
 		"fmt": "biome format --write ."
 	},
 	"dependencies": {
-		"@oh-my-pi/pi-ai": "17.2.14",
-		"@oh-my-pi/pi-catalog": "17.2.14",
-		"@oh-my-pi/pi-utils": "17.2.14",
+		"@oh-my-pi/pi-ai": "17.3.0",
+		"@oh-my-pi/pi-catalog": "17.3.0",
+		"@oh-my-pi/pi-utils": "17.3.0",
 		"@tailwindcss/node": "^4.3.2",
 		"chart.js": "^4.5.1",
 		"lucide-react": "^1.24.0",

package/src/client/routes/ModelsRoute.tsx

@@ -322,7 +322,7 @@ function ModelsTable({
 								<div className="grid gap-4" style={{ gridTemplateColumns: "200px 1fr" }}>
 									<div className="space-y-4 text-sm">
 										<div>
-											<div className="text-[var(--text-primary)] font-medium mb-2">Quality</div>
+											<div className="text-[var(--text-primary)] font-medium mb-2">Efficiency</div>
 											<div className="space-y-1 text-[var(--text-secondary)]">
 												<div className="flex items-center justify-between">
 													<span>Error rate</span>
@@ -336,8 +336,18 @@ function ModelsTable({
 												</div>
 												<div className="flex items-center justify-between">
 													<span>Cache rate</span>
-													<span className="text-[var(--accent-cyan)]">
-														{(model.cacheRate * 100).toFixed(1)}%
+													<span className="font-mono">{(model.cacheRate * 100).toFixed(1)}%</span>
+												</div>
+												<div className="flex items-center justify-between">
+													<span>Cache savings</span>
+													<span
+														className={
+															model.cacheSavings < 0
+																? "text-[var(--accent-red)]"
+																: "text-[var(--accent-green)]"
+														}
+													>
+														{(model.cacheSavings * 100).toFixed(1)}%
 													</span>
 												</div>
 											</div>

package/src/client/routes/ProjectsRoute.tsx

@@ -87,8 +87,16 @@ export function ProjectsRoute({ active, range, refreshTrigger }: ProjectsRoutePr
 				key: "cacheRate",
 				header: "Cache Rate",
 				numeric: true,
+				render: (item: FolderRowView) => <span className="font-mono">{formatPercent(item.cacheRate)}</span>,
+			},
+			{
+				key: "cacheSavings",
+				header: "Cache Savings",
+				numeric: true,
 				render: (item: FolderRowView) => (
-					<span className="stats-text-success font-medium">{formatPercent(item.cacheRate)}</span>
+					<span className={`${item.cacheSavings < 0 ? "stats-text-danger" : "stats-text-success"} font-medium`}>
+						{formatPercent(item.cacheSavings)}
+					</span>
 				),
 			},
 			{
@@ -129,9 +137,13 @@ export function ProjectsRoute({ active, range, refreshTrigger }: ProjectsRoutePr
 					<div className="stats-mobile-card-value font-mono">{formatCost(item.totalCost)}</div>
 				</div>
 				<div>
-					<div className="stats-mobile-card-label">Cache</div>
+					<div className="stats-mobile-card-label">Cache Rate</div>
 					<div className="stats-mobile-card-value">{formatPercent(item.cacheRate)}</div>
 				</div>
+				<div>
+					<div className="stats-mobile-card-label">Cache Savings</div>
+					<div className="stats-mobile-card-value">{formatPercent(item.cacheSavings)}</div>
+				</div>
 				<div>
 					<div className="stats-mobile-card-label">Duration</div>
 					<div className="stats-mobile-card-value">{formatDurationMs(item.avgDuration)}</div>

package/src/client/styles.css

@@ -563,7 +563,7 @@
 
 .stats-metric-primary-grid {
   display: grid;
-  grid-template-columns: repeat(4, 1fr);
+  grid-template-columns: repeat(5, minmax(0, 1fr));
   gap: 16px;
 }
 

package/src/client/ui/MetricCluster.tsx

@@ -29,7 +29,17 @@ export function MetricCluster({ stats }: MetricClusterProps) {
 					<div className="stats-metric-label">Requests</div>
 					<div className="stats-metric-value">{formatInteger(stats.totalRequests)}</div>
 				</div>
-				<div className="stats-metric-card primary">
+				<div
+					className="stats-metric-card primary"
+					title="Prompt-input cost saved versus billing the same tokens uncached; cache writes can make this negative"
+				>
+					<div className="stats-metric-label">Cache Savings</div>
+					<div className="stats-metric-value">{formatPercent(stats.cacheSavings)}</div>
+				</div>
+				<div
+					className="stats-metric-card primary"
+					title="Prompt input served from cache: cache reads / (uncached input + cache reads)"
+				>
 					<div className="stats-metric-label">Cache Rate</div>
 					<div className="stats-metric-value">{formatPercent(stats.cacheRate)}</div>
 				</div>

package/src/db.ts

@@ -2,7 +2,7 @@ import { Database } from "bun:sqlite";
 import * as fs from "node:fs/promises";
 import type { Usage } from "@oh-my-pi/pi-ai";
 import type { GeneratedProvider } from "@oh-my-pi/pi-catalog/models";
-import { getBundledModel } from "@oh-my-pi/pi-catalog/models";
+import { calculateUncachedInputCost, getBundledModel } from "@oh-my-pi/pi-catalog/models";
 import { getConfigRootDir, getStatsDbPath } from "@oh-my-pi/pi-utils";
 import { classifyAgentType } from "./parser";
 import type {
@@ -33,7 +33,7 @@ import type {
 
 type ModelCost = { input: number; output: number; cacheRead: number; cacheWrite: number };
 type UsageCost = Usage["cost"];
-type CostTokens = Pick<Usage, "input" | "output" | "cacheRead" | "cacheWrite">;
+type CostTokens = Pick<Usage, "input" | "output" | "cacheRead" | "cacheWrite" | "orchestration">;
 
 const ZERO_USAGE_COST: UsageCost = {
 	input: 0,
@@ -53,6 +53,42 @@ interface CostBackfillRow {
 	cache_write_tokens: number;
 }
 
+interface NoCacheInputCostBackfillRow {
+	id: number;
+	provider: string;
+	model: string;
+	input_tokens: number;
+	cache_read_tokens: number;
+	cache_write_tokens: number;
+}
+
+interface AggregatedStatsRow {
+	total_requests: number;
+	failed_requests: number | null;
+	total_input_tokens: number | null;
+	total_output_tokens: number | null;
+	total_cache_read_tokens: number | null;
+	total_cache_write_tokens: number | null;
+	total_premium_requests: number | null;
+	total_cost: number | null;
+	total_cached_prompt_cost: number | null;
+	total_no_cache_input_cost: number | null;
+	avg_duration: number | null;
+	avg_ttft: number | null;
+	avg_tokens_per_second: number | null;
+	first_timestamp: number | null;
+	last_timestamp: number | null;
+}
+
+interface ModelStatsRow extends AggregatedStatsRow {
+	model: string;
+	provider: string;
+}
+
+interface FolderStatsRow extends AggregatedStatsRow {
+	folder: string;
+}
+
 let db: Database | null = null;
 
 const BACKFILL_COMPLETE = "complete";
@@ -112,6 +148,7 @@ export async function initDb(): Promise<Database> {
 			cost_cache_read REAL NOT NULL,
 			cost_cache_write REAL NOT NULL,
 			cost_total REAL NOT NULL,
+			cost_no_cache_input REAL,
 			agent_type TEXT NOT NULL DEFAULT 'main',
 			UNIQUE(session_file, entry_id)
 		);
@@ -183,6 +220,9 @@ export async function initDb(): Promise<Database> {
 	if (!messageColumns.some(column => column.name === "premium_requests")) {
 		db.run("ALTER TABLE messages ADD COLUMN premium_requests REAL NOT NULL DEFAULT 0");
 	}
+	if (!messageColumns.some(column => column.name === "cost_no_cache_input")) {
+		db.run("ALTER TABLE messages ADD COLUMN cost_no_cache_input REAL");
+	}
 	db.run("UPDATE messages SET premium_requests = 0 WHERE premium_requests IS NULL");
 	// Token-usage-by-agent: each message is classified main / subagent / advisor
 	// from its transcript path. A brand-new table gets the column from CREATE
@@ -265,6 +305,7 @@ export async function initDb(): Promise<Database> {
 	backfillPriorityPremiumRequests(db);
 	backfillAgentType(db);
 	backfillMissingCatalogCosts(db);
+	backfillNoCacheInputCosts(db);
 	backfillForkDuplicates(db);
 	return db;
 }
@@ -323,6 +364,18 @@ function resolveStoredCost(stats: MessageStats): UsageCost {
 	return calculateCatalogCost(stats.provider, stats.model, stats.usage) ?? storedCost ?? ZERO_USAGE_COST;
 }
 
+function calculateNoCacheInputCost(provider: string, modelId: string, tokens: CostTokens): number | null {
+	const cost = getCatalogCost(provider, modelId);
+	if (!cost) return null;
+	const promptInputTokens =
+		tokens.input +
+		tokens.cacheRead +
+		tokens.cacheWrite +
+		(tokens.orchestration?.input ?? 0) +
+		(tokens.orchestration?.cacheRead ?? 0);
+	return calculateUncachedInputCost(cost, promptInputTokens);
+}
+
 function backfillMissingCatalogCosts(database: Database): void {
 	const rows = database
 		.prepare(`
@@ -358,6 +411,31 @@ function backfillMissingCatalogCosts(database: Database): void {
 	applyBackfill();
 }
 
+function backfillNoCacheInputCosts(database: Database): void {
+	const rows = database
+		.prepare(`
+			SELECT id, provider, model, input_tokens, cache_read_tokens, cache_write_tokens
+			FROM messages
+			WHERE cost_no_cache_input IS NULL
+		`)
+		.all() as NoCacheInputCostBackfillRow[];
+	if (rows.length === 0) return;
+
+	const update = database.prepare("UPDATE messages SET cost_no_cache_input = ? WHERE id = ?");
+	const applyBackfill = database.transaction(() => {
+		for (const row of rows) {
+			const cost = calculateNoCacheInputCost(row.provider, row.model, {
+				input: row.input_tokens,
+				output: 0,
+				cacheRead: row.cache_read_tokens,
+				cacheWrite: row.cache_write_tokens,
+			});
+			update.run(cost ?? 0, row.id);
+		}
+	});
+	applyBackfill();
+}
+
 /**
  * Get the stored offset for a session file.
  */
@@ -406,9 +484,9 @@ export function insertMessageStats(stats: MessageStats[]): number {
 			session_file, entry_id, folder, model, provider, api, timestamp,
 			duration, ttft, stop_reason, error_message,
 			input_tokens, output_tokens, cache_read_tokens, cache_write_tokens, total_tokens, premium_requests,
-			cost_input, cost_output, cost_cache_read, cost_cache_write, cost_total, agent_type
+			cost_input, cost_output, cost_cache_read, cost_cache_write, cost_total, cost_no_cache_input, agent_type
 		)
-		SELECT ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?
+		SELECT ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?
 		WHERE NOT EXISTS (
 			SELECT 1 FROM messages
 			WHERE entry_id = ? AND timestamp = ? AND session_file <> ?
@@ -422,6 +500,7 @@ export function insertMessageStats(stats: MessageStats[]): number {
 	const insert = db.transaction(() => {
 		for (const s of stats) {
 			const cost = resolveStoredCost(s);
+			const noCacheInputCost = calculateNoCacheInputCost(s.provider, s.model, s.usage) ?? 0;
 			const result = stmt.run(
 				s.sessionFile,
 				s.entryId,
@@ -445,6 +524,7 @@ export function insertMessageStats(stats: MessageStats[]): number {
 				cost.cacheRead,
 				cost.cacheWrite,
 				cost.total,
+				noCacheInputCost,
 				s.agentType,
 				// `WHERE NOT EXISTS` binds: skip when a different session_file
 				// already holds this (entry_id, timestamp).
@@ -463,7 +543,7 @@ export function insertMessageStats(stats: MessageStats[]): number {
 /**
  * Build aggregated stats from query results.
  */
-function buildAggregatedStats(rows: any[]): AggregatedStats {
+function buildAggregatedStats(rows: AggregatedStatsRow[]): AggregatedStats {
 	if (rows.length === 0) {
 		return {
 			totalRequests: 0,
@@ -475,6 +555,7 @@ function buildAggregatedStats(rows: any[]): AggregatedStats {
 			totalCacheReadTokens: 0,
 			totalCacheWriteTokens: 0,
 			cacheRate: 0,
+			cacheSavings: 0,
 			totalCost: 0,
 			totalPremiumRequests: 0,
 			avgDuration: null,
@@ -492,6 +573,8 @@ function buildAggregatedStats(rows: any[]): AggregatedStats {
 	const totalInputTokens = row.total_input_tokens || 0;
 	const totalCacheReadTokens = row.total_cache_read_tokens || 0;
 	const totalPremiumRequests = row.total_premium_requests || 0;
+	const noCacheInputCost = row.total_no_cache_input_cost || 0;
+	const cachedPromptCost = row.total_cached_prompt_cost || 0;
 
 	return {
 		totalRequests,
@@ -506,6 +589,7 @@ function buildAggregatedStats(rows: any[]): AggregatedStats {
 			totalInputTokens + totalCacheReadTokens > 0
 				? totalCacheReadTokens / (totalInputTokens + totalCacheReadTokens)
 				: 0,
+		cacheSavings: noCacheInputCost > 0 ? (noCacheInputCost - cachedPromptCost) / noCacheInputCost : 0,
 		totalCost: row.total_cost || 0,
 		totalPremiumRequests,
 		avgDuration: row.avg_duration,
@@ -533,6 +617,10 @@ export function getOverallStats(cutoff?: number): AggregatedStats {
 			SUM(cache_write_tokens) as total_cache_write_tokens,
 			SUM(premium_requests) as total_premium_requests,
 			SUM(cost_total) as total_cost,
+			SUM(CASE WHEN cost_no_cache_input > 0
+				THEN cost_input + cost_cache_read + cost_cache_write
+				ELSE 0 END) as total_cached_prompt_cost,
+			SUM(cost_no_cache_input) as total_no_cache_input_cost,
 			AVG(duration) as avg_duration,
 			AVG(ttft) as avg_ttft,
 			AVG(CASE WHEN duration > 0 THEN output_tokens * 1000.0 / duration ELSE NULL END) as avg_tokens_per_second,
@@ -543,7 +631,7 @@ export function getOverallStats(cutoff?: number): AggregatedStats {
 	`);
 
 	const rows = hasCutoff ? stmt.all(cutoff) : stmt.all();
-	return buildAggregatedStats(rows as any[]);
+	return buildAggregatedStats(rows as AggregatedStatsRow[]);
 }
 /**
  * Get stats grouped by model.
@@ -564,6 +652,10 @@ export function getStatsByModel(cutoff?: number): ModelStats[] {
 			SUM(cache_write_tokens) as total_cache_write_tokens,
 			SUM(premium_requests) as total_premium_requests,
 			SUM(cost_total) as total_cost,
+			SUM(CASE WHEN cost_no_cache_input > 0
+				THEN cost_input + cost_cache_read + cost_cache_write
+				ELSE 0 END) as total_cached_prompt_cost,
+			SUM(cost_no_cache_input) as total_no_cache_input_cost,
 			AVG(duration) as avg_duration,
 			AVG(ttft) as avg_ttft,
 			AVG(CASE WHEN duration > 0 THEN output_tokens * 1000.0 / duration ELSE NULL END) as avg_tokens_per_second,
@@ -575,7 +667,7 @@ export function getStatsByModel(cutoff?: number): ModelStats[] {
 		ORDER BY total_requests DESC
 	`);
 
-	const rows = (hasCutoff ? stmt.all(cutoff) : stmt.all()) as any[];
+	const rows = (hasCutoff ? stmt.all(cutoff) : stmt.all()) as ModelStatsRow[];
 	return rows.map(row => ({
 		model: row.model,
 		provider: row.provider,
@@ -601,6 +693,10 @@ export function getStatsByFolder(cutoff?: number): FolderStats[] {
 			SUM(cache_write_tokens) as total_cache_write_tokens,
 			SUM(premium_requests) as total_premium_requests,
 			SUM(cost_total) as total_cost,
+			SUM(CASE WHEN cost_no_cache_input > 0
+				THEN cost_input + cost_cache_read + cost_cache_write
+				ELSE 0 END) as total_cached_prompt_cost,
+			SUM(cost_no_cache_input) as total_no_cache_input_cost,
 			AVG(duration) as avg_duration,
 			AVG(ttft) as avg_ttft,
 			AVG(CASE WHEN duration > 0 THEN output_tokens * 1000.0 / duration ELSE NULL END) as avg_tokens_per_second,
@@ -612,7 +708,7 @@ export function getStatsByFolder(cutoff?: number): FolderStats[] {
 		ORDER BY total_requests DESC
 	`);
 
-	const rows = (hasCutoff ? stmt.all(cutoff) : stmt.all()) as any[];
+	const rows = (hasCutoff ? stmt.all(cutoff) : stmt.all()) as FolderStatsRow[];
 	return rows.map(row => ({
 		folder: row.folder,
 		...buildAggregatedStats([row]),

package/src/index.ts

@@ -68,6 +68,7 @@ async function printStats(): Promise<void> {
 	console.log(`  Input Tokens: ${formatNumber(overall.totalInputTokens)}`);
 	console.log(`  Output Tokens: ${formatNumber(overall.totalOutputTokens)}`);
 	console.log(`  Cache Rate: ${formatPercent(overall.cacheRate)}`);
+	console.log(`  Cache Savings: ${formatPercent(overall.cacheSavings)}`);
 	console.log(`  Total Cost: ${formatCost(overall.totalCost)}`);
 	console.log(`  Premium Requests: ${formatNumber(normalizePremiumRequests(overall.totalPremiumRequests ?? 0))}`);
 	console.log(`  Avg Duration: ${overall.avgDuration !== null ? formatDuration(overall.avgDuration) : "-"}`);
@@ -80,7 +81,7 @@ async function printStats(): Promise<void> {
 		console.log("\nBy Model:");
 		for (const m of byModel.slice(0, 10)) {
 			console.log(
-				`  ${m.model}: ${formatNumber(m.totalRequests)} reqs, ${formatCost(m.totalCost)}, ${formatPercent(m.cacheRate)} cache`,
+				`  ${m.model}: ${formatNumber(m.totalRequests)} reqs, ${formatCost(m.totalCost)}, ${formatPercent(m.cacheRate)} cache rate, ${formatPercent(m.cacheSavings)} cache savings`,
 			);
 		}
 	}

package/src/shared-types.ts

@@ -25,8 +25,13 @@ export interface AggregatedStats {
 	totalCacheReadTokens: number;
 	/** Total cache write tokens */
 	totalCacheWriteTokens: number;
-	/** Cache hit rate (0-1) */
+	/** Percentage of prompt input tokens served from cache (0-1). */
 	cacheRate: number;
+	/**
+	 * Prompt-input cost saved relative to billing the same tokens uncached
+	 * (0-1; negative when cache writes cost more than reads save).
+	 */
+	cacheSavings: number;
 	/** Total cost */
 	totalCost: number;
 	/** Total premium requests */