@syncended/dsh-usage 0.1.0 → 0.2.1

package/README.md

@@ -46,9 +46,13 @@ dsh plugin --profile web remove @syncended/dsh-usage
 
 ## Pricing
 
-Cost is an estimate derived from provider-reported token buckets and USD-per-million-token rules. The plugin ships starter public-list-price rules for common OpenAI GPT-5, Anthropic Claude 4, and DeepSeek routes. Pricing changes over time and negotiated or subscription plans may not map to token billing, so override the rules for your environment.
+Cost is an estimate derived from provider-reported token buckets and USD-per-million-token rules. The built-in catalog was verified on **2026-08-26 UTC** and contains 120 price/tier entries compiled into 259 exact provider-route rules for OpenAI GPT, Anthropic Claude, Google Gemini, DeepSeek, Z.AI GLM, Moonshot/Kimi, xAI Grok, Mistral, Cohere, Alibaba Qwen, and MiniMax.
 
-Rules are matched in order against `provider/model`. `*` is the only wildcard. A route without a matching rule remains visible as **UNPRICED** and is excluded from estimated spend; the dashboard reports pricing coverage.
+See the [complete generated catalog](docs/pricing-catalog.md) for every model, price, condition, caveat, and official source URL. The engine handles prompt-length tiers and DeepSeek's recurring UTC peak/off-peak windows. It deliberately does not guess broad future model families: a route without a matching rule remains visible as **UNPRICED** and is excluded from estimated spend, while the dashboard reports pricing coverage.
+
+Rules are matched in order against `provider/model`; `*` is the only route wildcard. Prompt tiers use `minPromptTokens` / `maxPromptTokens`, while known promotions and retirements use inclusive `validFrom` / exclusive `validTo` ISO-8601 instants. Calls outside a known validity interval remain unpriced rather than silently inheriting an expired rate.
+
+Pricing changes over time, and batch/flex/priority service tiers, regional uplifts, negotiated rates, subscription plans, tool fees, and cache-storage duration may not map to token billing, so override the catalog for your environment when necessary. Current rules without an explicit validity interval remain current-list-price estimates rather than a historical invoice reconstruction.
 
 Override the bundle row in `$DSH_HOME/profiles/web/cordis.patch.yml`:
 
@@ -57,11 +61,18 @@ Override the bundle row in `$DSH_HOME/profiles/web/cordis.patch.yml`:
   config:
     scanConcurrency: 4
     pricing:
-      - route: openai-codex/gpt-5*
-        input: 1.25
-        output: 10
-        cacheRead: 0.125
-        cacheWrite: 1.25
+      - route: openai-codex/gpt-5.6-sol
+        minPromptTokens: 272000
+        input: 8
+        output: 30
+        cacheRead: 0.8
+        cacheWrite: 10
+      - route: openai-codex/gpt-5.6-sol
+        maxPromptTokens: 271999
+        input: 4
+        output: 20
+        cacheRead: 0.4
+        cacheWrite: 5
       - route: my-provider/private-model
         input: 0.8
         output: 3.2

package/dist/aggregate.d.ts

@@ -1,12 +1,8 @@
 import type { SessionEvent, SessionHeader } from '@deepseek-ai/dsh-session';
 import type { ModelPrice, SessionUsage, UsageRange, UsageSnapshot } from './types.js';
-/**
- * Built-in public-list-price estimates (USD / 1M tokens).
- * Route rules are deliberately overridable through plugin config.
- */
-export declare const DEFAULT_PRICING: ModelPrice[];
+export { DEFAULT_PRICING } from './pricing-catalog.js';
 /** Fold one durable session into billable provider usage samples. */
 export declare function extractSessionUsage(meta: SessionHeader, events: readonly SessionEvent[]): SessionUsage;
 export declare function dateKey(timestamp: number, timeZone: string): string;
-export declare function priceFor(route: string, pricing: readonly ModelPrice[]): ModelPrice | undefined;
+export declare function priceFor(route: string, pricing: readonly ModelPrice[], promptTokens?: number, timestamp?: number): ModelPrice | undefined;
 export declare function aggregateUsage(sessions: readonly SessionUsage[], pricing: readonly ModelPrice[], range: UsageRange, timeZone: string, now?: number, errors?: number): UsageSnapshot;

package/dist/aggregate.js

@@ -1,15 +1,5 @@
-/**
- * Built-in public-list-price estimates (USD / 1M tokens).
- * Route rules are deliberately overridable through plugin config.
- */
-export const DEFAULT_PRICING = [
-    { route: 'openai-codex/gpt-5', input: 1.25, output: 10, cacheRead: 0.125, cacheWrite: 1.25 },
-    { route: 'openai/gpt-5', input: 1.25, output: 10, cacheRead: 0.125, cacheWrite: 1.25 },
-    { route: 'anthropic/claude-sonnet-4-20250514', input: 3, output: 15, cacheRead: 0.3, cacheWrite: 3.75 },
-    { route: 'anthropic/claude-opus-4-20250514', input: 15, output: 75, cacheRead: 1.5, cacheWrite: 18.75 },
-    { route: 'deepseek/deepseek-chat', input: 0.28, output: 0.42, cacheRead: 0.028, cacheWrite: 0.28 },
-    { route: 'deepseek/deepseek-reasoner', input: 0.55, output: 2.19, cacheRead: 0.14, cacheWrite: 0.55 },
-];
+import { DEFAULT_PRICING } from './pricing-catalog.js';
+export { DEFAULT_PRICING } from './pricing-catalog.js';
 const DAY_MS = 86_400_000;
 const dateFormatterCache = new Map();
 function finiteToken(value) {
@@ -43,6 +33,8 @@ function assistantRoute(event) {
 export function extractSessionUsage(meta, events) {
     let currentRoute = null;
     const samples = new Map();
+    const stepStarts = new Map();
+    const compactionStarts = new Map();
     const seedLength = meta.seedLength ?? 0;
     for (let index = 0; index < events.length; index += 1) {
         const event = events[index];
@@ -58,11 +50,19 @@ export function extractSessionUsage(meta, events) {
         }
         if (index < seedLength)
             continue;
+        if (event.type === 'step/start') {
+            stepStarts.set(`${event.data.turn}:${event.data.step}`, event.time);
+            continue;
+        }
+        if (event.type === 'compaction/start') {
+            compactionStarts.set(String(event.data.compactionId), event.time);
+            continue;
+        }
         if (event.type === 'compaction/summary' && event.data.usage !== undefined) {
             const amount = buckets(event.data.usage);
             samples.set(`compaction:${event.seq}`, {
                 sessionId: String(meta.id),
-                timestamp: event.time,
+                timestamp: compactionStarts.get(String(event.data.compactionId)) ?? event.time,
                 provider: event.data.provider,
                 model: event.data.model,
                 ...amount,
@@ -87,9 +87,10 @@ export function extractSessionUsage(meta, events) {
             continue;
         const amount = buckets(usage);
         const route = assistantRoute(event) ?? currentRoute ?? { provider: 'unknown', model: 'unknown' };
-        samples.set(`${turn}:${step}`, {
+        const sampleKey = `${turn}:${step}`;
+        samples.set(sampleKey, {
             sessionId: String(meta.id),
-            timestamp: event.time,
+            timestamp: stepStarts.get(sampleKey) ?? event.time,
             provider: route.provider,
             model: route.model,
             ...amount,
@@ -130,12 +131,51 @@ function datesBetween(start, end) {
     const days = Math.max(0, Math.round((Date.parse(`${end}T00:00:00Z`) - Date.parse(`${start}T00:00:00Z`)) / DAY_MS));
     return Array.from({ length: days + 1 }, (_, index) => shiftDate(start, index));
 }
+const wildcardRegexCache = new Map();
+const pricingBoundaryCache = new Map();
 function wildcardMatches(pattern, value) {
-    const escaped = pattern.replace(/[.+?^${}()|[\]\\]/g, '\\$&').replace(/\*/g, '.*');
-    return new RegExp(`^${escaped}$`, 'i').test(value);
+    let regex = wildcardRegexCache.get(pattern);
+    if (regex === undefined) {
+        const escaped = pattern.replace(/[.+?^${}()|[\]\\]/g, '\\$&').replace(/\*/g, '.*');
+        regex = new RegExp(`^${escaped}$`, 'i');
+        wildcardRegexCache.set(pattern, regex);
+    }
+    return regex.test(value);
 }
-export function priceFor(route, pricing) {
-    return pricing.find((price) => wildcardMatches(price.route, route));
+function validityMatches(price, timestamp) {
+    if (price.validFrom === undefined && price.validTo === undefined)
+        return true;
+    const instant = timestamp ?? Date.now();
+    if (!Number.isFinite(instant))
+        return false;
+    const boundary = (value) => {
+        let parsed = pricingBoundaryCache.get(value);
+        if (parsed === undefined) {
+            parsed = Date.parse(value);
+            pricingBoundaryCache.set(value, parsed);
+        }
+        return parsed;
+    };
+    return (price.validFrom === undefined || instant >= boundary(price.validFrom)) &&
+        (price.validTo === undefined || instant < boundary(price.validTo));
+}
+function scheduleMatches(price, timestamp) {
+    if (price.utcWindows === undefined || price.utcWindows.length === 0)
+        return true;
+    if (timestamp === undefined || !Number.isFinite(timestamp))
+        return false;
+    const date = new Date(timestamp);
+    const day = date.getUTCDay();
+    const hour = date.getUTCHours();
+    const inside = price.utcWindows.some((window) => window.days.includes(day) && hour >= window.startHour && hour < window.endHour);
+    return price.outsideUtcWindows === true ? !inside : inside;
+}
+export function priceFor(route, pricing, promptTokens = 0, timestamp) {
+    return pricing.find((price) => wildcardMatches(price.route, route) &&
+        (price.minPromptTokens === undefined || promptTokens >= price.minPromptTokens) &&
+        (price.maxPromptTokens === undefined || promptTokens <= price.maxPromptTokens) &&
+        validityMatches(price, timestamp) &&
+        scheduleMatches(price, timestamp));
 }
 function tokensOf(value) {
     return value.input + value.output + value.cacheRead + value.cacheWrite;
@@ -191,18 +231,14 @@ export function aggregateUsage(sessions, pricing, range, timeZone, now = Date.no
     const dailySessions = new Map();
     const rangeSessionIds = new Set();
     const modelRows = new Map();
-    const priceCache = new Map();
     for (const session of sessions) {
         for (const record of session.records) {
             const date = dateKey(record.timestamp, timeZone);
             if (date < allStart || date > endDate)
                 continue;
             const route = `${record.provider}/${record.model}`;
-            let price = priceCache.get(route);
-            if (!priceCache.has(route)) {
-                price = priceFor(route, pricing);
-                priceCache.set(route, price);
-            }
+            const promptTokens = record.input + record.cacheRead + record.cacheWrite;
+            const price = priceFor(route, pricing, promptTokens, record.timestamp);
             const totalTokens = tokensOf(record);
             const pricedTokens = price === undefined ? 0 : totalTokens;
             const cost = costOf(record, price);

package/dist/index.d.ts

@@ -2,6 +2,7 @@ import { Context, Service } from '@deepseek-ai/cordis';
 import z from '@deepseek-ai/schemastery';
 import type { UsagePluginConfig, UsageRange, UsageSnapshot } from './types.js';
 export * from './aggregate.js';
+export * from './pricing-catalog.js';
 export * from './types.js';
 export declare const name = "usage";
 export declare const Config: z<UsagePluginConfig>;

package/dist/index.js

@@ -2,19 +2,42 @@ import { Context, Service } from '@deepseek-ai/cordis';
 import z from '@deepseek-ai/schemastery';
 import { aggregateUsage, DEFAULT_PRICING, extractSessionUsage } from './aggregate.js';
 import { createUsageHttpHandler } from './http.js';
+import { validatePricing } from './pricing-catalog.js';
 export * from './aggregate.js';
+export * from './pricing-catalog.js';
 export * from './types.js';
 export const name = 'usage';
 const API_PREFIX = '/api/usage';
+const UtcWindowSchema = z.object({
+    days: z.array(z.number().min(0).max(6)).required(),
+    startHour: z.number().min(0).max(23).required(),
+    endHour: z.number().min(1).max(24).required(),
+});
 const PriceSchema = z.object({
     route: z.string().required(),
     input: z.number().min(0).default(0),
     output: z.number().min(0).default(0),
     cacheRead: z.number().min(0).default(0),
     cacheWrite: z.number().min(0).default(0),
+    minPromptTokens: z.number().min(0),
+    maxPromptTokens: z.number().min(0),
+    // Schemastery arrays default to [], so wrap this optional field in a union
+    // to preserve undefined for ordinary flat/context-tiered prices.
+    utcWindows: z.union([z.array(UtcWindowSchema), z.const(undefined)]),
+    outsideUtcWindows: z.boolean(),
+    validFrom: z.string(),
+    validTo: z.string(),
 });
+function clonePrice(price) {
+    return {
+        ...price,
+        ...(price.utcWindows === undefined ? {} : {
+            utcWindows: price.utcWindows.map((window) => ({ ...window, days: [...window.days] })),
+        }),
+    };
+}
 export const Config = z.object({
-    pricing: z.array(PriceSchema).default(DEFAULT_PRICING.map((price) => ({ ...price }))),
+    pricing: z.array(PriceSchema).default(DEFAULT_PRICING.map(clonePrice)),
     scanConcurrency: z.number().min(1).max(16).default(4),
 });
 /** Read-only analytics over the canonical durable Harness session log. */
@@ -27,7 +50,9 @@ export class UsageService extends Service {
     refreshPromise;
     constructor(ctx, config) {
         super(ctx, 'usage');
-        this.pricing = (config.pricing ?? DEFAULT_PRICING).map((price) => ({ ...price }));
+        const pricing = (config.pricing ?? DEFAULT_PRICING).map(clonePrice);
+        validatePricing(pricing);
+        this.pricing = pricing;
         this.scanConcurrency = config.scanConcurrency ?? 4;
     }
     async *[Service.init]() {

package/dist/pricing-catalog.d.ts

@@ -0,0 +1,35 @@
+import type { ModelPrice, UtcPricingWindow } from './types.js';
+export declare const PRICING_CATALOG_VERIFIED_AT = "2026-08-26";
+export declare const PRICING_SOURCES: Readonly<{
+    openai: "https://developers.openai.com/api/docs/pricing";
+    anthropic: "https://platform.claude.com/docs/en/about-claude/pricing";
+    gemini: "https://ai.google.dev/gemini-api/docs/pricing";
+    deepseek: "https://api-docs.deepseek.com/quick_start/pricing/";
+    zai: "https://docs.z.ai/guides/overview/pricing";
+    kimi: "https://platform.kimi.ai/docs/pricing";
+    xai: "https://docs.x.ai/developers/pricing";
+    mistral: "https://mistral.ai/pricing/api/";
+    cohere: "https://cohere.com/pricing";
+    qwen: "https://www.alibabacloud.com/help/en/model-studio/model-pricing";
+    minimax: "https://platform.minimax.io/docs/guides/pricing-paygo";
+}>;
+export interface PricingCatalogEntry {
+    family: string;
+    providers: string[];
+    models: string[];
+    input: number;
+    output: number;
+    cacheRead: number;
+    cacheWrite: number;
+    source: keyof typeof PRICING_SOURCES;
+    note?: string;
+    minPromptTokens?: number;
+    maxPromptTokens?: number;
+    utcWindows?: UtcPricingWindow[];
+    outsideUtcWindows?: boolean;
+    validFrom?: string;
+    validTo?: string;
+}
+export declare const PRICING_CATALOG: readonly PricingCatalogEntry[];
+export declare function validatePricing(pricing: readonly ModelPrice[]): void;
+export declare const DEFAULT_PRICING: ModelPrice[];

package/dist/pricing-catalog.js

@@ -0,0 +1,277 @@
+export const PRICING_CATALOG_VERIFIED_AT = '2026-08-26';
+export const PRICING_SOURCES = Object.freeze({
+    openai: 'https://developers.openai.com/api/docs/pricing',
+    anthropic: 'https://platform.claude.com/docs/en/about-claude/pricing',
+    gemini: 'https://ai.google.dev/gemini-api/docs/pricing',
+    deepseek: 'https://api-docs.deepseek.com/quick_start/pricing/',
+    zai: 'https://docs.z.ai/guides/overview/pricing',
+    kimi: 'https://platform.kimi.ai/docs/pricing',
+    xai: 'https://docs.x.ai/developers/pricing',
+    mistral: 'https://mistral.ai/pricing/api/',
+    cohere: 'https://cohere.com/pricing',
+    qwen: 'https://www.alibabacloud.com/help/en/model-studio/model-pricing',
+    minimax: 'https://platform.minimax.io/docs/guides/pricing-paygo',
+});
+const WEEKDAY_DEEPSEEK_PEAK = [
+    { days: [1, 2, 3, 4, 5], startHour: 1, endHour: 4 },
+    { days: [1, 2, 3, 4, 5], startHour: 6, endHour: 10 },
+];
+const entries = [];
+function add(entry) {
+    entries.push(entry);
+}
+function addFlat(source, family, providers, models, input, cacheRead, output, options = {}) {
+    add({
+        source,
+        family,
+        providers,
+        models,
+        input,
+        cacheRead,
+        cacheWrite: options.cacheWrite ?? input,
+        output,
+        ...(options.note === undefined ? {} : { note: options.note }),
+        ...(options.validFrom === undefined ? {} : { validFrom: options.validFrom }),
+        ...(options.validTo === undefined ? {} : { validTo: options.validTo }),
+    });
+}
+function addContextTiered(source, family, providers, models, threshold, short, long, note, validity = {}) {
+    add({
+        source,
+        family: `${family} · long context`,
+        providers,
+        models,
+        input: long.input,
+        cacheRead: long.cacheRead,
+        cacheWrite: long.cacheWrite ?? long.input,
+        output: long.output,
+        minPromptTokens: threshold,
+        ...(note === undefined ? {} : { note }),
+        ...(validity.validFrom === undefined ? {} : { validFrom: validity.validFrom }),
+        ...(validity.validTo === undefined ? {} : { validTo: validity.validTo }),
+    });
+    add({
+        source,
+        family,
+        providers,
+        models,
+        input: short.input,
+        cacheRead: short.cacheRead,
+        cacheWrite: short.cacheWrite ?? short.input,
+        output: short.output,
+        maxPromptTokens: threshold - 1,
+        ...(note === undefined ? {} : { note }),
+        ...(validity.validFrom === undefined ? {} : { validFrom: validity.validFrom }),
+        ...(validity.validTo === undefined ? {} : { validTo: validity.validTo }),
+    });
+}
+const OPENAI = ['openai', 'openai-codex'];
+addContextTiered('openai', 'GPT-5.6 Sol', OPENAI, ['gpt-5.6-sol'], 272_000, { input: 4, cacheRead: 0.4, cacheWrite: 5, output: 20 }, { input: 8, cacheRead: 0.8, cacheWrite: 10, output: 30 }, 'Standard synchronous tier; promotional through at least 2026-11-21.', { validFrom: '2026-08-26T00:00:00.000Z', validTo: '2026-11-22T00:00:00.000Z' });
+addContextTiered('openai', 'GPT-5.6 Terra', OPENAI, ['gpt-5.6-terra'], 272_000, { input: 2, cacheRead: 0.2, cacheWrite: 2.5, output: 12 }, { input: 4, cacheRead: 0.4, cacheWrite: 5, output: 18 });
+addContextTiered('openai', 'GPT-5.6 Luna', OPENAI, ['gpt-5.6-luna'], 272_000, { input: 0.2, cacheRead: 0.02, cacheWrite: 0.25, output: 1.2 }, { input: 0.4, cacheRead: 0.04, cacheWrite: 0.5, output: 1.8 });
+addContextTiered('openai', 'GPT-5.5', OPENAI, ['gpt-5.5'], 272_000, { input: 5, cacheRead: 0.5, output: 30 }, { input: 10, cacheRead: 1, output: 45 });
+addContextTiered('openai', 'GPT-5.5 Pro', OPENAI, ['gpt-5.5-pro'], 272_000, { input: 30, cacheRead: 30, output: 180 }, { input: 60, cacheRead: 60, output: 270 }, 'No discounted cached-input SKU is published; cached buckets use input price.');
+addContextTiered('openai', 'GPT-5.4', OPENAI, ['gpt-5.4'], 272_000, { input: 2.5, cacheRead: 0.25, output: 15 }, { input: 5, cacheRead: 0.5, output: 22.5 });
+addContextTiered('openai', 'GPT-5.4 Pro', OPENAI, ['gpt-5.4-pro'], 272_000, { input: 30, cacheRead: 30, output: 180 }, { input: 60, cacheRead: 60, output: 270 }, 'No discounted cached-input SKU is published; cached buckets use input price.');
+addFlat('openai', 'GPT-5.4 Mini', OPENAI, ['gpt-5.4-mini'], 0.75, 0.075, 4.5);
+addFlat('openai', 'GPT-5.4 Nano', OPENAI, ['gpt-5.4-nano'], 0.2, 0.02, 1.25);
+addFlat('openai', 'GPT-4.1', OPENAI, ['gpt-4.1', 'gpt-4.1-20*'], 2, 0.5, 8);
+addFlat('openai', 'GPT-4.1 Mini', OPENAI, ['gpt-4.1-mini', 'gpt-4.1-mini-20*'], 0.4, 0.1, 1.6);
+addFlat('openai', 'GPT-4.1 Nano', OPENAI, ['gpt-4.1-nano', 'gpt-4.1-nano-20*'], 0.1, 0.025, 0.4);
+addFlat('openai', 'GPT-5.3 Codex', OPENAI, ['gpt-5.3-codex'], 1.75, 0.175, 14, { note: 'Standard Codex tier.' });
+addFlat('openai', 'GPT-5', OPENAI, ['gpt-5'], 1.25, 0.125, 10, { note: 'Prior-generation exact ID retained for existing sessions.' });
+addFlat('openai', 'GPT-5 Mini', OPENAI, ['gpt-5-mini'], 0.25, 0.025, 2, { note: 'Prior-generation exact ID retained for existing sessions.' });
+addFlat('openai', 'GPT-5 Nano', OPENAI, ['gpt-5-nano'], 0.05, 0.005, 0.4, { note: 'Prior-generation exact ID retained for existing sessions.' });
+addFlat('openai', 'GPT-4o', OPENAI, ['gpt-4o', 'gpt-4o-20*'], 2.5, 1.25, 10);
+addFlat('openai', 'GPT-4o Mini', OPENAI, ['gpt-4o-mini', 'gpt-4o-mini-20*'], 0.15, 0.075, 0.6);
+const ANTHROPIC = ['anthropic'];
+addFlat('anthropic', 'Claude Fable 5', ANTHROPIC, ['claude-fable-5*'], 10, 1, 50, { cacheWrite: 12.5, note: 'Cache-write estimate uses the standard 5-minute cache rate.' });
+addFlat('anthropic', 'Claude Mythos 5', ANTHROPIC, ['claude-mythos-5*'], 10, 1, 50, { cacheWrite: 12.5, note: 'Limited availability; cache-write estimate uses the 5-minute rate.' });
+for (const [family, model] of [
+    ['Claude Opus 5', 'claude-opus-5*'],
+    ['Claude Opus 4.8', 'claude-opus-4-8*'],
+    ['Claude Opus 4.7', 'claude-opus-4-7*'],
+    ['Claude Opus 4.6', 'claude-opus-4-6*'],
+    ['Claude Opus 4.5', 'claude-opus-4-5*'],
+])
+    addFlat('anthropic', family, ANTHROPIC, [model], 5, 0.5, 25, { cacheWrite: 6.25, note: 'Global standard inference; cache-write estimate uses the 5-minute rate.' });
+for (const [family, model] of [
+    ['Claude Opus 4.1', 'claude-opus-4-1*'],
+    ['Claude Opus 4', 'claude-opus-4-2025*'],
+])
+    addFlat('anthropic', family, ANTHROPIC, [model], 15, 1.5, 75, { cacheWrite: 18.75, note: 'Retired on the first-party API; retained for historical logs.' });
+addFlat('anthropic', 'Claude Sonnet 5', ANTHROPIC, ['claude-sonnet-5*'], 2, 0.2, 10, { cacheWrite: 2.5, note: 'Cache-write estimate uses the 5-minute rate.' });
+for (const [family, model] of [
+    ['Claude Sonnet 4.6', 'claude-sonnet-4-6*'],
+    ['Claude Sonnet 4.5', 'claude-sonnet-4-5*'],
+    ['Claude Sonnet 4', 'claude-sonnet-4-2025*'],
+])
+    addFlat('anthropic', family, ANTHROPIC, [model], 3, 0.3, 15, { cacheWrite: 3.75, note: 'Cache-write estimate uses the 5-minute rate.' });
+addFlat('anthropic', 'Claude Haiku 4.5', ANTHROPIC, ['claude-haiku-4-5*'], 1, 0.1, 5, { cacheWrite: 1.25, note: 'Cache-write estimate uses the 5-minute rate.' });
+addFlat('anthropic', 'Claude Haiku 3.5', ANTHROPIC, ['claude-3-5-haiku*', 'claude-haiku-3-5*'], 0.8, 0.08, 4, { cacheWrite: 1, note: 'Retired on the first-party API; retained for historical logs.' });
+const GEMINI = ['google', 'gemini', 'google-ai'];
+addFlat('gemini', 'Gemini 3.7 Flash', GEMINI, ['gemini-3.7-flash'], 0.75, 0.075, 3.75, { note: 'Promotional through 2026-12-31; cache storage token-hours are excluded.', validFrom: '2026-08-26T00:00:00.000Z', validTo: '2027-01-01T00:00:00.000Z' });
+addFlat('gemini', 'Gemini 3.7 Flash · 2027 rate', GEMINI, ['gemini-3.7-flash'], 1.5, 0.15, 7.5, { note: 'Official rate beginning 2027-01-01; cache storage token-hours are excluded.', validFrom: '2027-01-01T00:00:00.000Z' });
+addFlat('gemini', 'Gemini 3.6 Flash', GEMINI, ['gemini-3.6-flash'], 0.75, 0.075, 3.75, { note: 'Promotional through 2026-12-31; cache storage token-hours are excluded.', validFrom: '2026-08-26T00:00:00.000Z', validTo: '2027-01-01T00:00:00.000Z' });
+addFlat('gemini', 'Gemini 3.6 Flash · 2027 rate', GEMINI, ['gemini-3.6-flash'], 1.5, 0.15, 7.5, { note: 'Official rate beginning 2027-01-01; cache storage token-hours are excluded.', validFrom: '2027-01-01T00:00:00.000Z' });
+addFlat('gemini', 'Gemini 3.5 Flash', GEMINI, ['gemini-3.5-flash'], 1.5, 0.15, 9, { note: 'Cache storage token-hours are excluded.' });
+addFlat('gemini', 'Gemini 3.5 Flash-Lite', GEMINI, ['gemini-3.5-flash-lite'], 0.3, 0.03, 2.5, { note: 'Cache storage token-hours are excluded.' });
+addContextTiered('gemini', 'Gemini 3.1 Pro Preview', GEMINI, ['gemini-3.1-pro-preview', 'gemini-3.1-pro-preview-customtools'], 200_001, { input: 2, cacheRead: 0.2, output: 12 }, { input: 4, cacheRead: 0.4, output: 18 }, 'Cache storage token-hours are excluded.');
+addFlat('gemini', 'Gemini 3.1 Flash-Lite', GEMINI, ['gemini-3.1-flash-lite'], 0.25, 0.025, 1.5, { note: 'Text/image/video rate; audio and cache storage are excluded.' });
+addContextTiered('gemini', 'Gemini 2.5 Pro', GEMINI, ['gemini-2.5-pro'], 200_001, { input: 1.25, cacheRead: 0.125, output: 10 }, { input: 2.5, cacheRead: 0.25, output: 15 }, 'Cache storage token-hours are excluded.');
+addFlat('gemini', 'Gemini 2.5 Flash', GEMINI, ['gemini-2.5-flash'], 0.3, 0.03, 2.5, { note: 'Text/image/video rate; audio and cache storage are excluded.' });
+addFlat('gemini', 'Gemini 2.5 Flash-Lite', GEMINI, ['gemini-2.5-flash-lite'], 0.1, 0.01, 0.4, { note: 'Text/image/video rate; audio and cache storage are excluded.' });
+const DEEPSEEK = ['deepseek', 'deepseek-api'];
+function addDeepSeek(family, model, offPeak, peak) {
+    add({ source: 'deepseek', family: `${family} · peak`, providers: DEEPSEEK, models: [model], input: peak[0], cacheRead: peak[1], cacheWrite: peak[0], output: peak[2], utcWindows: WEEKDAY_DEEPSEEK_PEAK });
+    add({ source: 'deepseek', family: `${family} · off-peak`, providers: DEEPSEEK, models: [model], input: offPeak[0], cacheRead: offPeak[1], cacheWrite: offPeak[0], output: offPeak[2], utcWindows: WEEKDAY_DEEPSEEK_PEAK, outsideUtcWindows: true });
+}
+addDeepSeek('DeepSeek V4 Flash', 'deepseek-v4-flash', [0.22, 0.007, 0.66], [0.44, 0.014, 1.32]);
+addDeepSeek('DeepSeek V4 Pro', 'deepseek-v4-pro', [0.66, 0.022, 1.98], [1.32, 0.044, 3.96]);
+addFlat('deepseek', 'DeepSeek Chat legacy', ['deepseek'], ['deepseek-chat'], 0.28, 0.028, 0.42, { note: 'Retired after 2026-07-24 15:59 UTC; retained for historical logs.', validTo: '2026-07-24T16:00:00.000Z' });
+addFlat('deepseek', 'DeepSeek Reasoner legacy', ['deepseek'], ['deepseek-reasoner'], 0.55, 0.14, 2.19, { note: 'Retired after 2026-07-24 15:59 UTC; retained for historical logs.', validTo: '2026-07-24T16:00:00.000Z' });
+const ZAI = ['zai', 'z-ai', 'zhipu', 'bigmodel'];
+for (const [family, model, input, cacheRead, output] of [
+    ['GLM-5.3', 'glm-5.3', 1.4, 0.26, 4.4],
+    ['GLM-5.2', 'glm-5.2', 1.4, 0.26, 4.4],
+    ['GLM-5.1', 'glm-5.1', 1.4, 0.26, 4.4],
+    ['GLM-5', 'glm-5', 1, 0.2, 3.2],
+    ['GLM-5 Turbo', 'glm-5-turbo', 1.2, 0.24, 4],
+    ['GLM-4.7', 'glm-4.7', 0.6, 0.11, 2.2],
+    ['GLM-4.7 FlashX', 'glm-4.7-flashx', 0.07, 0.01, 0.4],
+    ['GLM-4.6', 'glm-4.6', 0.6, 0.11, 2.2],
+    ['GLM-4.5', 'glm-4.5', 0.6, 0.11, 2.2],
+    ['GLM-4.5 X', 'glm-4.5-x', 2.2, 0.45, 8.9],
+    ['GLM-4.5 Air', 'glm-4.5-air', 0.2, 0.03, 1.1],
+    ['GLM-4.5 AirX', 'glm-4.5-airx', 1.1, 0.22, 4.5],
+    ['GLM-4 32B', 'glm-4-32b-0414-128k', 0.1, 0.1, 0.1],
+    ['GLM-4.7 Flash', 'glm-4.7-flash', 0, 0, 0],
+    ['GLM-4.5 Flash', 'glm-4.5-flash', 0, 0, 0],
+])
+    addFlat('zai', family, ZAI, [model], input, cacheRead, output, { note: 'Global Z.AI endpoint; cached-input storage is currently free.' });
+const KIMI = ['kimi', 'moonshot'];
+for (const [family, model, input, cacheRead, output, note, validTo] of [
+    ['Kimi K3', 'kimi-k3', 3, 0.3, 15, 'Standard realtime tier.', undefined],
+    ['Kimi K2.7 Code', 'kimi-k2.7-code', 0.95, 0.19, 4, 'Standard realtime tier.', undefined],
+    ['Kimi K2.7 Code Highspeed', 'kimi-k2.7-code-highspeed', 1.9, 0.38, 8, 'High-speed serving tier.', undefined],
+    ['Kimi K2.6', 'kimi-k2.6', 0.95, 0.16, 4, 'Standard realtime tier.', undefined],
+    ['Kimi K2.5', 'kimi-k2.5', 0.6, 0.1, 3, 'Scheduled for retirement on 2026-08-31; retained for historical logs.', '2026-09-01T00:00:00.000Z'],
+    ['Moonshot V1 8K', 'moonshot-v1-8k', 0.2, 0.2, 2, 'Scheduled for retirement on 2026-08-31; no cache discount is published.', '2026-09-01T00:00:00.000Z'],
+    ['Moonshot V1 32K', 'moonshot-v1-32k', 1, 1, 3, 'Scheduled for retirement on 2026-08-31; no cache discount is published.', '2026-09-01T00:00:00.000Z'],
+    ['Moonshot V1 128K', 'moonshot-v1-128k', 2, 2, 5, 'Scheduled for retirement on 2026-08-31; no cache discount is published.', '2026-09-01T00:00:00.000Z'],
+])
+    addFlat('kimi', family, KIMI, [model], input, cacheRead, output, { note, ...(validTo === undefined ? {} : { validTo }) });
+const XAI = ['xai'];
+for (const [family, model, shortInput, shortCache, shortOutput, longInput, longCache, longOutput] of [
+    ['Grok 4.6', 'grok-4.6', 2, 0.5, 6, 4, 1, 12],
+    ['Grok Build 0.1', 'grok-build-0.1', 1, 0.2, 2, 2, 0.4, 4],
+    ['Grok 4.5', 'grok-4.5', 2, 0.3, 6, 4, 0.6, 12],
+    ['Grok 4.3', 'grok-4.3', 1.25, 0.2, 2.5, 2.5, 0.4, 5],
+    ['Grok 4.20 Reasoning', 'grok-4.20-0309-reasoning', 1.25, 0.2, 2.5, 2.5, 0.4, 5],
+    ['Grok 4.20 Non-Reasoning', 'grok-4.20-0309-non-reasoning', 1.25, 0.2, 2.5, 2.5, 0.4, 5],
+    ['Grok 4.20 Multi-Agent', 'grok-4.20-multi-agent-0309', 1.25, 0.2, 2.5, 2.5, 0.4, 5],
+])
+    addContextTiered('xai', family, XAI, [model], 200_000, { input: shortInput, cacheRead: shortCache, output: shortOutput }, { input: longInput, cacheRead: longCache, output: longOutput }, 'Standard text API; tool-call fees are excluded.');
+const MISTRAL = ['mistral'];
+for (const [family, models, input, output] of [
+    ['Mistral Medium 3.5', ['mistral-medium-3-5'], 1.5, 7.5],
+    ['Mistral Large 3', ['mistral-large-2512'], 0.5, 1.5],
+    ['Mistral Small 4', ['mistral-small-2603'], 0.15, 0.6],
+    ['Codestral', ['codestral-2508'], 0.3, 0.9],
+])
+    addFlat('mistral', family, MISTRAL, [...models], input, input, output, { note: 'Exact cache discount is not published per model; cached buckets conservatively use input price.' });
+const COHERE = ['cohere'];
+for (const [family, model, input, output, note] of [
+    ['Command A', 'command-a-03-2025', 2.5, 10, 'Current paid production model.'],
+    ['Command R7B', 'command-r7b-12-2024', 0.0375, 0.15, 'Pinned paid model.'],
+    ['Command R', 'command-r-08-2024', 0.15, 0.6, 'Pinned paid model.'],
+    ['Command R+', 'command-r-plus-08-2024', 2.5, 10, 'Pinned paid model.'],
+    ['Command legacy', 'command', 1, 2, 'Deprecated; retained for historical logs.'],
+    ['Command Light legacy', 'command-light', 0.3, 0.6, 'Deprecated; retained for historical logs.'],
+    ['Command R legacy', 'command-r-03-2024', 0.5, 1.5, 'Deprecated; retained for historical logs.'],
+    ['Command R+ legacy', 'command-r-plus-04-2024', 3, 15, 'Deprecated; retained for historical logs.'],
+])
+    addFlat('cohere', family, COHERE, [model], input, input, output, { note: `${note} No cache discount is published.` });
+const QWEN = ['dashscope', 'alibaba', 'qwen'];
+add({ source: 'qwen', family: 'Qwen 3.7 Max', providers: QWEN, models: ['qwen3.7-max-2026-06-08'], input: 2.5, cacheRead: 0.5, cacheWrite: 3.125, output: 7.5, maxPromptTokens: 1_000_000, note: 'Singapore international list price; cache read uses the implicit-cache rate.' });
+addContextTiered('qwen', 'Qwen 3.7 Plus', QWEN, ['qwen3.7-plus-2026-05-26'], 256_001, { input: 0.4, cacheRead: 0.08, cacheWrite: 0.5, output: 1.6 }, { input: 1.2, cacheRead: 0.24, cacheWrite: 1.5, output: 4.8 }, 'Singapore international list price; cache read uses the implicit-cache rate.');
+function addQwenTier(family, model, min, max, input, output) {
+    const cacheRead = Number((input * 0.2).toFixed(6));
+    const cacheWrite = Number((input * 1.25).toFixed(6));
+    add({ source: 'qwen', family, providers: QWEN, models: [model], input, cacheRead, cacheWrite, output, minPromptTokens: min, maxPromptTokens: max, note: 'Singapore international list price; cache read uses the implicit-cache rate.' });
+}
+addQwenTier('Qwen 3 Max · ≤32K', 'qwen3-max-2026-01-23', 0, 32_000, 1.2, 6);
+addQwenTier('Qwen 3 Max · 32K–128K', 'qwen3-max-2026-01-23', 32_001, 128_000, 2.4, 12);
+addQwenTier('Qwen 3 Max · 128K–256K', 'qwen3-max-2026-01-23', 128_001, 256_000, 3, 15);
+addQwenTier('Qwen 3 Coder Plus · ≤32K', 'qwen3-coder-plus-2025-09-23', 0, 32_000, 1, 5);
+addQwenTier('Qwen 3 Coder Plus · 32K–128K', 'qwen3-coder-plus-2025-09-23', 32_001, 128_000, 1.8, 9);
+addQwenTier('Qwen 3 Coder Plus · 128K–256K', 'qwen3-coder-plus-2025-09-23', 128_001, 256_000, 3, 15);
+addQwenTier('Qwen 3 Coder Plus · 256K–1M', 'qwen3-coder-plus-2025-09-23', 256_001, 1_000_000, 6, 60);
+addQwenTier('Qwen 3 Coder Flash · ≤32K', 'qwen3-coder-flash-2025-07-28', 0, 32_000, 0.3, 1.5);
+addQwenTier('Qwen 3 Coder Flash · 32K–128K', 'qwen3-coder-flash-2025-07-28', 32_001, 128_000, 0.5, 2.5);
+addQwenTier('Qwen 3 Coder Flash · 128K–256K', 'qwen3-coder-flash-2025-07-28', 128_001, 256_000, 0.8, 4);
+addQwenTier('Qwen 3 Coder Flash · 256K–1M', 'qwen3-coder-flash-2025-07-28', 256_001, 1_000_000, 1.6, 9.6);
+const MINIMAX = ['minimax'];
+addContextTiered('minimax', 'MiniMax M3', MINIMAX, ['MiniMax-M3'], 512_001, { input: 0.3, cacheRead: 0.06, output: 1.2 }, { input: 0.6, cacheRead: 0.12, output: 2.4 }, 'Standard tier; no separate cache-write price is published.');
+for (const [family, model, input, cacheRead, cacheWrite, output] of [
+    ['MiniMax M2.7', 'MiniMax-M2.7', 0.3, 0.06, 0.375, 1.2],
+    ['MiniMax M2.7 Highspeed', 'MiniMax-M2.7-highspeed', 0.6, 0.06, 0.375, 2.4],
+    ['MiniMax M2.5', 'MiniMax-M2.5', 0.3, 0.03, 0.375, 1.2],
+    ['MiniMax M2.5 Highspeed', 'MiniMax-M2.5-highspeed', 0.6, 0.03, 0.375, 2.4],
+])
+    addFlat('minimax', family, MINIMAX, [model], input, cacheRead, output, { cacheWrite });
+export const PRICING_CATALOG = Object.freeze(entries.map((entry) => Object.freeze({ ...entry })));
+export function validatePricing(pricing) {
+    for (const [index, price] of pricing.entries()) {
+        const label = `pricing[${index}] (${price.route})`;
+        if (price.route === '')
+            throw new Error(`${label}: route must not be empty`);
+        for (const [bucket, value] of Object.entries({ input: price.input, output: price.output, cacheRead: price.cacheRead, cacheWrite: price.cacheWrite })) {
+            if (!Number.isFinite(value) || value < 0)
+                throw new Error(`${label}: ${bucket} must be a non-negative finite number`);
+        }
+        for (const [field, value] of [['minPromptTokens', price.minPromptTokens], ['maxPromptTokens', price.maxPromptTokens]]) {
+            if (value !== undefined && (!Number.isSafeInteger(value) || value < 0))
+                throw new Error(`${label}: ${field} must be a non-negative safe integer`);
+        }
+        if (price.minPromptTokens !== undefined && price.maxPromptTokens !== undefined && price.minPromptTokens > price.maxPromptTokens) {
+            throw new Error(`${label}: minPromptTokens must not exceed maxPromptTokens`);
+        }
+        if (price.utcWindows !== undefined && price.utcWindows.length === 0) {
+            throw new Error(`${label}: utcWindows requires at least one utcWindow`);
+        }
+        if (price.outsideUtcWindows === true && price.utcWindows === undefined) {
+            throw new Error(`${label}: outsideUtcWindows requires at least one utcWindow`);
+        }
+        for (const window of price.utcWindows ?? []) {
+            if (window.days.length === 0 || new Set(window.days).size !== window.days.length || window.days.some((day) => !Number.isInteger(day) || day < 0 || day > 6)) {
+                throw new Error(`${label}: utcWindow days must be unique integers from 0 through 6`);
+            }
+            if (!Number.isInteger(window.startHour) || !Number.isInteger(window.endHour) || window.startHour < 0 || window.endHour > 24 || window.startHour >= window.endHour) {
+                throw new Error(`${label}: utcWindow requires integer hours with 0 <= startHour < endHour <= 24`);
+            }
+        }
+        const validFrom = price.validFrom === undefined ? undefined : Date.parse(price.validFrom);
+        const validTo = price.validTo === undefined ? undefined : Date.parse(price.validTo);
+        if (validFrom !== undefined && !Number.isFinite(validFrom))
+            throw new Error(`${label}: validFrom must be ISO-8601`);
+        if (validTo !== undefined && !Number.isFinite(validTo))
+            throw new Error(`${label}: validTo must be ISO-8601`);
+        if (validFrom !== undefined && validTo !== undefined && validFrom >= validTo)
+            throw new Error(`${label}: validFrom must precede validTo`);
+    }
+}
+export const DEFAULT_PRICING = PRICING_CATALOG.flatMap((entry) => entry.providers.flatMap((provider) => entry.models.map((model) => ({
+    route: `${provider}/${model}`,
+    input: entry.input,
+    output: entry.output,
+    cacheRead: entry.cacheRead,
+    cacheWrite: entry.cacheWrite,
+    ...(entry.minPromptTokens === undefined ? {} : { minPromptTokens: entry.minPromptTokens }),
+    ...(entry.maxPromptTokens === undefined ? {} : { maxPromptTokens: entry.maxPromptTokens }),
+    ...(entry.utcWindows === undefined ? {} : { utcWindows: entry.utcWindows.map((window) => ({ ...window, days: [...window.days] })) }),
+    ...(entry.outsideUtcWindows === undefined ? {} : { outsideUtcWindows: entry.outsideUtcWindows }),
+    ...(entry.validFrom === undefined ? {} : { validFrom: entry.validFrom }),
+    ...(entry.validTo === undefined ? {} : { validTo: entry.validTo }),
+}))));
+validatePricing(DEFAULT_PRICING);

package/dist/types.d.ts

@@ -1,4 +1,12 @@
 export type UsageRange = '30d' | '90d' | '365d' | 'all';
+export interface UtcPricingWindow {
+    /** UTC weekday numbers, Sunday = 0. */
+    days: number[];
+    /** Inclusive UTC hour, 0–23. */
+    startHour: number;
+    /** Exclusive UTC hour, 1–24. */
+    endHour: number;
+}
 export interface ModelPrice {
     /** Provider/model glob. Only `*` is special. */
     route: string;
@@ -10,6 +18,18 @@ export interface ModelPrice {
     cacheRead: number;
     /** USD per one million cache-write tokens. */
     cacheWrite: number;
+    /** Match calls whose total prompt buckets are at least this size. */
+    minPromptTokens?: number;
+    /** Match calls whose total prompt buckets are no larger than this size. */
+    maxPromptTokens?: number;
+    /** Optional recurring UTC billing windows. */
+    utcWindows?: UtcPricingWindow[];
+    /** Match outside utcWindows instead of inside them. */
+    outsideUtcWindows?: boolean;
+    /** Inclusive ISO-8601 instant when this price becomes valid. */
+    validFrom?: string;
+    /** Exclusive ISO-8601 instant when this price stops being valid. */
+    validTo?: string;
 }
 export interface UsagePluginConfig {
     pricing?: ModelPrice[];

package/docs/pricing-catalog.md

@@ -0,0 +1,164 @@
+# Built-in pricing catalog
+
+Verified against official public pricing pages on **2026-08-26 UTC**. Prices are USD per one million tokens.
+
+The catalog estimates standard, synchronous, first-party API usage. Batch/flex/priority modes, regional uplifts, negotiated discounts, subscriptions, tool-call fees, taxes, and cache-storage token-hours are excluded unless a row note explicitly says otherwise.
+
+Catalog entries: **128**. Generated route rules include provider aliases and context/time tiers.
+
+| Family / tier | Provider routes | Models | Input | Cache read | Cache write | Output | Match | Notes |
+|---|---|---|---:|---:|---:|---:|---|---|
+| GPT-5.6 Sol · long context | openai, openai-codex | gpt-5.6-sol | $8 | $0.8 | $10 | $30 | prompt ≥ 272,000; from 2026-08-26T00:00:00.000Z; before 2026-11-22T00:00:00.000Z | Standard synchronous tier; promotional through at least 2026-11-21. |
+| GPT-5.6 Sol | openai, openai-codex | gpt-5.6-sol | $4 | $0.4 | $5 | $20 | prompt ≤ 271,999; from 2026-08-26T00:00:00.000Z; before 2026-11-22T00:00:00.000Z | Standard synchronous tier; promotional through at least 2026-11-21. |
+| GPT-5.6 Terra · long context | openai, openai-codex | gpt-5.6-terra | $4 | $0.4 | $5 | $18 | prompt ≥ 272,000 |  |
+| GPT-5.6 Terra | openai, openai-codex | gpt-5.6-terra | $2 | $0.2 | $2.5 | $12 | prompt ≤ 271,999 |  |
+| GPT-5.6 Luna · long context | openai, openai-codex | gpt-5.6-luna | $0.4 | $0.04 | $0.5 | $1.8 | prompt ≥ 272,000 |  |
+| GPT-5.6 Luna | openai, openai-codex | gpt-5.6-luna | $0.2 | $0.02 | $0.25 | $1.2 | prompt ≤ 271,999 |  |
+| GPT-5.5 · long context | openai, openai-codex | gpt-5.5 | $10 | $1 | $10 | $45 | prompt ≥ 272,000 |  |
+| GPT-5.5 | openai, openai-codex | gpt-5.5 | $5 | $0.5 | $5 | $30 | prompt ≤ 271,999 |  |
+| GPT-5.5 Pro · long context | openai, openai-codex | gpt-5.5-pro | $60 | $60 | $60 | $270 | prompt ≥ 272,000 | No discounted cached-input SKU is published; cached buckets use input price. |
+| GPT-5.5 Pro | openai, openai-codex | gpt-5.5-pro | $30 | $30 | $30 | $180 | prompt ≤ 271,999 | No discounted cached-input SKU is published; cached buckets use input price. |
+| GPT-5.4 · long context | openai, openai-codex | gpt-5.4 | $5 | $0.5 | $5 | $22.5 | prompt ≥ 272,000 |  |
+| GPT-5.4 | openai, openai-codex | gpt-5.4 | $2.5 | $0.25 | $2.5 | $15 | prompt ≤ 271,999 |  |
+| GPT-5.4 Pro · long context | openai, openai-codex | gpt-5.4-pro | $60 | $60 | $60 | $270 | prompt ≥ 272,000 | No discounted cached-input SKU is published; cached buckets use input price. |
+| GPT-5.4 Pro | openai, openai-codex | gpt-5.4-pro | $30 | $30 | $30 | $180 | prompt ≤ 271,999 | No discounted cached-input SKU is published; cached buckets use input price. |
+| GPT-5.4 Mini | openai, openai-codex | gpt-5.4-mini | $0.75 | $0.075 | $0.75 | $4.5 | standard |  |
+| GPT-5.4 Nano | openai, openai-codex | gpt-5.4-nano | $0.2 | $0.02 | $0.2 | $1.25 | standard |  |
+| GPT-4.1 | openai, openai-codex | gpt-4.1, gpt-4.1-20* | $2 | $0.5 | $2 | $8 | standard |  |
+| GPT-4.1 Mini | openai, openai-codex | gpt-4.1-mini, gpt-4.1-mini-20* | $0.4 | $0.1 | $0.4 | $1.6 | standard |  |
+| GPT-4.1 Nano | openai, openai-codex | gpt-4.1-nano, gpt-4.1-nano-20* | $0.1 | $0.025 | $0.1 | $0.4 | standard |  |
+| GPT-5.3 Codex | openai, openai-codex | gpt-5.3-codex | $1.75 | $0.175 | $1.75 | $14 | standard | Standard Codex tier. |
+| GPT-5 | openai, openai-codex | gpt-5 | $1.25 | $0.125 | $1.25 | $10 | standard | Prior-generation exact ID retained for existing sessions. |
+| GPT-5 Mini | openai, openai-codex | gpt-5-mini | $0.25 | $0.025 | $0.25 | $2 | standard | Prior-generation exact ID retained for existing sessions. |
+| GPT-5 Nano | openai, openai-codex | gpt-5-nano | $0.05 | $0.005 | $0.05 | $0.4 | standard | Prior-generation exact ID retained for existing sessions. |
+| GPT-4o | openai, openai-codex | gpt-4o, gpt-4o-20* | $2.5 | $1.25 | $2.5 | $10 | standard |  |
+| GPT-4o Mini | openai, openai-codex | gpt-4o-mini, gpt-4o-mini-20* | $0.15 | $0.075 | $0.15 | $0.6 | standard |  |
+| Claude Fable 5 | anthropic | claude-fable-5* | $10 | $1 | $12.5 | $50 | standard | Cache-write estimate uses the standard 5-minute cache rate. |
+| Claude Mythos 5 | anthropic | claude-mythos-5* | $10 | $1 | $12.5 | $50 | standard | Limited availability; cache-write estimate uses the 5-minute rate. |
+| Claude Opus 5 | anthropic | claude-opus-5* | $5 | $0.5 | $6.25 | $25 | standard | Global standard inference; cache-write estimate uses the 5-minute rate. |
+| Claude Opus 4.8 | anthropic | claude-opus-4-8* | $5 | $0.5 | $6.25 | $25 | standard | Global standard inference; cache-write estimate uses the 5-minute rate. |
+| Claude Opus 4.7 | anthropic | claude-opus-4-7* | $5 | $0.5 | $6.25 | $25 | standard | Global standard inference; cache-write estimate uses the 5-minute rate. |
+| Claude Opus 4.6 | anthropic | claude-opus-4-6* | $5 | $0.5 | $6.25 | $25 | standard | Global standard inference; cache-write estimate uses the 5-minute rate. |
+| Claude Opus 4.5 | anthropic | claude-opus-4-5* | $5 | $0.5 | $6.25 | $25 | standard | Global standard inference; cache-write estimate uses the 5-minute rate. |
+| Claude Opus 4.1 | anthropic | claude-opus-4-1* | $15 | $1.5 | $18.75 | $75 | standard | Retired on the first-party API; retained for historical logs. |
+| Claude Opus 4 | anthropic | claude-opus-4-2025* | $15 | $1.5 | $18.75 | $75 | standard | Retired on the first-party API; retained for historical logs. |
+| Claude Sonnet 5 | anthropic | claude-sonnet-5* | $2 | $0.2 | $2.5 | $10 | standard | Cache-write estimate uses the 5-minute rate. |
+| Claude Sonnet 4.6 | anthropic | claude-sonnet-4-6* | $3 | $0.3 | $3.75 | $15 | standard | Cache-write estimate uses the 5-minute rate. |
+| Claude Sonnet 4.5 | anthropic | claude-sonnet-4-5* | $3 | $0.3 | $3.75 | $15 | standard | Cache-write estimate uses the 5-minute rate. |
+| Claude Sonnet 4 | anthropic | claude-sonnet-4-2025* | $3 | $0.3 | $3.75 | $15 | standard | Cache-write estimate uses the 5-minute rate. |
+| Claude Haiku 4.5 | anthropic | claude-haiku-4-5* | $1 | $0.1 | $1.25 | $5 | standard | Cache-write estimate uses the 5-minute rate. |
+| Claude Haiku 3.5 | anthropic | claude-3-5-haiku*, claude-haiku-3-5* | $0.8 | $0.08 | $1 | $4 | standard | Retired on the first-party API; retained for historical logs. |
+| Gemini 3.7 Flash | google, gemini, google-ai | gemini-3.7-flash | $0.75 | $0.075 | $0.75 | $3.75 | from 2026-08-26T00:00:00.000Z; before 2027-01-01T00:00:00.000Z | Promotional through 2026-12-31; cache storage token-hours are excluded. |
+| Gemini 3.7 Flash · 2027 rate | google, gemini, google-ai | gemini-3.7-flash | $1.5 | $0.15 | $1.5 | $7.5 | from 2027-01-01T00:00:00.000Z | Official rate beginning 2027-01-01; cache storage token-hours are excluded. |
+| Gemini 3.6 Flash | google, gemini, google-ai | gemini-3.6-flash | $0.75 | $0.075 | $0.75 | $3.75 | from 2026-08-26T00:00:00.000Z; before 2027-01-01T00:00:00.000Z | Promotional through 2026-12-31; cache storage token-hours are excluded. |
+| Gemini 3.6 Flash · 2027 rate | google, gemini, google-ai | gemini-3.6-flash | $1.5 | $0.15 | $1.5 | $7.5 | from 2027-01-01T00:00:00.000Z | Official rate beginning 2027-01-01; cache storage token-hours are excluded. |
+| Gemini 3.5 Flash | google, gemini, google-ai | gemini-3.5-flash | $1.5 | $0.15 | $1.5 | $9 | standard | Cache storage token-hours are excluded. |
+| Gemini 3.5 Flash-Lite | google, gemini, google-ai | gemini-3.5-flash-lite | $0.3 | $0.03 | $0.3 | $2.5 | standard | Cache storage token-hours are excluded. |
+| Gemini 3.1 Pro Preview · long context | google, gemini, google-ai | gemini-3.1-pro-preview, gemini-3.1-pro-preview-customtools | $4 | $0.4 | $4 | $18 | prompt ≥ 200,001 | Cache storage token-hours are excluded. |
+| Gemini 3.1 Pro Preview | google, gemini, google-ai | gemini-3.1-pro-preview, gemini-3.1-pro-preview-customtools | $2 | $0.2 | $2 | $12 | prompt ≤ 200,000 | Cache storage token-hours are excluded. |
+| Gemini 3.1 Flash-Lite | google, gemini, google-ai | gemini-3.1-flash-lite | $0.25 | $0.025 | $0.25 | $1.5 | standard | Text/image/video rate; audio and cache storage are excluded. |
+| Gemini 2.5 Pro · long context | google, gemini, google-ai | gemini-2.5-pro | $2.5 | $0.25 | $2.5 | $15 | prompt ≥ 200,001 | Cache storage token-hours are excluded. |
+| Gemini 2.5 Pro | google, gemini, google-ai | gemini-2.5-pro | $1.25 | $0.125 | $1.25 | $10 | prompt ≤ 200,000 | Cache storage token-hours are excluded. |
+| Gemini 2.5 Flash | google, gemini, google-ai | gemini-2.5-flash | $0.3 | $0.03 | $0.3 | $2.5 | standard | Text/image/video rate; audio and cache storage are excluded. |
+| Gemini 2.5 Flash-Lite | google, gemini, google-ai | gemini-2.5-flash-lite | $0.1 | $0.01 | $0.1 | $0.4 | standard | Text/image/video rate; audio and cache storage are excluded. |
+| DeepSeek V4 Flash · peak | deepseek, deepseek-api | deepseek-v4-flash | $0.44 | $0.014 | $0.44 | $1.32 | inside listed UTC windows |  |
+| DeepSeek V4 Flash · off-peak | deepseek, deepseek-api | deepseek-v4-flash | $0.22 | $0.007 | $0.22 | $0.66 | outside listed UTC windows |  |
+| DeepSeek V4 Pro · peak | deepseek, deepseek-api | deepseek-v4-pro | $1.32 | $0.044 | $1.32 | $3.96 | inside listed UTC windows |  |
+| DeepSeek V4 Pro · off-peak | deepseek, deepseek-api | deepseek-v4-pro | $0.66 | $0.022 | $0.66 | $1.98 | outside listed UTC windows |  |
+| DeepSeek Chat legacy | deepseek | deepseek-chat | $0.28 | $0.028 | $0.28 | $0.42 | before 2026-07-24T16:00:00.000Z | Retired after 2026-07-24 15:59 UTC; retained for historical logs. |
+| DeepSeek Reasoner legacy | deepseek | deepseek-reasoner | $0.55 | $0.14 | $0.55 | $2.19 | before 2026-07-24T16:00:00.000Z | Retired after 2026-07-24 15:59 UTC; retained for historical logs. |
+| GLM-5.3 | zai, z-ai, zhipu, bigmodel | glm-5.3 | $1.4 | $0.26 | $1.4 | $4.4 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-5.2 | zai, z-ai, zhipu, bigmodel | glm-5.2 | $1.4 | $0.26 | $1.4 | $4.4 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-5.1 | zai, z-ai, zhipu, bigmodel | glm-5.1 | $1.4 | $0.26 | $1.4 | $4.4 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-5 | zai, z-ai, zhipu, bigmodel | glm-5 | $1 | $0.2 | $1 | $3.2 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-5 Turbo | zai, z-ai, zhipu, bigmodel | glm-5-turbo | $1.2 | $0.24 | $1.2 | $4 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-4.7 | zai, z-ai, zhipu, bigmodel | glm-4.7 | $0.6 | $0.11 | $0.6 | $2.2 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-4.7 FlashX | zai, z-ai, zhipu, bigmodel | glm-4.7-flashx | $0.07 | $0.01 | $0.07 | $0.4 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-4.6 | zai, z-ai, zhipu, bigmodel | glm-4.6 | $0.6 | $0.11 | $0.6 | $2.2 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-4.5 | zai, z-ai, zhipu, bigmodel | glm-4.5 | $0.6 | $0.11 | $0.6 | $2.2 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-4.5 X | zai, z-ai, zhipu, bigmodel | glm-4.5-x | $2.2 | $0.45 | $2.2 | $8.9 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-4.5 Air | zai, z-ai, zhipu, bigmodel | glm-4.5-air | $0.2 | $0.03 | $0.2 | $1.1 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-4.5 AirX | zai, z-ai, zhipu, bigmodel | glm-4.5-airx | $1.1 | $0.22 | $1.1 | $4.5 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-4 32B | zai, z-ai, zhipu, bigmodel | glm-4-32b-0414-128k | $0.1 | $0.1 | $0.1 | $0.1 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-4.7 Flash | zai, z-ai, zhipu, bigmodel | glm-4.7-flash | $0 | $0 | $0 | $0 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| GLM-4.5 Flash | zai, z-ai, zhipu, bigmodel | glm-4.5-flash | $0 | $0 | $0 | $0 | standard | Global Z.AI endpoint; cached-input storage is currently free. |
+| Kimi K3 | kimi, moonshot | kimi-k3 | $3 | $0.3 | $3 | $15 | standard | Standard realtime tier. |
+| Kimi K2.7 Code | kimi, moonshot | kimi-k2.7-code | $0.95 | $0.19 | $0.95 | $4 | standard | Standard realtime tier. |
+| Kimi K2.7 Code Highspeed | kimi, moonshot | kimi-k2.7-code-highspeed | $1.9 | $0.38 | $1.9 | $8 | standard | High-speed serving tier. |
+| Kimi K2.6 | kimi, moonshot | kimi-k2.6 | $0.95 | $0.16 | $0.95 | $4 | standard | Standard realtime tier. |
+| Kimi K2.5 | kimi, moonshot | kimi-k2.5 | $0.6 | $0.1 | $0.6 | $3 | before 2026-09-01T00:00:00.000Z | Scheduled for retirement on 2026-08-31; retained for historical logs. |
+| Moonshot V1 8K | kimi, moonshot | moonshot-v1-8k | $0.2 | $0.2 | $0.2 | $2 | before 2026-09-01T00:00:00.000Z | Scheduled for retirement on 2026-08-31; no cache discount is published. |
+| Moonshot V1 32K | kimi, moonshot | moonshot-v1-32k | $1 | $1 | $1 | $3 | before 2026-09-01T00:00:00.000Z | Scheduled for retirement on 2026-08-31; no cache discount is published. |
+| Moonshot V1 128K | kimi, moonshot | moonshot-v1-128k | $2 | $2 | $2 | $5 | before 2026-09-01T00:00:00.000Z | Scheduled for retirement on 2026-08-31; no cache discount is published. |
+| Grok 4.6 · long context | xai | grok-4.6 | $4 | $1 | $4 | $12 | prompt ≥ 200,000 | Standard text API; tool-call fees are excluded. |
+| Grok 4.6 | xai | grok-4.6 | $2 | $0.5 | $2 | $6 | prompt ≤ 199,999 | Standard text API; tool-call fees are excluded. |
+| Grok Build 0.1 · long context | xai | grok-build-0.1 | $2 | $0.4 | $2 | $4 | prompt ≥ 200,000 | Standard text API; tool-call fees are excluded. |
+| Grok Build 0.1 | xai | grok-build-0.1 | $1 | $0.2 | $1 | $2 | prompt ≤ 199,999 | Standard text API; tool-call fees are excluded. |
+| Grok 4.5 · long context | xai | grok-4.5 | $4 | $0.6 | $4 | $12 | prompt ≥ 200,000 | Standard text API; tool-call fees are excluded. |
+| Grok 4.5 | xai | grok-4.5 | $2 | $0.3 | $2 | $6 | prompt ≤ 199,999 | Standard text API; tool-call fees are excluded. |
+| Grok 4.3 · long context | xai | grok-4.3 | $2.5 | $0.4 | $2.5 | $5 | prompt ≥ 200,000 | Standard text API; tool-call fees are excluded. |
+| Grok 4.3 | xai | grok-4.3 | $1.25 | $0.2 | $1.25 | $2.5 | prompt ≤ 199,999 | Standard text API; tool-call fees are excluded. |
+| Grok 4.20 Reasoning · long context | xai | grok-4.20-0309-reasoning | $2.5 | $0.4 | $2.5 | $5 | prompt ≥ 200,000 | Standard text API; tool-call fees are excluded. |
+| Grok 4.20 Reasoning | xai | grok-4.20-0309-reasoning | $1.25 | $0.2 | $1.25 | $2.5 | prompt ≤ 199,999 | Standard text API; tool-call fees are excluded. |
+| Grok 4.20 Non-Reasoning · long context | xai | grok-4.20-0309-non-reasoning | $2.5 | $0.4 | $2.5 | $5 | prompt ≥ 200,000 | Standard text API; tool-call fees are excluded. |
+| Grok 4.20 Non-Reasoning | xai | grok-4.20-0309-non-reasoning | $1.25 | $0.2 | $1.25 | $2.5 | prompt ≤ 199,999 | Standard text API; tool-call fees are excluded. |
+| Grok 4.20 Multi-Agent · long context | xai | grok-4.20-multi-agent-0309 | $2.5 | $0.4 | $2.5 | $5 | prompt ≥ 200,000 | Standard text API; tool-call fees are excluded. |
+| Grok 4.20 Multi-Agent | xai | grok-4.20-multi-agent-0309 | $1.25 | $0.2 | $1.25 | $2.5 | prompt ≤ 199,999 | Standard text API; tool-call fees are excluded. |
+| Mistral Medium 3.5 | mistral | mistral-medium-3-5 | $1.5 | $1.5 | $1.5 | $7.5 | standard | Exact cache discount is not published per model; cached buckets conservatively use input price. |
+| Mistral Large 3 | mistral | mistral-large-2512 | $0.5 | $0.5 | $0.5 | $1.5 | standard | Exact cache discount is not published per model; cached buckets conservatively use input price. |
+| Mistral Small 4 | mistral | mistral-small-2603 | $0.15 | $0.15 | $0.15 | $0.6 | standard | Exact cache discount is not published per model; cached buckets conservatively use input price. |
+| Codestral | mistral | codestral-2508 | $0.3 | $0.3 | $0.3 | $0.9 | standard | Exact cache discount is not published per model; cached buckets conservatively use input price. |
+| Command A | cohere | command-a-03-2025 | $2.5 | $2.5 | $2.5 | $10 | standard | Current paid production model. No cache discount is published. |
+| Command R7B | cohere | command-r7b-12-2024 | $0.0375 | $0.0375 | $0.0375 | $0.15 | standard | Pinned paid model. No cache discount is published. |
+| Command R | cohere | command-r-08-2024 | $0.15 | $0.15 | $0.15 | $0.6 | standard | Pinned paid model. No cache discount is published. |
+| Command R+ | cohere | command-r-plus-08-2024 | $2.5 | $2.5 | $2.5 | $10 | standard | Pinned paid model. No cache discount is published. |
+| Command legacy | cohere | command | $1 | $1 | $1 | $2 | standard | Deprecated; retained for historical logs. No cache discount is published. |
+| Command Light legacy | cohere | command-light | $0.3 | $0.3 | $0.3 | $0.6 | standard | Deprecated; retained for historical logs. No cache discount is published. |
+| Command R legacy | cohere | command-r-03-2024 | $0.5 | $0.5 | $0.5 | $1.5 | standard | Deprecated; retained for historical logs. No cache discount is published. |
+| Command R+ legacy | cohere | command-r-plus-04-2024 | $3 | $3 | $3 | $15 | standard | Deprecated; retained for historical logs. No cache discount is published. |
+| Qwen 3.7 Max | dashscope, alibaba, qwen | qwen3.7-max-2026-06-08 | $2.5 | $0.5 | $3.125 | $7.5 | prompt ≤ 1,000,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3.7 Plus · long context | dashscope, alibaba, qwen | qwen3.7-plus-2026-05-26 | $1.2 | $0.24 | $1.5 | $4.8 | prompt ≥ 256,001 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3.7 Plus | dashscope, alibaba, qwen | qwen3.7-plus-2026-05-26 | $0.4 | $0.08 | $0.5 | $1.6 | prompt ≤ 256,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3 Max · ≤32K | dashscope, alibaba, qwen | qwen3-max-2026-01-23 | $1.2 | $0.24 | $1.5 | $6 | prompt ≥ 0; prompt ≤ 32,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3 Max · 32K–128K | dashscope, alibaba, qwen | qwen3-max-2026-01-23 | $2.4 | $0.48 | $3 | $12 | prompt ≥ 32,001; prompt ≤ 128,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3 Max · 128K–256K | dashscope, alibaba, qwen | qwen3-max-2026-01-23 | $3 | $0.6 | $3.75 | $15 | prompt ≥ 128,001; prompt ≤ 256,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3 Coder Plus · ≤32K | dashscope, alibaba, qwen | qwen3-coder-plus-2025-09-23 | $1 | $0.2 | $1.25 | $5 | prompt ≥ 0; prompt ≤ 32,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3 Coder Plus · 32K–128K | dashscope, alibaba, qwen | qwen3-coder-plus-2025-09-23 | $1.8 | $0.36 | $2.25 | $9 | prompt ≥ 32,001; prompt ≤ 128,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3 Coder Plus · 128K–256K | dashscope, alibaba, qwen | qwen3-coder-plus-2025-09-23 | $3 | $0.6 | $3.75 | $15 | prompt ≥ 128,001; prompt ≤ 256,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3 Coder Plus · 256K–1M | dashscope, alibaba, qwen | qwen3-coder-plus-2025-09-23 | $6 | $1.2 | $7.5 | $60 | prompt ≥ 256,001; prompt ≤ 1,000,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3 Coder Flash · ≤32K | dashscope, alibaba, qwen | qwen3-coder-flash-2025-07-28 | $0.3 | $0.06 | $0.375 | $1.5 | prompt ≥ 0; prompt ≤ 32,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3 Coder Flash · 32K–128K | dashscope, alibaba, qwen | qwen3-coder-flash-2025-07-28 | $0.5 | $0.1 | $0.625 | $2.5 | prompt ≥ 32,001; prompt ≤ 128,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3 Coder Flash · 128K–256K | dashscope, alibaba, qwen | qwen3-coder-flash-2025-07-28 | $0.8 | $0.16 | $1 | $4 | prompt ≥ 128,001; prompt ≤ 256,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| Qwen 3 Coder Flash · 256K–1M | dashscope, alibaba, qwen | qwen3-coder-flash-2025-07-28 | $1.6 | $0.32 | $2 | $9.6 | prompt ≥ 256,001; prompt ≤ 1,000,000 | Singapore international list price; cache read uses the implicit-cache rate. |
+| MiniMax M3 · long context | minimax | MiniMax-M3 | $0.6 | $0.12 | $0.6 | $2.4 | prompt ≥ 512,001 | Standard tier; no separate cache-write price is published. |
+| MiniMax M3 | minimax | MiniMax-M3 | $0.3 | $0.06 | $0.3 | $1.2 | prompt ≤ 512,000 | Standard tier; no separate cache-write price is published. |
+| MiniMax M2.7 | minimax | MiniMax-M2.7 | $0.3 | $0.06 | $0.375 | $1.2 | standard |  |
+| MiniMax M2.7 Highspeed | minimax | MiniMax-M2.7-highspeed | $0.6 | $0.06 | $0.375 | $2.4 | standard |  |
+| MiniMax M2.5 | minimax | MiniMax-M2.5 | $0.3 | $0.03 | $0.375 | $1.2 | standard |  |
+| MiniMax M2.5 Highspeed | minimax | MiniMax-M2.5-highspeed | $0.6 | $0.03 | $0.375 | $2.4 | standard |  |
+
+## Official sources
+
+- **openai**: https://developers.openai.com/api/docs/pricing
+- **anthropic**: https://platform.claude.com/docs/en/about-claude/pricing
+- **gemini**: https://ai.google.dev/gemini-api/docs/pricing
+- **deepseek**: https://api-docs.deepseek.com/quick_start/pricing/
+- **zai**: https://docs.z.ai/guides/overview/pricing
+- **kimi**: https://platform.kimi.ai/docs/pricing
+- **xai**: https://docs.x.ai/developers/pricing
+- **mistral**: https://mistral.ai/pricing/api/
+- **cohere**: https://cohere.com/pricing
+- **qwen**: https://www.alibabacloud.com/help/en/model-studio/model-pricing
+- **minimax**: https://platform.minimax.io/docs/guides/pricing-paygo
+
+## Accuracy boundaries
+
+- Route matching is case-insensitive. Only explicit model aliases and narrowly scoped version-suffix globs are included.
+- Prompt length is the sum of uncached input, cache-read, and cache-write buckets reported for a call.
+- DeepSeek V4 peak windows are evaluated from each model-call start timestamp in UTC.
+- Known promotions and retirements use inclusive `validFrom` / exclusive `validTo` instants; calls outside them remain unpriced unless a successor rule is published.
+- Anthropic cache writes use the 5-minute rate because Harness usage records do not expose cache TTL.
+- Qwen cache reads use the implicit-cache rate; explicit cache hits can be cheaper.
+- Unknown routes remain unpriced rather than inheriting a broad family wildcard.
+- A custom `pricing` array replaces the built-in catalog for that plugin instance.
+

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@syncended/dsh-usage",
-  "version": "0.1.0",
+  "version": "0.2.1",
   "description": "Token usage, model cost analytics, trends, and activity heatmaps for DeepSeek Harness",
   "type": "module",
   "main": "dist/index.js",
@@ -16,6 +16,7 @@
   },
   "files": [
     "dist",
+    "docs",
     "lib/client.js",
     "cordis.patch.yml",
     "README.md",
@@ -42,7 +43,9 @@
     "build": "pnpm clean && tsc -p tsconfig.json && node --check lib/client.js",
     "typecheck": "tsc -p tsconfig.json --noEmit",
     "test": "pnpm build && node --test test/*.test.mjs",
-    "check": "pnpm typecheck && pnpm test",
+    "docs:pricing": "node scripts/render-pricing-catalog.mjs",
+    "docs:check": "node scripts/render-pricing-catalog.mjs --check",
+    "check": "pnpm typecheck && pnpm test && pnpm docs:check",
     "prepare": "pnpm build"
   },
   "keywords": [