Show a spinner in the app browser embed until its apps arrive

The embed stayed blank while it checked the session with /whoami and loaded the SDK. <puter-user-menu> hides its own spinner once the embed's page loads, so its panel went blank for that whole time. The empty embed now shows the same spinner in the same place until it renders its apps or its unavailable message.

.github/workflows/puterjs-tests.yaml

@@ -176,7 +176,10 @@ jobs:
       - name: Checkout repository
         uses: actions/checkout@v7
 
+      # Missing when the PR leg failed; `coverage (pr)` already reports that.
       - name: Download PR coverage
+        id: pr-coverage
+        continue-on-error: true
         uses: actions/download-artifact@v8
         with:
           name: puterjs-coverage-pr
@@ -186,6 +189,7 @@ jobs:
       # base run) — then report without the comparison column.
       - name: Download base coverage
         id: base-coverage
+        if: steps.pr-coverage.outcome == 'success'
         continue-on-error: true
         uses: actions/download-artifact@v8
         with:
@@ -193,7 +197,7 @@ jobs:
           path: coverage-base
 
       - name: Report coverage on PR
-        if: steps.base-coverage.outcome == 'success'
+        if: steps.pr-coverage.outcome == 'success' && steps.base-coverage.outcome == 'success'
         uses: davelosert/vitest-coverage-report-action@v2
         with:
           name: puter.js SDK
@@ -202,7 +206,7 @@ jobs:
           json-summary-compare-path: coverage-base/coverage-summary.json
 
       - name: Report coverage on PR (no base comparison)
-        if: steps.base-coverage.outcome != 'success'
+        if: steps.pr-coverage.outcome == 'success' && steps.base-coverage.outcome != 'success'
         uses: davelosert/vitest-coverage-report-action@v2
         with:
           name: puter.js SDK

AGENTS.md

@@ -159,7 +159,9 @@ The reference has two kinds of page:
 
 An overview page answers "what is this API and is it what I need?", and a function page answers "how exactly does this call behave?". A change to one function usually only touches that function's page, so a new `kv.list()` option goes in `KV/list.md` without a matching update to `KV.md`. Don't copy a function's options, limits, or edge cases onto the overview page for completeness. Where a detail goes is a judgment call, not a hard rule: update the overview when a change affects how someone would describe the API as a whole.
 
-When you add a page, add it to [src/docs/src/sidebar.js](src/docs/src/sidebar.js) under its area, or it won't show up in the navigation. When you add a runnable example, add its HTML to `src/docs/src/playground/examples/<slug>.html`, register the slug in [src/docs/src/examples.js](src/docs/src/examples.js), and link the code block to it by putting the slug after the language with a semicolon (` ```html;kv-set `).
+When you add a page, add it to [src/docs/src/sidebar.js](src/docs/src/sidebar.js) under its area, or it won't show up in the navigation.
+
+**Sidebar order.** Order the API areas in the sidebar by how often developers use them, most common first. Overview stays first and Objects stays last. Inside an area, order the pages by use too, and keep pages that make up one feature next to each other, such as `share()`, `unshare()`, `listShared()`, and `listSharedByMe()`. A group goes where its most-used page would go. A helper that only supports another function, such as `listModels()` for `chat()`, goes by its own use. Put create, list, get, update, and delete in that order, and constants such as `MAX_KEY_SIZE` last. The function list on the area's overview page uses the sidebar's order. When you add a runnable example, add its HTML to `src/docs/src/playground/examples/<slug>.html`, register the slug in [src/docs/src/examples.js](src/docs/src/examples.js), and link the code block to it by putting the slug after the language with a semicolon (` ```html;kv-set `).
 
 ### Function page layout
 
@@ -200,6 +202,7 @@ These apply to reference pages and recipes alike.
 - **Describe behavior directly.** No quoted inner-voice phrases. Write "`isNew` is `true` when the person did not have access before", not "this tells 'I just gave this to someone' from 'they already had it'".
 - **Keep references easy to follow.** Name what a list or code block belongs to in the sentence directly above it. Name a thing once and keep that name. Introduce an actor as "a user" before giving them a role: "A user can create a team. The user who creates the team becomes its owner."
 - **Sentence mechanics.** Never start a sentence with inline code: write "To store data, use the [`set()`](/KV/set/) method." Link every inline mention of an API to its page, every time. No em dashes, and no colons or semicolons mid-sentence; a colon that introduces a code block is fine.
+- **No `##` or `###` headings inside tabs.** The table of contents lists every `##` and `###` heading on the page, including the ones in a hidden tab (an `example-content` block switched by the `example-group` pills), so its links lead to content the reader can't see. Put the heading above the tabs, or use `####` inside a tab, which the table of contents skips.
 - **Show the correct way.** No "don't" and no wrong-way-first examples, in recipes and playground examples alike. Mention a wrong approach only when developers already hit it in practice, and state how it fails as a fact: "`add()` reads a plain object as a path map, so the call rejects with …".
 
 ### Recipes

doc/embedding-app-browser.md

@@ -0,0 +1,147 @@
+# Embedded app browser
+
+`GET /embed/apps` on the deployment's main GUI origin renders the same app browser
+as the dashboard's Apps tab, without the desktop or dashboard navigation.
+
+It supports search, paging, saved folders, and saved ordering. It uses the same
+installed-app and recommendation data as the dashboard, including the user's
+removed recommendations. Editing, adding, reordering, renaming, and uninstalling
+are disabled. Selecting an app opens it in a new browser tab at its own URL, the
+same way the dashboard does, and the host page stays open. The app's token isn't
+put in the URL: the app's SDK asks the embed for it once the tab loads, so the
+embed must still be open then. If the iframe is removed first, the app falls back
+to its normal sign-in. Apps that need the desktop (built-in and godmode apps,
+private apps the user can't open, or a failed token request) open `/app/<name>`
+on the GUI origin, cut off from the embed. Website shortcuts open their external
+destination with no opener.
+The embed follows the OS's light or dark setting. A host with a fixed theme adds
+`theme=dark` or `theme=light` to the fragment (`#user=<uuid>&theme=dark`), since
+Chromium doesn't pass an iframe's `color-scheme` on to the page inside it.
+Apps opened from the embed carry `utm_source=puter.com&utm_medium=referral&utm_campaign=app_browser_embed`;
+the desktop and dashboard send `utm_campaign=desktop` and `utm_campaign=dashboard`.
+
+## Session and framing
+
+The embed shows the GUI session only to the user the host page is signed in as.
+The host names that user's `uuid` (from `puter.auth.getUser()`) in the iframe
+URL's fragment, as `/embed/apps#user=<uuid>`. The embed reads the session, asks
+`/whoami` which user it belongs to, and shows the signed-out view unless the two
+match. So a host that names no user, a malformed one, or a user other than the
+session's, gets the signed-out view, and so does a session the API refuses. A
+failed `/whoami` check shows the error state. The fragment never reaches the
+server and can only narrow what the embed shows: naming the session's own user
+gives the same view as a direct visit to the GUI. Changing the fragment of a
+loaded embed reloads it.
+
+The embed reads the existing GUI session from local storage on its own origin.
+If that storage holds no session, the embed gets the session from the GUI origin's session cookie through
+`/get-gui-token`. This matters in Safari, which gives a framed origin separate
+local storage under each top-level origin, so even a sibling subdomain's iframe
+can't see the GUI's stored token. The session cookie is still sent, because the
+host is on the same site. The embed doesn't write that token back to storage. It
+does not create accounts, show a login dialog, or accept a token from its parent.
+Being signed in on a subdomain alone does not create a session on the main GUI
+origin.
+
+Without a session it can show, the embed shows the recommended apps (`RecommendedAppsService`)
+that the server put in the page, and nothing else: no installed apps, folders, or
+saved order, and no puter.js. Selecting one opens it at the same URL a signed-in
+launch uses, with no opener and no token on offer, so the app signs in on its own.
+Apps that need the desktop open `/app/<name>`. If the page carries no recommended
+apps, or a load fails, the embed displays an unavailable/error state.
+
+The route's `Content-Security-Policy: frame-ancestors` allows the GUI origin and
+the embedding page's origin when it is exactly one label under `config.domain`
+with the scheme and port from `config.origin`. Subdomains the server answers
+itself (`api`, `dav`, `js`, `www`, the native-app hosts, and the host of
+`api_base_url`) and the user-content hosting domains (`static_hosting_domain`,
+`private_app_hosting_domain`, and their `_alt` variants) are refused. For
+example, `https://puter.com` permits `https://spreadsheet.puter.com` but not
+`https://api.puter.com`, and `http://puter.localhost:4100` permits
+`http://spreadsheet.puter.localhost:4100` but not
+`http://someone.site.puter.localhost:4100`. The embedding page is read from the
+iframe request's referrer, so the host must not suppress it (the browser default
+`strict-origin-when-cross-origin` is enough). All ancestors in a nested iframe
+chain must be allowed. Other GUI routes retain `X-Frame-Options: SAMEORIGIN`.
+
+The embed accepts no query parameters. Requests with a query string redirect to
+the same embed path so SDK bootstrap parameters cannot override the session or API
+origin. The host needs to allow the GUI origin in its own `frame-src` CSP if it
+sets one. If sandboxing the iframe, allow scripts, same-origin storage, popups,
+and popups to escape the sandbox.
+
+## Host integration
+
+A first-party app that signs in with puter.js can place `<puter-user-menu>` from
+the SDK in its header instead (see `src/puter-js/src/ui/components.md`). It opens
+this embed with the signed-in user named and closes it when the app's sign-in
+changes. The rest of this section is for a host that builds the iframe itself.
+
+The iframe can be shown whether or not the user is signed in. Create it when
+“More apps” is opened, naming the signed-in user in its fragment, and remove it
+on close so reopening loads current apps and preferences. Remove it when the
+host's own sign-in changes too, such as on sign-out, so an open embed doesn't
+keep the previous user. The host owns the modal and close control;
+there is no parent messaging API and no app list or credential is sent to it.
+The app browser fills its iframe and responds to its size. It shows a spinner until its apps load.
+
+This example assumes `puterOrigin` comes from the host app's deployment
+settings. Translate the button and frame labels using the host app's
+localization system.
+
+```html
+<button id="more-apps" type="button">More apps</button>
+<dialog id="apps-dialog" aria-label="More apps">
+  <button id="close-apps" type="button">Close</button>
+  <div id="apps-frame"></div>
+</dialog>
+<style>
+  #apps-dialog { padding: 12px; width: min(900px, 90vw); }
+  #apps-frame iframe { width: 100%; height: min(650px, 75vh); border: 0; }
+</style>
+<script type="module">
+  const puterOrigin = 'https://puter.com'; // Use this deployment's configured origin.
+  const dialog = document.querySelector('#apps-dialog');
+  const container = document.querySelector('#apps-frame');
+  const trigger = document.querySelector('#more-apps');
+
+  trigger.addEventListener('click', async () => {
+    const url = new URL('/embed/apps', puterOrigin);
+    // Signed out, the embed shows the recommended apps.
+    const user = puter.auth.isSignedIn() ? await puter.auth.getUser().catch(() => null) : null;
+    if (user) url.hash = `user=${encodeURIComponent(user.uuid)}`;
+    const frame = document.createElement('iframe');
+    frame.title = 'Your apps';
+    frame.src = url.href;
+    container.replaceChildren(frame);
+    dialog.showModal();
+  });
+  document.querySelector('#close-apps').addEventListener('click', () => dialog.close());
+  dialog.addEventListener('close', () => {
+    container.replaceChildren();
+    trigger.focus();
+  });
+</script>
+```
+
+## Implementation and checks
+
+`createAppBrowser` in `src/gui/src/UI/Dashboard/AppBrowser.js` owns the shared UI
+and data behavior; its `guestApps` option is the signed-out list. `TabApps.js`
+injects desktop actions. `appsEmbed.js` supplies new-tab launching through
+`launchAppInTab` (signed in) and `guestAppLaunchUrl` (signed out) in
+`src/gui/src/helpers/appTabLaunch.js`, which build the same app URL as
+`launch_app`, and loads only the embed entry, its SDK, and shared styles. `embedHostUser` in
+`src/gui/src/helpers/embedHostUser.js` reads the host's user from the fragment, and
+`embedHostTheme.js` beside it applies the host's theme.
+The normal GUI webpack build emits `bundle.min.js` and `apps-embed.min.js`.
+The `/embed/apps` route puts the recommended list in the page's config as
+`guestApps`. No puter.js API or app-list endpoint changes are needed.
+
+Run the GUI component/browser tests (with the SDK bundle built and Playwright
+Chromium installed), and the backend route tests:
+
+```sh
+npx vitest run --config src/gui/vitest.config.js src/gui/src/UI/Dashboard/AppBrowser
+npx vitest run --config src/backend/vitest.config.ts src/backend/server.test.ts
+```

doc/push-client.md

@@ -1,6 +1,6 @@
 # Browser push client
 
-Puter's desktop shows pushes through `/sw.js`. Core registers it, and shows Settings → Notifications, only when the `pushEnabled` gui param is `true`; core never sets it. A backend extension that delivers pushes sets it with `services.homepage.setGuiParam('pushEnabled', true)` and can load its own code into the worker with `services.homepage.registerServiceWorkerScript(path)`.
+Puter's desktop shows pushes through `/sw.js`. Core shows Settings → Notifications only when the `pushEnabled` gui param is `true`; core never sets it. The worker is registered only when push is turned on in the browser, and a load removes one left without a subscription. A backend extension that delivers pushes sets it with `services.homepage.setGuiParam('pushEnabled', true)` and can load its own code into the worker with `services.homepage.registerServiceWorkerScript(path)`.
 
 Core subscribes the browser to a push service only when the server also publishes `pushWebPushKey` (see [Built-in Web Push client](#built-in-web-push-client)), and it has no `/push/*` routes of its own. The device, app, preference and subscription endpoints below come from that extension. On a deployment without one, such as most self-hosted servers, they answer core's plain 404 (`code: 'not_found'`), which `puter.notifications` reports as `push_unavailable` and the Notifications tab shows as unavailable. A device the extension can't find is `device_not_found`.
 
@@ -22,12 +22,13 @@ When the browser replaces the subscription (`pushsubscriptionchange`), `sw.js` p
 
 Before a notification is pushed, the server asks the open clients whether they received it live. It sends `notif.receipt` with `{ uid }` on the user's sockets for the app that sent it, or on the desktop's sockets for Puter's own notifications. The SDK remembers the uids its page received through `onNotification()`, `onPush()` or the desktop's feed, and answers once its page has that notification, or with `{ received: false }` when it never gets it. The server waits up to 5 seconds (`notificationAckTimeoutMs`).
 
-An answer carries `visibilityState`, `focused` and `showsOwn`, the last being whether the page shows the notification itself. A website page shows it itself when it is hidden or unfocused, has an `onPush()` handler and notification permission, and no other tab of the same site is focused. Inside the desktop, the desktop shows it, and an app's `onPush()` result is never shown as a browser notification.
+An answer carries `visibilityState` and `focused` as they were when the page received the notification. Pages never show a browser notification themselves; pushes do.
 
-- An answer from a page that is visible and focused, or that shows the notification itself, stops the push.
-- Otherwise the notification is pushed. A website's pushes go only to the browsers that subscribed through its own worker, never to Puter's `/sw.js`. An app inside the desktop pushes through `/sw.js`, which shows the app's name and icon.
-- A hidden desktop tab shows no banner, so answers only from hidden desktop tabs don't stop a push of Puter's own notifications.
-- A client that can't answer, such as an older puter.js, holds the push back while it is connected.
+- An answer from a page that was visible and focused stops the push (`handled` on `notif.delivery.settled`). An older client that doesn't send `focused` counts while visible.
+- Otherwise the notification is pushed, to every browser the user subscribed. A website's pushes go only to the browsers that subscribed through its own worker, never to Puter's `/sw.js`. An app inside the desktop pushes through `/sw.js`, which shows the app's name and icon.
+- A client that can't answer, such as an older puter.js, doesn't hold the push back.
+
+The ask reaches only the sockets in the asking region. So that other regions hear about a page too, a page that receives a notification live while visible and focused also sends `notif.handled` with `{ uid }` on the socket it came on. Core relays it as the node-local `notif.handled` event with the socket's user and app, at most 60 a minute per user; an extension that delivers pushes decides what to do with it.
 
 ## Registering
 

src/backend/clients/event/types.ts

@@ -117,7 +117,7 @@ export type EventMap = {
         model_used: string;
         service_used: string;
         intended_service: string;
-        model_details: {
+        model_details?: {
             id: string;
             provider: string;
             input_cost_key: string;
@@ -636,6 +636,18 @@ export type EventMap = {
         /** A peer region holds a socket for the user. */
         peerRegions: boolean;
         responders: NotifDeliveryResponder[];
+        /** A page here got it live while visible and focused. */
+        handled: boolean;
+    };
+    /**
+     * A page reported that it got a notification live while visible and
+     * focused. Node-local; `appUid` is the reporting socket's app, absent for
+     * the desktop.
+     */
+    'notif.handled': {
+        userId: number;
+        appUid?: string;
+        uid: string;
     };
 
     // ---- Metering ----
@@ -861,6 +873,12 @@ export type EventMap = {
     /** One access token was revoked; only its own connections should drop. */
     'auth.access-token.revoked': { token_uid: string };
 
+    /** Same-cluster eviction for raw realtime WebSockets. */
+    'pubsub.realtime.auth-revoked': {
+        sessionUids?: string[];
+        tokenUid?: string;
+    };
+
     /**
      * A grant was withdrawn, so whatever was standing on it has to be settled
      * rather than left to fail its next check.
@@ -1061,10 +1079,11 @@ export type EventKey = keyof EventMap & string;
 // Generates a wildcard for every non-final dot-separated prefix of K.
 export type WildcardPrefixes<K extends string> =
     K extends `${infer Head}.${infer Tail}`
-        ? | `${Head}.*`
-          | (Tail extends `${string}.${string}`
-                ? `${Head}.${WildcardPrefixes<Tail>}`
-                : never)
+        ?
+              | `${Head}.*`
+              | (Tail extends `${string}.${string}`
+                    ? `${Head}.${WildcardPrefixes<Tail>}`
+                    : never)
         : never;
 
 export type ListenKey = EventKey | WildcardPrefixes<EventKey>;
@@ -1075,18 +1094,16 @@ export type MatchingEvents<P extends ListenKey> = P extends `${infer Prefix}.*`
 
 /** Acked, not acked, no socket anywhere, or sockets only in a peer region. */
 export type NotifDeliveryOutcome =
-    'received' | 'unreceived' | 'no-socket' | 'remote-only';
+    | 'received'
+    | 'unreceived'
+    | 'no-socket'
+    | 'remote-only';
 
 /** One socket's receipt for a live notification. */
 export interface NotifDeliveryResponder {
     visibilityState: 'visible' | 'hidden' | 'unknown';
     /** The page's window has focus. Absent from older clients. */
     focused?: boolean;
-    /**
-     * The page shows its own browser notification while hidden or unfocused.
-     * Absent from older clients.
-     */
-    showsOwn?: boolean;
 }
 
 export type EventMetadata = {

src/backend/controllers/homepage/HomepageController.test.ts

@@ -27,6 +27,7 @@ import type { Actor } from '../../core/actor.js';
 import { PuterRouter } from '../../core/http/PuterRouter.js';
 import { PuterServer } from '../../server.js';
 import { setupTestServer } from '../../testUtil.js';
+import { embedFrameAncestors } from './HomepageController.js';
 
 // ── Test harness ────────────────────────────────────────────────────
 //
@@ -529,3 +530,23 @@ describe('HomepageController GET /sw.js', () => {
         expect(route?.options.subdomain).toBe('');
     });
 });
+
+describe('embedFrameAncestors', () => {
+    const config = {
+        origin: 'https://example.test',
+        domain: 'example.test',
+        api_base_url: 'https://backend.example.test',
+    };
+
+    it('names a sibling subdomain that embeds the browser', () => {
+        expect(
+            embedFrameAncestors(config, 'https://sheets.example.test/doc/1'),
+        ).toBe("'self' https://sheets.example.test");
+    });
+
+    it('refuses the configured API host even off the api label', () => {
+        expect(
+            embedFrameAncestors(config, 'https://backend.example.test/x'),
+        ).toBe("'self'");
+    });
+});

src/backend/controllers/homepage/HomepageController.ts

@@ -22,6 +22,11 @@ import path from 'node:path';
 import { PuterController } from '../types.js';
 import { toAppShellView } from '../../util/appShellView.js';
 import type { PuterRouter } from '../../core/http/PuterRouter';
+import {
+    isReservedSubdomain,
+    subdomainLabelOf,
+} from '../../core/http/middleware/hostRedirects.js';
+import type { IConfig } from '../../types';
 import type {
     PuterHomepageService,
     PageMeta,
@@ -43,9 +48,9 @@ const APP_LANDING_LIMIT = {
 };
 
 /**
- * App descriptions can run long; search snippets and social cards want a
- * short blurb. Collapses whitespace, then cuts at a word boundary with an
- * ellipsis so the result never exceeds `APP_META_DESCRIPTION_MAX`.
+ * App descriptions can run long; search snippets and social cards want a short
+ * blurb. Collapses whitespace, then cuts at a word boundary with an ellipsis so
+ * the result never exceeds `APP_META_DESCRIPTION_MAX`.
  */
 export function appMetaDescription(text: string): string {
     const clean = text.replace(/\s+/g, ' ').trim();
@@ -56,6 +61,53 @@ export function appMetaDescription(text: string): string {
     return `${cut.replace(/[\s,;:.!?…—-]+$/, '')}…`;
 }
 
+/**
+ * `frame-ancestors` for the app browser embed. A wildcard over `*.<domain>`
+ * would also admit user-hosted sites whenever the hosting domains sit under it,
+ * so the embedding page is named exactly, from the request's referrer. Only a
+ * single label under `<domain>` qualifies, and not one this server answers
+ * itself (the API, WebDAV, SDK and native-app hosts).
+ */
+export function embedFrameAncestors(
+    config: Pick<
+        IConfig,
+        | 'origin'
+        | 'domain'
+        | 'api_base_url'
+        | 'static_hosting_domain'
+        | 'static_hosting_domain_alt'
+        | 'private_app_hosting_domain'
+        | 'private_app_hosting_domain_alt'
+    >,
+    referer: string | undefined,
+): string {
+    const gui = new URL(config.origin ?? `https://${config.domain}`);
+    let parent: URL;
+    try {
+        parent = new URL(referer ?? '');
+    } catch {
+        return "'self'";
+    }
+    const host = parent.hostname.toLowerCase();
+    const label = subdomainLabelOf(host, config.domain);
+    const hostingDomains = [
+        config.static_hosting_domain,
+        config.static_hosting_domain_alt,
+        config.private_app_hosting_domain,
+        config.private_app_hosting_domain_alt,
+    ]
+        .map((d) => (d ?? '').toLowerCase().split(':')[0])
+        .filter(Boolean);
+    const allowed =
+        parent.protocol === gui.protocol &&
+        parent.port === gui.port &&
+        label !== undefined &&
+        !isReservedSubdomain(label) &&
+        host !== new URL(config.api_base_url).hostname &&
+        !hostingDomains.some((d) => host === d || host.endsWith(`.${d}`));
+    return allowed ? `'self' ${parent.origin}` : "'self'";
+}
+
 /**
  * Routes that render the Puter GUI shell, plus a catch-all static fallback
  * under `<gui_assets_root>/src` for non-dist/src paths (images, fonts, lib
@@ -105,6 +157,27 @@ export class HomepageController extends PuterController {
 
         router.get('/', {}, (req, res) => sendShell(req, res));
 
+        router.get('/embed/apps', {}, async (req, res) => {
+            // SDK bootstrap parameters must not change this page's account or API origin.
+            // Checked on the raw URL; `req.query` drops pairs past the parser's limit.
+            if (req.originalUrl.includes('?')) {
+                res.redirect('/embed/apps');
+                return;
+            }
+            res.removeHeader('X-Frame-Options');
+            res.setHeader(
+                'Content-Security-Policy',
+                `frame-ancestors ${embedFrameAncestors(this.config, req.get('referer'))}`,
+            );
+            res.setHeader('Cache-Control', 'no-store');
+            // Shown to visitors without a session; a signed-in embed fetches
+            // its own list, so a failure here only costs signed-out visitors.
+            const guestApps = await this.services.recommendedApps
+                .getRecommendedApps()
+                .catch((): Array<Record<string, unknown>> => []);
+            res.send(homepage.renderAppBrowserEmbed({ guestApps }));
+        });
+
         router.get('/settings', {}, (req, res) => sendShell(req, res));
         router.get('/settings/*splat', {}, (req, res) => sendShell(req, res));
 

src/backend/controllers/puterai/PuterAIController.shapes.test.ts

@@ -201,6 +201,7 @@ describe('PuterAIController.openaiResponses parameter forwarding', () => {
                     context_management: [{ type: 'compaction' }],
                     prompt_cache_key: 'ck',
                     prompt_cache_retention: '24h',
+                    prompt_cache_options: { ttl: '12h', prewarm: true },
                     store: true,
                     truncation: 'auto',
                     service_tier: 'default',
@@ -226,6 +227,9 @@ describe('PuterAIController.openaiResponses parameter forwarding', () => {
             context_management: [{ type: 'compaction' }],
             prompt_cache_key: 'ck',
             prompt_cache_retention: '24h',
+            // Forwarded as sent; the provider keeps only the fields it
+            // accepts.
+            prompt_cache_options: { ttl: '12h', prewarm: true },
             // forced false regardless of the client's request.
             store: false,
             truncation: 'auto',
@@ -359,13 +363,18 @@ describe('PuterAIController.openaiResponses input normalization', () => {
                 role: 'user',
                 content: [
                     'a bare string part',
-                    { type: 'input_text', text: 'typed text' },
+                    {
+                        type: 'input_text',
+                        text: 'typed text',
+                        prompt_cache_breakpoint: { mode: 'explicit', ttl: '12h' },
+                    },
                     { type: 'output_text', text: 'echoed text' },
                     {
                         type: 'input_image',
                         detail: 'high',
                         image_url: 'https://img.test/a.png',
                         file_id: 'file_img',
+                        prompt_cache_breakpoint: { mode: 'explicit' },
                     },
                     { type: 'input_audio', input_audio: { data: 'AAA' } },
                     {
@@ -386,13 +395,18 @@ describe('PuterAIController.openaiResponses input normalization', () => {
                 role: 'user',
                 content: [
                     { type: 'text', text: 'a bare string part' },
-                    { type: 'text', text: 'typed text' },
+                    {
+                        type: 'text',
+                        text: 'typed text',
+                        prompt_cache_breakpoint: { mode: 'explicit', ttl: '12h' },
+                    },
                     { type: 'text', text: 'echoed text' },
                     {
                         type: 'image_url',
                         detail: 'high',
                         image_url: { url: 'https://img.test/a.png' },
                         file_id: 'file_img',
+                        prompt_cache_breakpoint: { mode: 'explicit' },
                     },
                     { type: 'input_audio', input_audio: { data: 'AAA' } },
                     {
@@ -639,6 +653,33 @@ describe('PuterAIController.openaiResponses output items', () => {
 // -- /openai/v1/chat/completions edges -------------------------------
 
 describe('PuterAIController.openaiChatCompletions translation edges', () => {
+    it('forwards the prompt-cache fields to the driver', async () => {
+        const completeSpy = stubChatComplete({
+            message: { role: 'assistant', content: 'ok' },
+        });
+
+        const { res } = makeRes();
+        await controller.openaiChatCompletions(
+            makeReq({
+                body: {
+                    model: 'gpt-test',
+                    messages: [{ role: 'user', content: 'hi' }],
+                    prompt_cache_key: 'ck',
+                    prompt_cache_retention: 'in_memory',
+                    prompt_cache_options: { ttl: '30m' },
+                },
+            }),
+            res,
+        );
+
+        const args = completeSpy.mock.calls[0]![0] as Record<string, unknown>;
+        expect(args).toMatchObject({
+            prompt_cache_key: 'ck',
+            prompt_cache_retention: 'in_memory',
+            prompt_cache_options: { ttl: '30m' },
+        });
+    });
+
     it('derives tool_calls from tool_use content blocks when the message has none', async () => {
         stubChatComplete({
             message: {

src/backend/controllers/puterai/PuterAIController.ts

@@ -447,6 +447,7 @@ export class PuterAIController extends PuterController {
                           body.reasoning_effort as ICompleteArguments['reasoning_effort'],
                   }
                 : {}),
+            ...promptCacheArgsFromBody(body),
             ...openaiCompatProvider(body),
             ...(body.fallback === false ? { fallback: false } : {}),
         };
@@ -842,15 +843,7 @@ export class PuterAIController extends PuterController {
                           body.compaction as ICompleteArguments['compaction'],
                   }
                 : {}),
-            ...(body.prompt_cache_key
-                ? { prompt_cache_key: String(body.prompt_cache_key) }
-                : {}),
-            ...(body.prompt_cache_retention
-                ? {
-                      prompt_cache_retention:
-                          body.prompt_cache_retention as ICompleteArguments['prompt_cache_retention'],
-                  }
-                : {}),
+            ...promptCacheArgsFromBody(body),
             // forced off regardless of what the caller sent — OpenAI
             // defaults `store` to true, which would persist state server-side
             // on OpenAI's end that this route gives no way to retrieve.
@@ -1343,6 +1336,33 @@ const openaiCompatProvider = (
     return { provider: DEFAULTS.openaiChat };
 };
 
+/**
+ * The prompt-cache fields both OpenAI-shaped routes forward as sent; the
+ * provider checks the values.
+ */
+const promptCacheArgsFromBody = (
+    body: Record<string, unknown>,
+): Pick<
+    ICompleteArguments,
+    'prompt_cache_key' | 'prompt_cache_retention' | 'prompt_cache_options'
+> => ({
+    ...(body.prompt_cache_key
+        ? { prompt_cache_key: String(body.prompt_cache_key) }
+        : {}),
+    ...(body.prompt_cache_retention
+        ? {
+              prompt_cache_retention:
+                  body.prompt_cache_retention as ICompleteArguments['prompt_cache_retention'],
+          }
+        : {}),
+    ...(body.prompt_cache_options
+        ? {
+              prompt_cache_options:
+                  body.prompt_cache_options as ICompleteArguments['prompt_cache_options'],
+          }
+        : {}),
+});
+
 const randomId = (): string => crypto.randomUUID().replace(/-/g, '');
 const generateId = (prefix: string): string => `${prefix}_${randomId()}`;
 
@@ -1590,12 +1610,24 @@ const parseJsonMaybe = (value: unknown): unknown => {
     }
 };
 
+/** An OpenAI `prompt_cache_breakpoint` marker, kept on the translated part. */
+const cacheBreakpointOf = (
+    p: Record<string, unknown>,
+): { prompt_cache_breakpoint?: unknown } =>
+    p.prompt_cache_breakpoint && typeof p.prompt_cache_breakpoint === 'object'
+        ? { prompt_cache_breakpoint: p.prompt_cache_breakpoint }
+        : {};
+
 const normalizeContentPart = (part: unknown): Record<string, unknown> => {
     if (typeof part === 'string') return { type: 'text', text: part };
     if (!part || typeof part !== 'object') return { type: 'text', text: '' };
     const p = part as Record<string, unknown>;
     if (p.type === 'input_text' || p.type === 'output_text') {
-        return { type: 'text', text: String(p.text ?? '') };
+        return {
+            type: 'text',
+            text: String(p.text ?? ''),
+            ...cacheBreakpointOf(p),
+        };
     }
     if (p.type === 'input_image') {
         return {
@@ -1603,6 +1635,7 @@ const normalizeContentPart = (part: unknown): Record<string, unknown> => {
             ...(p.detail ? { detail: p.detail } : {}),
             ...(p.image_url ? { image_url: { url: p.image_url } } : {}),
             ...(p.file_id ? { file_id: p.file_id } : {}),
+            ...cacheBreakpointOf(p),
         };
     }
     if (p.type === 'input_audio')
@@ -1614,6 +1647,7 @@ const normalizeContentPart = (part: unknown): Record<string, unknown> => {
             ...(p.file_id ? { file_id: p.file_id } : {}),
             ...(p.file_url ? { file_url: p.file_url } : {}),
             ...(p.filename ? { filename: p.filename } : {}),
+            ...cacheBreakpointOf(p),
         };
     }
     return p;

src/backend/core/http/middleware/hostRedirects.ts

@@ -27,15 +27,7 @@ import { HttpError, isHttpError } from '../HttpError';
 import { isLocalWorkerHost } from './localWorkerProxy';
 
 /** Native-app subdomains served via `nativeAppStatic`. */
-const NATIVE_APP_SUBDOMAINS = [
-    'about',
-    'developer',
-    'docs',
-    'editor',
-    'markus',
-    'pdf',
-    'apps',
-] as const;
+const NATIVE_APP_SUBDOMAINS = ['about', 'developer', 'docs', 'apps'] as const;
 
 /** Subset served out of a `dist/` subdirectory rather than the app root. */
 const NATIVE_APPS_WITH_DIST = new Set(['docs', 'developer']);
@@ -111,6 +103,10 @@ export const subdomainLabelOf = (
     return label && !label.includes('.') ? label : undefined;
 };
 
+/** Whether core itself answers `<name>.<domain>` (`www` redirects to the root). */
+export const isReservedSubdomain = (name: string): boolean =>
+    name === 'www' || RESERVED_SUBDOMAINS.has(name);
+
 /**
  * Record a claim on `<name>.<domain>`. Throws on a malformed or uppercase
  * label, a subdomain core already handles, or a name that is already claimed.
@@ -123,7 +119,7 @@ export const addClaimedSubdomain = (
     if (typeof name !== 'string' || !SUBDOMAIN_LABEL.test(name)) {
         throw new Error(`Invalid subdomain label: ${JSON.stringify(name)}`);
     }
-    if (name === 'www' || RESERVED_SUBDOMAINS.has(name)) {
+    if (isReservedSubdomain(name)) {
         throw new Error(`Subdomain '${name}' is reserved`);
     }
     if (claims.has(name)) {

src/backend/drivers/ai-chat/providers/azure/AzureResponsesProvider.test.ts

@@ -449,6 +449,22 @@ describe('AzureResponsesProvider.complete request shape', () => {
         expect(args.service_tier).toBe('default');
     });
 
+    it('sends the wire spelling of the in-memory cache retention', async () => {
+        const provider = makeProvider();
+        responsesCreateMock.mockResolvedValueOnce(okResponse);
+
+        await withTestActor(() =>
+            provider.complete({
+                model: 'gpt-5.3-codex',
+                messages: [{ role: 'user', content: 'hi' }],
+                prompt_cache_retention: 'in-memory',
+            } as never),
+        );
+
+        const [args] = responsesCreateMock.mock.calls[0]!;
+        expect(args.prompt_cache_retention).toBe('in_memory');
+    });
+
     it.each([
         ['previous_response_id', 'resp_1'],
         ['conversation', 'conv_1'],

src/backend/drivers/ai-chat/providers/azure/AzureResponsesProvider.ts

@@ -35,6 +35,7 @@ import * as OpenAiUtil from '../../utils/OpenAIUtil.js';
 import {
     clampReasoningEffort,
     openAICompatParams,
+    promptCacheRetentionToWire,
     rejectStatefulResponsesFields,
 } from '../../utils/openaiParams.js';
 import { buildCostsOverride } from '../../utils/pricing.js';
@@ -209,6 +210,9 @@ export class AzureResponsesProvider implements IChatProvider {
         const userIdentifier = upstreamUserIdentifier(actor);
         // Cache key defaults to the actor identifier; see upstreamUserIdentifier.
         const cacheKey = prompt_cache_key ?? userIdentifier;
+        const cacheRetention = promptCacheRetentionToWire(
+            prompt_cache_retention,
+        );
 
         // Resolve any `puter_path` content parts into inline base64 data URLs
         // before the Responses API sees them.
@@ -307,8 +311,8 @@ export class AzureResponsesProvider implements IChatProvider {
             ...(instructions !== undefined ? { instructions } : {}),
             ...(metadata !== undefined ? { metadata } : {}),
             ...(cacheKey !== undefined ? { prompt_cache_key: cacheKey } : {}),
-            ...(prompt_cache_retention !== undefined
-                ? { prompt_cache_retention }
+            ...(cacheRetention !== undefined
+                ? { prompt_cache_retention: cacheRetention }
                 : {}),
             ...(store !== undefined ? { store } : {}),
             ...(max_tokens !== undefined

src/backend/drivers/ai-chat/providers/meta/MetaProvider.ts

@@ -23,6 +23,7 @@ import { HttpError } from '../../../../core/http/HttpError.js';
 import type { MeteringService } from '../../../../services/metering/MeteringService.js';
 import { upstreamUserIdentifier } from '../../../util/upstreamIdentifier.js';
 import type { IChatModel, ICompleteArguments } from '../../types.js';
+import { promptCacheRetentionToWire } from '../../utils/openaiParams.js';
 import { sdkClientOptions } from '../../utils/sdkClient.js';
 import {
     type ChatProviderConfig,
@@ -107,12 +108,9 @@ export class MetaProvider extends OpenAICompatProvider {
                 ? requestedEffort
                 : undefined;
 
-        // Puter spells the in-memory retention with a hyphen; Meta's enum
-        // uses an underscore.
-        const cacheRetention =
-            args.prompt_cache_retention === 'in-memory'
-                ? 'in_memory'
-                : args.prompt_cache_retention;
+        const cacheRetention = promptCacheRetentionToWire(
+            args.prompt_cache_retention,
+        );
 
         // The identifier is Puter's abuse attribution, so `custom` can't
         // override it. Cache key defaults to it; see upstreamUserIdentifier.

src/backend/drivers/ai-chat/providers/openai/OpenAiChatCompletionsProvider.test.ts

@@ -249,6 +249,130 @@ describe('OpenAiChatProvider.complete argument validation', () => {
         expect(result).toEqual({ delegated: true });
     });
 
+    it('delegates a 12h cache lifetime to the Responses provider, which is the only API that writes it', async () => {
+        const { provider } = makeProvider();
+        const sibling = {
+            complete: vi.fn().mockResolvedValue({ delegated: true }),
+        };
+        provider.setResponsesProvider(
+            sibling as unknown as Parameters<
+                typeof provider.setResponsesProvider
+            >[0],
+        );
+
+        const params = {
+            model: 'gpt-6-luna',
+            messages: [{ role: 'user', content: 'hi' }],
+            prompt_cache_options: { ttl: '12h' as const },
+        };
+        const result = await withTestActor(() => provider.complete(params));
+
+        expect(createMock).not.toHaveBeenCalled();
+        expect(sibling.complete).toHaveBeenCalledWith(params);
+        expect(result).toEqual({ delegated: true });
+    });
+
+    it.each([
+        ['30m', { mode: 'explicit', ttl: '30m' }],
+        ['12h', { mode: 'explicit', ttl: '12h' }],
+    ])(
+        'delegates a %s breakpoint lifetime to the Responses provider, since Chat Completions rejects the field',
+        async (_ttl, breakpoint) => {
+            const { provider } = makeProvider();
+            const sibling = {
+                complete: vi.fn().mockResolvedValue({ delegated: true }),
+            };
+            provider.setResponsesProvider(
+                sibling as unknown as Parameters<
+                    typeof provider.setResponsesProvider
+                >[0],
+            );
+
+            const params = {
+                model: 'gpt-6-luna',
+                messages: [
+                    {
+                        role: 'system',
+                        content: [
+                            {
+                                type: 'text',
+                                text: 'long prefix',
+                                prompt_cache_breakpoint: breakpoint,
+                            },
+                        ],
+                    },
+                    { role: 'user', content: 'hi' },
+                ],
+            };
+            const result = await withTestActor(() =>
+                provider.complete(params as never),
+            );
+
+            expect(createMock).not.toHaveBeenCalled();
+            expect(sibling.complete).toHaveBeenCalledWith(params);
+            expect(result).toEqual({ delegated: true });
+        },
+    );
+
+    it('keeps a breakpoint without a lifetime on the Chat Completions path, intact', async () => {
+        const { provider } = makeProvider();
+        createMock.mockResolvedValueOnce({
+            choices: [
+                {
+                    message: { content: 'ok', role: 'assistant' },
+                    finish_reason: 'stop',
+                },
+            ],
+            usage: { prompt_tokens: 1, completion_tokens: 1 },
+        });
+
+        await withTestActor(() =>
+            provider.complete({
+                model: 'gpt-6-luna',
+                messages: [
+                    {
+                        role: 'system',
+                        content: [
+                            {
+                                type: 'text',
+                                text: 'long prefix',
+                                prompt_cache_breakpoint: { mode: 'explicit' },
+                            },
+                        ],
+                    },
+                    { role: 'user', content: 'hi' },
+                ],
+            } as never),
+        );
+
+        const [args] = createMock.mock.calls[0]!;
+        expect(args.messages[0]).toMatchObject({
+            role: 'system',
+            content: [
+                {
+                    type: 'text',
+                    text: 'long prefix',
+                    prompt_cache_breakpoint: { mode: 'explicit' },
+                },
+            ],
+        });
+    });
+
+    it('throws 400 for a 12h cache lifetime without a Responses sibling provider', async () => {
+        const { provider } = makeProvider();
+
+        await expect(
+            withTestActor(() =>
+                provider.complete({
+                    model: 'gpt-6-luna',
+                    messages: [{ role: 'user', content: 'hi' }],
+                    prompt_cache_options: { ttl: '12h' },
+                }),
+            ),
+        ).rejects.toMatchObject({ statusCode: 400 });
+        expect(createMock).not.toHaveBeenCalled();
+    });
+
     it('requestPricing delegates to the Responses sibling (web_search\'s own hold), and no-ops without one', async () => {
         const { provider } = makeProvider();
         const args = {
@@ -349,6 +473,31 @@ describe('OpenAiChatProvider.complete request shape', () => {
         expect(args.safety_identifier).toBe('puter-u42');
     });
 
+    it('forwards 30m prompt_cache_options and the wire spelling of prompt_cache_retention', async () => {
+        const { provider } = makeProvider();
+        createMock.mockResolvedValueOnce(baseCompletion);
+
+        await withTestActor(() =>
+            provider.complete({
+                model: 'gpt-6-luna',
+                messages: [{ role: 'user', content: 'hello' }],
+                prompt_cache_retention: 'in-memory',
+                prompt_cache_options: {
+                    mode: 'explicit',
+                    ttl: '30m',
+                    prewarm: true,
+                } as never,
+            }),
+        );
+
+        const [args] = createMock.mock.calls[0]!;
+        expect(args.prompt_cache_retention).toBe('in_memory');
+        expect(args.prompt_cache_options).toEqual({
+            mode: 'explicit',
+            ttl: '30m',
+        });
+    });
+
     it.each([
         'gpt-6-astra',
         'gpt-6.1-sol',
@@ -665,7 +814,7 @@ describe('OpenAiChatProvider.complete non-stream output', () => {
             },
         });
 
-        await withTestActor(() =>
+        const result = await withTestActor(() =>
             provider.complete({
                 model: 'gpt-6-luna',
                 messages: [{ role: 'user', content: 'hi' }],
@@ -685,6 +834,13 @@ describe('OpenAiChatProvider.complete non-stream output', () => {
             cached_tokens: 1000 * Number(luna.costs.cached_tokens),
             cache_write_tokens: 3000 * Number(luna.costs.cache_write_tokens),
         });
+        // Chat Completions only writes OpenAI's default 30m entries.
+        expect((result as { usageDetails: unknown }).usageDetails).toEqual({
+            inputTokens: 1000,
+            outputTokens: 12,
+            cacheReadTokens: 1000,
+            cacheWrite30mTokens: 3000,
+        });
     });
 
     it('zeroes cached_tokens when prompt_tokens_details is missing', async () => {

src/backend/drivers/ai-chat/providers/openai/OpenAiChatCompletionsProvider.ts

@@ -35,7 +35,10 @@ import { make_openai_tools } from '../../utils/FunctionCalling.js';
 import * as OpenAiUtil from '../../utils/OpenAIUtil.js';
 import {
     clampReasoningEffort,
+    messagesHaveCacheBreakpointTtl,
     openAICompatParams,
+    promptCacheOptionsToWire,
+    promptCacheRetentionToWire,
 } from '../../utils/openaiParams.js';
 import { OPEN_AI_MODELS } from './models.js';
 import type { OpenAiResponsesChatProvider } from './OpenAiChatResponsesProvider.js';
@@ -126,8 +129,11 @@ export class OpenAiChatProvider implements IChatProvider {
             temperature,
             text,
             prompt_cache_key,
+            prompt_cache_retention,
+            prompt_cache_options,
         } = params;
         let { messages, model } = params;
+        const cacheOptions = promptCacheOptionsToWire(prompt_cache_options);
         if (
             tools?.filter((e) => isWebSearchTool(e as Record<string, unknown>))
                 .length
@@ -157,6 +163,21 @@ export class OpenAiChatProvider implements IChatProvider {
             }
             return await this.#responsesProvider.complete(params);
         }
+        // Only the Responses API writes 12h cache entries or takes a lifetime
+        // on a `prompt_cache_breakpoint`; Chat Completions rejects both.
+        if (
+            cacheOptions?.ttl === '12h' ||
+            messagesHaveCacheBreakpointTtl(messages)
+        ) {
+            if (!this.#responsesProvider) {
+                throw new HttpError(
+                    400,
+                    'a prompt cache lifetime requires the OpenAI Responses provider, which is not configured',
+                    { legacyCode: 'bad_request' },
+                );
+            }
+            return await this.#responsesProvider.complete(params);
+        }
         // Validate messages
         if (!Array.isArray(messages)) {
             throw new HttpError(400, '`messages` must be an array', {
@@ -180,6 +201,9 @@ export class OpenAiChatProvider implements IChatProvider {
         const userIdentifier = upstreamUserIdentifier(actor);
         // Cache key defaults to the actor identifier; see upstreamUserIdentifier.
         const cacheKey = prompt_cache_key ?? userIdentifier;
+        const cacheRetention = promptCacheRetentionToWire(
+            prompt_cache_retention,
+        );
 
         // Strip Anthropic-only shape a fallback-replayed message can carry
         // (thinking/server-tool blocks, cache_control, citations) before the
@@ -216,6 +240,12 @@ export class OpenAiChatProvider implements IChatProvider {
             user: userIdentifier,
             safety_identifier: userIdentifier,
             ...(cacheKey !== undefined ? { prompt_cache_key: cacheKey } : {}),
+            ...(cacheRetention !== undefined
+                ? { prompt_cache_retention: cacheRetention }
+                : {}),
+            ...(cacheOptions !== undefined
+                ? { prompt_cache_options: cacheOptions }
+                : {}),
             messages: messages,
             model: modelUsed.id,
             ...(mappedTools?.length ? { tools: mappedTools } : {}),
@@ -254,28 +284,11 @@ export class OpenAiChatProvider implements IChatProvider {
         );
 
         return OpenAiUtil.handle_completion_output({
-            usage_calculator: ({ usage, setUsageCosts }) => {
-                const cachedTokens =
-                    usage.prompt_tokens_details?.cached_tokens ?? 0;
-                // GPT-5.6 and later bill cache writes at 1.25x input. They're
-                // reported inside `prompt_tokens`, like cached reads.
-                // The SDK doesn't type `cache_write_tokens` yet.
-                const cacheWriteTokens =
-                    (
-                        usage.prompt_tokens_details as
-                            { cache_write_tokens?: number } | undefined
-                    )?.cache_write_tokens ?? 0;
-                const trackedUsage = {
-                    prompt_tokens:
-                        (usage.prompt_tokens ?? 0) -
-                        cachedTokens -
-                        cacheWriteTokens,
-                    completion_tokens: usage.completion_tokens ?? 0,
-                    cached_tokens: cachedTokens,
-                    ...(cacheWriteTokens
-                        ? { cache_write_tokens: cacheWriteTokens }
-                        : {}),
-                };
+            usage_calculator: ({ usage, setUsageCosts, setUsageDetails }) => {
+                // GPT-5.6 and later report cache writes inside
+                // `prompt_tokens`, like cached reads; Chat Completions only
+                // writes the default 30m entries, billed at 1.25x input.
+                const trackedUsage = OpenAiUtil.splitCachedPrompt(usage);
 
                 const metered = meterChatUsage(
                     this.#meteringService,
@@ -285,6 +298,11 @@ export class OpenAiChatProvider implements IChatProvider {
                     trackedUsage,
                 );
                 setUsageCosts(metered.costs);
+                setUsageDetails(
+                    OpenAiUtil.usageDetailsFromTrackedUsage(metered.usage, {
+                        writeTtl: '30m',
+                    })!,
+                );
                 return metered.usage;
             },
             stream,

src/backend/drivers/ai-chat/providers/openai/OpenAiChatResponsesProvider.test.ts

@@ -246,6 +246,7 @@ describe('OpenAiResponsesChatProvider model catalog', () => {
                     prompt_tokens: input,
                     cached_tokens: cached,
                     cache_write_tokens: input * 1.25,
+                    cache_write_12h_tokens: input * 2.5,
                     completion_tokens: output,
                 },
                 context: 1_050_000,
@@ -844,7 +845,7 @@ describe('OpenAiResponsesChatProvider.complete non-stream output', () => {
             },
         });
 
-        await withTestActor(() =>
+        const result = await withTestActor(() =>
             provider.complete({
                 model: 'gpt-6-luna',
                 messages: [{ role: 'user', content: 'hi' }],
@@ -867,6 +868,115 @@ describe('OpenAiResponsesChatProvider.complete non-stream output', () => {
             cached_tokens: 1000 * Number(luna.costs.cached_tokens),
             cache_write_tokens: 3000 * Number(luna.costs.cache_write_tokens),
         });
+        // Writes without a per-lifetime breakdown are OpenAI's default 30m
+        // entries, and the neutral breakdown says so.
+        expect((result as { usageDetails: unknown }).usageDetails).toEqual({
+            inputTokens: 1000,
+            outputTokens: 20,
+            cacheReadTokens: 1000,
+            cacheWrite30mTokens: 3000,
+        });
+    });
+
+    it('bills 12h cache writes at 2.5x input and the other writes at 1.25x', async () => {
+        const luna = OPEN_AI_MODELS.find((m) => m.id === 'gpt-6-luna')!;
+        const { provider } = makeProvider();
+        // One request can write both lifetimes: an explicit 12h breakpoint
+        // plus the implicit 30m one OpenAI adds at the end of the prompt.
+        responsesCreateMock.mockResolvedValueOnce({
+            output: [{ role: 'assistant' }],
+            output_text: 'ok',
+            usage: {
+                input_tokens: 6026,
+                output_tokens: 5,
+                input_tokens_details: {
+                    cached_tokens: 0,
+                    cache_write_tokens: 6023,
+                    cache_write_tokens_details: {
+                        ttl_12h_tokens: 6012,
+                        ttl_30m_tokens: 11,
+                    },
+                },
+            },
+        });
+
+        const result = await withTestActor(() =>
+            provider.complete({
+                model: 'gpt-6-luna',
+                messages: [{ role: 'user', content: 'hi' }],
+                prompt_cache_options: { ttl: '12h' },
+            }),
+        );
+
+        const [usage, , , overrides] = recordSpy.mock.calls[0]!;
+        expect(usage).toEqual({
+            prompt_tokens: 3,
+            completion_tokens: 5,
+            cached_tokens: 0,
+            cache_write_tokens: 11,
+            cache_write_12h_tokens: 6012,
+        });
+        expect(luna.costs.cache_write_12h_tokens).toBe(
+            Number(luna.costs.prompt_tokens) * 2.5,
+        );
+        expect(overrides).toEqual({
+            prompt_tokens: 3 * Number(luna.costs.prompt_tokens),
+            completion_tokens: 5 * Number(luna.costs.completion_tokens),
+            cached_tokens: 0,
+            cache_write_tokens: 11 * Number(luna.costs.cache_write_tokens),
+            cache_write_12h_tokens:
+                6012 * Number(luna.costs.cache_write_12h_tokens),
+        });
+        expect((result as { usageDetails: unknown }).usageDetails).toEqual({
+            inputTokens: 3,
+            outputTokens: 5,
+            cacheWrite30mTokens: 11,
+            cacheWrite12hTokens: 6012,
+        });
+    });
+
+    it('forwards prompt_cache_options and the wire spelling of prompt_cache_retention', async () => {
+        const { provider } = makeProvider();
+        responsesCreateMock.mockResolvedValueOnce({
+            output: [{ role: 'assistant' }],
+            output_text: 'ok',
+            usage: { input_tokens: 1, output_tokens: 1 },
+        });
+
+        await withTestActor(() =>
+            provider.complete({
+                model: 'gpt-6-luna',
+                messages: [{ role: 'user', content: 'hi' }],
+                prompt_cache_retention: 'in-memory',
+                prompt_cache_options: {
+                    mode: 'explicit',
+                    ttl: '12h',
+                    comparison_response_id: 'resp_1',
+                } as never,
+            }),
+        );
+
+        const [args] = responsesCreateMock.mock.calls[0]!;
+        expect(args.prompt_cache_retention).toBe('in_memory');
+        expect(args.prompt_cache_options).toEqual({
+            mode: 'explicit',
+            ttl: '12h',
+        });
+    });
+
+    it('rejects a cache lifetime OpenAI does not offer before calling upstream', async () => {
+        const { provider } = makeProvider();
+
+        await expect(
+            withTestActor(() =>
+                provider.complete({
+                    model: 'gpt-6-luna',
+                    messages: [{ role: 'user', content: 'hi' }],
+                    prompt_cache_options: { ttl: '2h' } as never,
+                }),
+            ),
+        ).rejects.toMatchObject({ statusCode: 400 });
+        expect(responsesCreateMock).not.toHaveBeenCalled();
     });
 
     it.each([
@@ -1068,6 +1178,69 @@ describe('OpenAiResponsesChatProvider.complete streaming', () => {
         });
     });
 
+    it('labels streamed cache writes by lifetime on the final usage chunk', async () => {
+        const luna = OPEN_AI_MODELS.find((m) => m.id === 'gpt-6-luna')!;
+        const { provider } = makeProvider();
+        responsesCreateMock.mockReturnValueOnce(
+            asAsyncIterable([
+                { type: 'response.output_text.delta', delta: 'ok' },
+                {
+                    type: 'response.completed',
+                    response: {
+                        usage: {
+                            input_tokens: 6026,
+                            output_tokens: 5,
+                            input_tokens_details: {
+                                cached_tokens: 0,
+                                cache_write_tokens: 6023,
+                                cache_write_tokens_details: {
+                                    ttl_12h_tokens: 6012,
+                                    ttl_30m_tokens: 11,
+                                },
+                            },
+                        },
+                    },
+                },
+            ]),
+        );
+
+        const result = await withTestActor(() =>
+            provider.complete({
+                model: 'gpt-6-luna',
+                messages: [{ role: 'user', content: 'say ok' }],
+                stream: true,
+                prompt_cache_options: { ttl: '12h' },
+            }),
+        );
+        const harness = makeCapturingChatStream();
+        await (
+            result as {
+                init_chat_stream: (p: { chatStream: unknown }) => Promise<void>;
+            }
+        ).init_chat_stream({ chatStream: harness.chatStream });
+
+        const usageEvent = harness.events().find((e) => e.type === 'usage');
+        expect(usageEvent?.usage).toEqual({
+            prompt_tokens: 3,
+            completion_tokens: 5,
+            cached_tokens: 0,
+            cache_write_tokens: 11,
+            cache_write_12h_tokens: 6012,
+        });
+        expect(usageEvent?.usageDetails).toEqual({
+            inputTokens: 3,
+            outputTokens: 5,
+            cacheWrite30mTokens: 11,
+            cacheWrite12hTokens: 6012,
+        });
+        const [, , , overrides] = recordSpy.mock.calls[0]!;
+        expect(overrides).toMatchObject({
+            cache_write_tokens: 11 * Number(luna.costs.cache_write_tokens),
+            cache_write_12h_tokens:
+                6012 * Number(luna.costs.cache_write_12h_tokens),
+        });
+    });
+
     it('emits a tool_use block when response.output_item.done arrives with a function_call', async () => {
         const { provider } = makeProvider();
         responsesCreateMock.mockReturnValueOnce(

src/backend/drivers/ai-chat/providers/openai/OpenAiChatResponsesProvider.ts

@@ -32,6 +32,8 @@ import * as OpenAiUtil from '../../utils/OpenAIUtil.js';
 import {
     clampReasoningEffort,
     openAICompatParams,
+    promptCacheOptionsToWire,
+    promptCacheRetentionToWire,
     rejectStatefulResponsesFields,
 } from '../../utils/openaiParams.js';
 import { responseSamplingParams } from '../../utils/responseSampling.js';
@@ -150,6 +152,7 @@ export class OpenAiResponsesChatProvider implements IChatProvider {
         prompt,
         prompt_cache_key,
         prompt_cache_retention,
+        prompt_cache_options,
         store,
         top_p,
         truncation,
@@ -195,6 +198,10 @@ export class OpenAiResponsesChatProvider implements IChatProvider {
         const userIdentifier = upstreamUserIdentifier(actor);
         // Cache key defaults to the actor identifier; see upstreamUserIdentifier.
         const cacheKey = prompt_cache_key ?? userIdentifier;
+        const cacheRetention = promptCacheRetentionToWire(
+            prompt_cache_retention,
+        );
+        const cacheOptions = promptCacheOptionsToWire(prompt_cache_options);
 
         if (tools) {
             // Unravel tools to OpenAI Responses API format
@@ -288,8 +295,11 @@ export class OpenAiResponsesChatProvider implements IChatProvider {
             ...(instructions !== undefined ? { instructions } : {}),
             ...(metadata !== undefined ? { metadata } : {}),
             ...(cacheKey !== undefined ? { prompt_cache_key: cacheKey } : {}),
-            ...(prompt_cache_retention !== undefined
-                ? { prompt_cache_retention }
+            ...(cacheRetention !== undefined
+                ? { prompt_cache_retention: cacheRetention }
+                : {}),
+            ...(cacheOptions !== undefined
+                ? { prompt_cache_options: cacheOptions }
                 : {}),
             ...(store !== undefined ? { store } : {}),
             ...(max_tokens !== undefined
@@ -310,24 +320,17 @@ export class OpenAiResponsesChatProvider implements IChatProvider {
             { signal: Context.get('upstreamSignal') },
         );
         return OpenAiUtil.handle_completion_output_responses_api({
-            usage_calculator: ({ usage, webSearchCalls, setUsageCosts }) => {
-                const cachedTokens =
-                    (usage as any).input_tokens_details?.cached_tokens ?? 0;
-                // GPT-5.6 and later bill cache writes at 1.25x input. They're
-                // reported inside `input_tokens`, like cached reads.
-                const cacheWriteTokens =
-                    (usage as any).input_tokens_details?.cache_write_tokens ??
-                    0;
+            usage_calculator: ({
+                usage,
+                webSearchCalls,
+                setUsageCosts,
+                setUsageDetails,
+            }) => {
+                // GPT-5.6 and later report cache writes inside `input_tokens`,
+                // like cached reads, split by lifetime (30m and 12h entries
+                // cost 1.25x and 2.5x input to write).
                 const trackedUsage: Record<string, number> = {
-                    prompt_tokens:
-                        ((usage as any).input_tokens ?? 0) -
-                        cachedTokens -
-                        cacheWriteTokens,
-                    completion_tokens: (usage as any).output_tokens ?? 0,
-                    cached_tokens: cachedTokens,
-                    ...(cacheWriteTokens
-                        ? { cache_write_tokens: cacheWriteTokens }
-                        : {}),
+                    ...OpenAiUtil.splitCachedPrompt(usage),
                     ...(webSearchCalls
                         ? { web_search_calls: webSearchCalls }
                         : {}),
@@ -356,6 +359,11 @@ export class OpenAiResponsesChatProvider implements IChatProvider {
                         : {},
                 );
                 setUsageCosts?.(metered.costs);
+                setUsageDetails?.(
+                    OpenAiUtil.usageDetailsFromTrackedUsage(metered.usage, {
+                        writeTtl: '30m',
+                    })!,
+                );
                 return metered.usage;
             },
             stream,

src/backend/drivers/ai-chat/providers/openai/models.ts

@@ -23,6 +23,8 @@ import type { IChatModel } from '../../types.js';
 
 // Prompts over 272K input tokens are billed at 2x input (cached reads and
 // cache writes included) and 1.5x output for the full request.
+// GPT-5.6 and later bill cache writes by the lifetime requested: 1.25x input
+// for the default 30m entries, 2.5x for 12h entries.
 const GPT_LONG_CONTEXT_PRICING = {
     threshold: 272_000,
     input_multiplier: 2,
@@ -49,6 +51,7 @@ export const OPEN_AI_MODELS: IChatModel[] = [
             prompt_tokens: 100,
             cached_tokens: 5,
             cache_write_tokens: 100 * 1.25,
+            cache_write_12h_tokens: 100 * 2.5,
             completion_tokens: 500,
         },
         long_context_pricing: GPT_LONG_CONTEXT_PRICING,
@@ -74,6 +77,7 @@ export const OPEN_AI_MODELS: IChatModel[] = [
             prompt_tokens: 200,
             cached_tokens: 10,
             cache_write_tokens: 200 * 1.25,
+            cache_write_12h_tokens: 200 * 2.5,
             completion_tokens: 1000,
         },
         long_context_pricing: GPT_LONG_CONTEXT_PRICING,
@@ -99,6 +103,7 @@ export const OPEN_AI_MODELS: IChatModel[] = [
             prompt_tokens: 200,
             cached_tokens: 20,
             cache_write_tokens: 200 * 1.25,
+            cache_write_12h_tokens: 200 * 2.5,
             completion_tokens: 1000,
         },
         long_context_pricing: GPT_LONG_CONTEXT_PRICING,
@@ -124,6 +129,7 @@ export const OPEN_AI_MODELS: IChatModel[] = [
             prompt_tokens: 10,
             cached_tokens: 1,
             cache_write_tokens: 10 * 1.25,
+            cache_write_12h_tokens: 10 * 2.5,
             completion_tokens: 50,
         },
         long_context_pricing: GPT_LONG_CONTEXT_PRICING,
@@ -149,6 +155,7 @@ export const OPEN_AI_MODELS: IChatModel[] = [
             prompt_tokens: 1000,
             cached_tokens: 100,
             cache_write_tokens: 1000 * 1.25,
+            cache_write_12h_tokens: 1000 * 2.5,
             completion_tokens: 5000,
         },
         long_context_pricing: GPT_LONG_CONTEXT_PRICING,
@@ -174,6 +181,7 @@ export const OPEN_AI_MODELS: IChatModel[] = [
             prompt_tokens: 400,
             cached_tokens: 40,
             cache_write_tokens: 400 * 1.25,
+            cache_write_12h_tokens: 400 * 2.5,
             completion_tokens: 2000,
         },
         long_context_pricing: GPT_LONG_CONTEXT_PRICING,
@@ -198,6 +206,7 @@ export const OPEN_AI_MODELS: IChatModel[] = [
             prompt_tokens: 200,
             cached_tokens: 20,
             cache_write_tokens: 200 * 1.25,
+            cache_write_12h_tokens: 200 * 2.5,
             completion_tokens: 1200,
         },
         long_context_pricing: GPT_LONG_CONTEXT_PRICING,
@@ -222,6 +231,7 @@ export const OPEN_AI_MODELS: IChatModel[] = [
             prompt_tokens: 20,
             cached_tokens: 2,
             cache_write_tokens: 20 * 1.25,
+            cache_write_12h_tokens: 20 * 2.5,
             completion_tokens: 120,
         },
         long_context_pricing: GPT_LONG_CONTEXT_PRICING,

src/backend/drivers/ai-chat/types.ts

@@ -145,13 +145,19 @@ export type StopReason =
     | 'model_context_window_exceeded'
     | (string & {});
 
-/** Provider-neutral usage breakdown. `inputTokens` excludes cache reads/writes. */
+/**
+ * Provider-neutral usage breakdown. `inputTokens` excludes cache reads/writes.
+ * Cache writes are keyed by the lifetime the entry was written with: Claude and
+ * Moonshot write 5m and 1h entries, OpenAI writes 30m and 12h entries.
+ */
 export interface UsageDetails {
     inputTokens: number;
     outputTokens: number; // includes reasoning
     cacheReadTokens?: number;
     cacheWrite5mTokens?: number;
+    cacheWrite30mTokens?: number;
     cacheWrite1hTokens?: number;
+    cacheWrite12hTokens?: number;
     reasoningTokens?: number;
     webSearchRequests?: number;
     webFetchRequests?: number;
@@ -168,10 +174,22 @@ export interface UsageDetails {
         outputTokens: number;
         cacheReadTokens?: number;
         cacheWrite5mTokens?: number;
+        cacheWrite30mTokens?: number;
         cacheWrite1hTokens?: number;
+        cacheWrite12hTokens?: number;
     }>;
 }
 
+/**
+ * OpenAI `prompt_cache_options`, GPT-5.6 and later. `ttl` is the minimum
+ * lifetime of every cache entry the request writes; `12h` entries cost twice
+ * what `30m` entries do to write and only the Responses API writes them.
+ */
+export interface PromptCacheOptions {
+    mode?: 'implicit' | 'explicit';
+    ttl?: '30m' | '12h';
+}
+
 export interface ICompleteArguments {
     messages: PuterMessage[];
     provider?: string;
@@ -234,7 +252,9 @@ export interface ICompleteArguments {
     metadata?: Record<string, string>;
     prompt?: unknown;
     prompt_cache_key?: string;
-    prompt_cache_retention?: 'in-memory' | '24h' | undefined;
+    /** `in-memory` is accepted as a spelling of OpenAI's `in_memory`. */
+    prompt_cache_retention?: 'in-memory' | 'in_memory' | '24h' | undefined;
+    prompt_cache_options?: PromptCacheOptions;
     store?: boolean;
     top_p?: number;
     truncation?: 'auto' | 'disabled' | undefined;

src/backend/drivers/ai-chat/utils/OpenAIUtil.js

@@ -481,43 +481,86 @@ export const extractMeteredUsage = (usage) => {
 };
 
 /**
- * Chat Completions usage as it's metered: `prompt_tokens` reports cached reads
- * inside its count, so they come out of it and are priced only once, under
- * `cached_tokens`.
+ * @typedef {{
+ *     cached_tokens?: number | null;
+ *     cache_write_tokens?: number | null;
+ *     cache_write_tokens_details?: {
+ *         ttl_30m_tokens?: number | null;
+ *         ttl_12h_tokens?: number | null;
+ *     } | null;
+ * }} OpenAiCacheDetails
+ */
+
+/**
+ * OpenAI-shaped usage as it's metered. The prompt count reports cached reads
+ * and cache writes inside it, so they come out of it and are priced once each
+ * under their own keys: `cached_tokens` for reads, `cache_write_tokens` for
+ * writes at the model's default lifetime and `cache_write_12h_tokens` for
+ * writes at OpenAI's 12h lifetime. Without the per-lifetime breakdown every
+ * write is a default-lifetime write: Chat Completions writes no 12h entries,
+ * and Responses reports the breakdown whenever a request asked for them. Reads
+ * the Chat Completions (`prompt_tokens`) and Responses (`input_tokens`)
+ * spellings alike.
  *
- * @param {import('openai/resources/completions.mjs').CompletionUsage | undefined} usage
+ * @param {{
+ *           prompt_tokens?: number;
+ *           completion_tokens?: number;
+ *           prompt_tokens_details?: OpenAiCacheDetails | null;
+ *           input_tokens?: number;
+ *           output_tokens?: number;
+ *           input_tokens_details?: OpenAiCacheDetails | null;
+ *       }
+ *     | undefined} usage
  */
 export const splitCachedPrompt = (usage) => {
-    const details = usage?.prompt_tokens_details;
+    const details = usage?.prompt_tokens_details ?? usage?.input_tokens_details;
     const cachedTokens = details?.cached_tokens ?? 0;
-    // Moonshot counts the prompt tokens it wrote to its prefix cache here;
-    // they're billed at the model's write rate, not the plain prompt rate.
+    // Moonshot and GPT-5.6+ count the prompt tokens they wrote to the prefix
+    // cache here; they're billed at the model's write rate, not the prompt rate.
     const writeTokens =
         typeof details?.cache_write_tokens === 'number'
             ? details.cache_write_tokens
             : undefined;
+    const reported12h = details?.cache_write_tokens_details?.ttl_12h_tokens;
+    const write12hTokens =
+        typeof reported12h === 'number'
+            ? Math.min(reported12h, writeTokens ?? 0)
+            : undefined;
     return {
         prompt_tokens:
-            (usage?.prompt_tokens ?? 0) - cachedTokens - (writeTokens ?? 0),
-        completion_tokens: usage?.completion_tokens ?? 0,
+            (usage?.prompt_tokens ?? usage?.input_tokens ?? 0) -
+            cachedTokens -
+            (writeTokens ?? 0),
+        completion_tokens:
+            usage?.completion_tokens ?? usage?.output_tokens ?? 0,
         cached_tokens: cachedTokens,
         ...(writeTokens !== undefined
-            ? { cache_write_tokens: writeTokens }
+            ? { cache_write_tokens: writeTokens - (write12hTokens ?? 0) }
+            : {}),
+        ...(write12hTokens !== undefined
+            ? { cache_write_12h_tokens: write12hTokens }
             : {}),
     };
 };
 
 /**
- * The `{prompt_tokens, completion_tokens, cached_tokens, cache_write_tokens}`
- * shape every OpenAI-family `usage_calculator` already returns for billing,
- * mapped to the provider-neutral `UsageDetails` the end-of-stream/non-stream
- * result carries. `inputTokens`/`outputTokens` are the same numbers already
- * billed — this never introduces a new figure, only exposes the breakdown.
+ * The `{prompt_tokens, completion_tokens, cached_tokens, cache_write_tokens,
+ * cache_write_12h_tokens}` shape every OpenAI-family `usage_calculator` already
+ * returns for billing, mapped to the provider-neutral `UsageDetails` the
+ * end-of-stream/non-stream result carries. `inputTokens`/`outputTokens` are the
+ * same numbers already billed — this never introduces a new figure, only
+ * exposes the breakdown.
  *
  * @param {Record<string, number> | undefined | null} trackedUsage
+ * @param {{ writeTtl?: '5m' | '30m' }} [opts] The lifetime `cache_write_tokens`
+ *   were written with: 5m unless the provider says otherwise (OpenAI's default
+ *   entries live 30m).
  * @returns {import('../types.js').UsageDetails | undefined}
  */
-export const usageDetailsFromTrackedUsage = (trackedUsage) => {
+export const usageDetailsFromTrackedUsage = (
+    trackedUsage,
+    { writeTtl = '5m' } = {},
+) => {
     if (!trackedUsage || typeof trackedUsage !== 'object') return undefined;
     // OpenRouter, Infron and Ollama track `{prompt, completion,
     // input_cache_read}` instead; any other shape is left to the driver.
@@ -526,12 +569,19 @@ export const usageDetailsFromTrackedUsage = (trackedUsage) => {
     if (input === undefined && output === undefined) return undefined;
     const cacheRead =
         trackedUsage.cached_tokens ?? trackedUsage.input_cache_read;
+    const cacheWrite = trackedUsage.cache_write_tokens;
     return {
         inputTokens: input ?? 0,
         outputTokens: output ?? 0,
         ...(cacheRead ? { cacheReadTokens: cacheRead } : {}),
-        ...(trackedUsage.cache_write_tokens
-            ? { cacheWrite5mTokens: trackedUsage.cache_write_tokens }
+        ...(cacheWrite && writeTtl === '30m'
+            ? { cacheWrite30mTokens: cacheWrite }
+            : {}),
+        ...(cacheWrite && writeTtl !== '30m'
+            ? { cacheWrite5mTokens: cacheWrite }
+            : {}),
+        ...(trackedUsage.cache_write_12h_tokens
+            ? { cacheWrite12hTokens: trackedUsage.cache_write_12h_tokens }
             : {}),
     };
 };
@@ -864,12 +914,16 @@ export const create_chat_stream_handler_responses_api =
         // Missing usage is left undefined rather than fed to the calculator —
         // see the sibling handler above, including why usage is reported
         // before the block flushes.
+        let usageDetails;
         const usage = last_usage
             ? usage_calculator({
                   usage: last_usage,
                   webSearchCalls,
                   tool_usage: toolUsage,
                   setUsageCosts: (c) => chatStream.setUsageCosts(c),
+                  setUsageDetails: (d) => {
+                      usageDetails = d;
+                  },
               })
             : undefined;
         chatStream.reportUsage(usage);
@@ -877,7 +931,7 @@ export const create_chat_stream_handler_responses_api =
             reason:
                 incompleteReason ?? (sawFunctionCall ? 'tool_use' : 'end_turn'),
         });
-        const usageDetails = usageDetailsFromTrackedUsage(usage);
+        usageDetails ??= usageDetailsFromTrackedUsage(usage);
         if (usageDetails) chatStream.setUsageDetails(usageDetails);
 
         textblock.end();
@@ -1015,6 +1069,7 @@ export const handle_completion_output = async (
  *     webSearchCalls?: number;
  *     tool_usage?: Record<string, unknown>;
  *     setUsageCosts?: (costs: Record<string, number>) => void;
+ *     setUsageDetails?: (details: import('../types').UsageDetails) => void;
  * }) => unknown} params.usage_calculator
  * @param {() => Promise<void>} [params.finally_fn]
  * @returns {ReturnType<import('../types').IChatProvider['complete']>}
@@ -1156,12 +1211,15 @@ export const handle_completion_output_responses_api = async ({
               setUsageCosts: (c) => {
                   ret.usageCosts = c;
               },
+              setUsageDetails: (d) => {
+                  ret.usageDetails = d;
+              },
           })
         : {
               input_tokens: completion.usage.input_tokens,
               output_tokens: completion.usage.output_tokens,
           };
-    if (usage_calculator) {
+    if (usage_calculator && ret.usageDetails === undefined) {
         const usageDetails = usageDetailsFromTrackedUsage(ret.usage);
         if (usageDetails) ret.usageDetails = usageDetails;
     }

src/backend/drivers/ai-chat/utils/OpenAIUtil.test.ts

@@ -247,6 +247,40 @@ describe('usageDetailsFromTrackedUsage', () => {
         });
     });
 
+    it('labels cache writes by the lifetime the provider says it wrote them with', () => {
+        expect(
+            usageDetailsFromTrackedUsage(
+                {
+                    prompt_tokens: 10,
+                    completion_tokens: 5,
+                    cached_tokens: 3,
+                    cache_write_tokens: 11,
+                    cache_write_12h_tokens: 6012,
+                },
+                { writeTtl: '30m' },
+            ),
+        ).toEqual({
+            inputTokens: 10,
+            outputTokens: 5,
+            cacheReadTokens: 3,
+            cacheWrite30mTokens: 11,
+            cacheWrite12hTokens: 6012,
+        });
+        // Zero counts, as OpenAI reports on a cache hit, add no field.
+        expect(
+            usageDetailsFromTrackedUsage(
+                {
+                    prompt_tokens: 10,
+                    completion_tokens: 5,
+                    cached_tokens: 3,
+                    cache_write_tokens: 0,
+                    cache_write_12h_tokens: 0,
+                },
+                { writeTtl: '30m' },
+            ),
+        ).toEqual({ inputTokens: 10, outputTokens: 5, cacheReadTokens: 3 });
+    });
+
     it('returns undefined for a missing/non-object usage', () => {
         expect(usageDetailsFromTrackedUsage(undefined)).toBeUndefined();
         expect(usageDetailsFromTrackedUsage(null)).toBeUndefined();
@@ -962,6 +996,66 @@ describe('splitCachedPrompt', () => {
             cached_tokens: 3,
         });
     });
+
+    it('reads the Responses usage spelling', () => {
+        expect(
+            splitCachedPrompt({
+                input_tokens: 6026,
+                output_tokens: 5,
+                input_tokens_details: { cached_tokens: 0, cache_write_tokens: 6023 },
+            }),
+        ).toEqual({
+            prompt_tokens: 3,
+            completion_tokens: 5,
+            cached_tokens: 0,
+            cache_write_tokens: 6023,
+        });
+    });
+
+    it('splits 12h writes out of the write count when the breakdown is reported', () => {
+        // One request can write both lifetimes: an explicit 12h breakpoint
+        // plus the implicit 30m one OpenAI adds at the end of the prompt.
+        expect(
+            splitCachedPrompt({
+                input_tokens: 6026,
+                output_tokens: 5,
+                input_tokens_details: {
+                    cached_tokens: 0,
+                    cache_write_tokens: 6023,
+                    cache_write_tokens_details: {
+                        ttl_12h_tokens: 6012,
+                        ttl_30m_tokens: 11,
+                    },
+                },
+            }),
+        ).toEqual({
+            prompt_tokens: 3,
+            completion_tokens: 5,
+            cached_tokens: 0,
+            cache_write_tokens: 11,
+            cache_write_12h_tokens: 6012,
+        });
+    });
+
+    it('never reports more 12h writes than writes in total', () => {
+        expect(
+            splitCachedPrompt({
+                input_tokens: 100,
+                output_tokens: 1,
+                input_tokens_details: {
+                    cached_tokens: 0,
+                    cache_write_tokens: 40,
+                    cache_write_tokens_details: { ttl_12h_tokens: 50 },
+                },
+            }),
+        ).toEqual({
+            prompt_tokens: 60,
+            completion_tokens: 1,
+            cached_tokens: 0,
+            cache_write_tokens: 0,
+            cache_write_12h_tokens: 40,
+        });
+    });
 });
 
 // ── recorded usage costs ────────────────────────────────────────────

src/backend/drivers/ai-chat/utils/openaiParams.test.ts

@@ -22,11 +22,14 @@ import { describe, expect, it } from 'vitest';
 import type { ICompleteArguments } from '../types.js';
 import {
     clampReasoningEffort,
+    messagesHaveCacheBreakpointTtl,
     openAICompatParams,
     outputFormatFromResponseFormat,
     outputFormatFromResponsesText,
     outputFormatToResponseFormat,
     outputFormatToResponsesText,
+    promptCacheOptionsToWire,
+    promptCacheRetentionToWire,
     rejectStatefulResponsesFields,
     toolChoiceFromWire,
     toolChoiceToWire,
@@ -334,3 +337,109 @@ describe('rejectStatefulResponsesFields', () => {
         ).toThrow(/previous_response_id/);
     });
 });
+
+describe('promptCacheRetentionToWire', () => {
+    it('spells the in-memory retention the way the upstreams expect', () => {
+        expect(promptCacheRetentionToWire('in-memory')).toBe('in_memory');
+        expect(promptCacheRetentionToWire('in_memory')).toBe('in_memory');
+        expect(promptCacheRetentionToWire('24h')).toBe('24h');
+        expect(promptCacheRetentionToWire(undefined)).toBeUndefined();
+    });
+});
+
+describe('promptCacheOptionsToWire', () => {
+    it('forwards mode and ttl and nothing else', () => {
+        expect(
+            promptCacheOptionsToWire({
+                mode: 'explicit',
+                ttl: '12h',
+                prewarm: true,
+                comparison_response_id: 'resp_1',
+            }),
+        ).toEqual({ mode: 'explicit', ttl: '12h' });
+        expect(promptCacheOptionsToWire({ ttl: '30m' })).toEqual({
+            ttl: '30m',
+        });
+    });
+
+    it('sends nothing for an absent or empty object', () => {
+        expect(promptCacheOptionsToWire(undefined)).toBeUndefined();
+        expect(promptCacheOptionsToWire(null)).toBeUndefined();
+        expect(promptCacheOptionsToWire({})).toBeUndefined();
+        expect(promptCacheOptionsToWire({ prewarm: true })).toBeUndefined();
+    });
+
+    it('rejects a lifetime or mode OpenAI does not offer', () => {
+        expect(() => promptCacheOptionsToWire({ ttl: '2h' })).toThrow(
+            expect.objectContaining({ statusCode: 400 }),
+        );
+        expect(() => promptCacheOptionsToWire({ ttl: '24h' })).toThrow(
+            /30m.*12h/,
+        );
+        expect(() => promptCacheOptionsToWire({ mode: 'auto' })).toThrow(
+            /implicit.*explicit/,
+        );
+        expect(() => promptCacheOptionsToWire('12h')).toThrow(
+            /must be an object/,
+        );
+    });
+});
+
+describe('messagesHaveCacheBreakpointTtl', () => {
+    it('finds a breakpoint lifetime on any content part', () => {
+        expect(
+            messagesHaveCacheBreakpointTtl([
+                { role: 'user', content: 'plain' },
+                {
+                    role: 'system',
+                    content: [
+                        { type: 'text', text: 'a' },
+                        {
+                            type: 'text',
+                            text: 'b',
+                            prompt_cache_breakpoint: {
+                                mode: 'explicit',
+                                ttl: '30m',
+                            },
+                        },
+                    ],
+                },
+            ]),
+        ).toBe(true);
+        expect(
+            messagesHaveCacheBreakpointTtl([
+                {
+                    role: 'user',
+                    content: {
+                        type: 'text',
+                        text: 'a',
+                        prompt_cache_breakpoint: { ttl: '12h' },
+                    },
+                },
+            ]),
+        ).toBe(true);
+    });
+
+    it('ignores breakpoints without a lifetime and odd message shapes', () => {
+        expect(
+            messagesHaveCacheBreakpointTtl([
+                {
+                    role: 'system',
+                    content: [
+                        {
+                            type: 'text',
+                            text: 'a',
+                            prompt_cache_breakpoint: { mode: 'explicit' },
+                        },
+                        'bare string part',
+                        null,
+                    ],
+                },
+                { role: 'user', content: null },
+                null,
+            ]),
+        ).toBe(false);
+        expect(messagesHaveCacheBreakpointTtl(undefined)).toBe(false);
+        expect(messagesHaveCacheBreakpointTtl('nope')).toBe(false);
+    });
+});

src/backend/drivers/ai-chat/utils/openaiParams.ts

@@ -29,6 +29,7 @@ import { HttpError } from '../../../core/http/HttpError.js';
 import type {
     ICompleteArguments,
     OutputFormat,
+    PromptCacheOptions,
     ReasoningEffort,
     ToolChoice,
 } from '../types.js';
@@ -305,6 +306,80 @@ export const openAICompatParams = (
     return out;
 };
 
+// -- prompt caching ----------------------------------------------------------
+
+/**
+ * The wire spelling of `prompt_cache_retention`. Puter also accepts the
+ * hyphenated `in-memory`, which every OpenAI-shaped upstream rejects.
+ */
+export const promptCacheRetentionToWire = (
+    retention: ICompleteArguments['prompt_cache_retention'],
+): 'in_memory' | '24h' | undefined =>
+    retention === 'in-memory' ? 'in_memory' : retention;
+
+const PROMPT_CACHE_MODES = new Set(['implicit', 'explicit']);
+const PROMPT_CACHE_TTLS = new Set(['30m', '12h']);
+
+/**
+ * `prompt_cache_options` as forwarded upstream: `mode` and `ttl` only, each
+ * checked against the values OpenAI accepts, so a bad value fails here with a
+ * clear message instead of as an upstream 400 the fallback loop may hide.
+ * `comparison_response_id` names a stored response, which Puter never keeps,
+ * and `prewarm` is dropped with it.
+ */
+export const promptCacheOptionsToWire = (
+    options: unknown,
+): PromptCacheOptions | undefined => {
+    if (options === undefined || options === null) return undefined;
+    if (typeof options !== 'object' || Array.isArray(options)) {
+        throw new HttpError(400, '`prompt_cache_options` must be an object', {
+            legacyCode: 'bad_request',
+        });
+    }
+    const { mode, ttl } = options as Record<string, unknown>;
+    if (mode !== undefined && !PROMPT_CACHE_MODES.has(mode as string)) {
+        throw new HttpError(
+            400,
+            '`prompt_cache_options.mode` must be `implicit` or `explicit`',
+            { legacyCode: 'bad_request' },
+        );
+    }
+    if (ttl !== undefined && !PROMPT_CACHE_TTLS.has(ttl as string)) {
+        throw new HttpError(
+            400,
+            '`prompt_cache_options.ttl` must be `30m` or `12h`',
+            { legacyCode: 'bad_request' },
+        );
+    }
+    const wire: PromptCacheOptions = {
+        ...(mode !== undefined
+            ? { mode: mode as PromptCacheOptions['mode'] }
+            : {}),
+        ...(ttl !== undefined ? { ttl: ttl as PromptCacheOptions['ttl'] } : {}),
+    };
+    return Object.keys(wire).length ? wire : undefined;
+};
+
+/**
+ * Whether any content part carries a `prompt_cache_breakpoint` with its own
+ * `ttl`. Chat Completions rejects the field whatever its value; only the
+ * Responses API honours a per-breakpoint lifetime.
+ */
+export const messagesHaveCacheBreakpointTtl = (messages: unknown): boolean =>
+    Array.isArray(messages) &&
+    messages.some((message) => {
+        const content = (message as { content?: unknown } | null)?.content;
+        const parts = Array.isArray(content) ? content : [content];
+        return parts.some(
+            (part) =>
+                (
+                    part as {
+                        prompt_cache_breakpoint?: { ttl?: unknown } | null;
+                    } | null
+                )?.prompt_cache_breakpoint?.ttl !== undefined,
+        );
+    });
+
 // -- Responses server-side state ---------------------------------------------
 
 const STATEFUL_RESPONSES_FIELDS = [

src/backend/drivers/ai-image/ImageGenerationDriver.test.ts

@@ -758,7 +758,10 @@ describe('ImageGenerationDriver.generate provider routing', () => {
             }),
         );
 
-        const reserveSpy = vi.spyOn(server.services.metering, 'reserveCredits');
+        const reserveSpy = vi.spyOn(
+            server.services.metering,
+            'reserveCreditsIfAffordable',
+        );
         const usageSpy = vi.spyOn(
             server.services.metering,
             'batchIncrementUsages',
@@ -852,7 +855,7 @@ describe('ImageGenerationDriver.generate provider routing', () => {
             });
             const reserveSpy = vi.spyOn(
                 server.services.metering,
-                'reserveCredits',
+                'reserveCreditsIfAffordable',
             );
 
             // Cloudflare answers the scheduled refresh with new rates at once;
@@ -1162,8 +1165,8 @@ describe('ImageGenerationDriver.generate Puter file inputs', () => {
         );
 
     beforeEach(() => {
-        vi.spyOn(server.services.metering, 'hasEnoughCredits').mockResolvedValue(
-            true,
+        vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable').mockImplementation((who, amount, opts) =>
+            server.services.metering.reserveCredits(who, amount, opts),
         );
     });
 
@@ -1665,7 +1668,10 @@ it.each([undefined, null, '', '  ', 42, {}])(
 
 describe('ImageGenerationDriver cost factor', () => {
     it('holds and meters at the AI cost factor', async () => {
-        const reserveSpy = vi.spyOn(server.services.metering, 'reserveCredits');
+        const reserveSpy = vi.spyOn(
+            server.services.metering,
+            'reserveCreditsIfAffordable',
+        );
         const usageSpy = vi.spyOn(
             server.services.metering,
             'batchIncrementUsages',

src/backend/drivers/ai-image/providers/byteplus/BytePlusImageProvider.test.ts

@@ -72,7 +72,7 @@ vi.mock('openai', () => {
 // -- Test harness ----------------------------------------------------
 
 let server: PuterServer;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let batchIncrementUsagesSpy: MockInstance<
     MeteringService['batchIncrementUsages']
 >;
@@ -107,7 +107,7 @@ const makeProvider = (
 beforeEach(() => {
     generateMock.mockReset();
     openAICtor.mockReset();
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     batchIncrementUsagesSpy = vi.spyOn(
         server.services.metering,
         'batchIncrementUsages',
@@ -177,7 +177,7 @@ describe('BytePlusImageProvider.generate gates', () => {
         expect(result).toBe(
             'https://puter-sample-data.puter.site/image_example.png',
         );
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(generateMock).not.toHaveBeenCalled();
     });
 
@@ -196,7 +196,7 @@ describe('BytePlusImageProvider.generate gates', () => {
     });
 
     it('throws 402 BEFORE hitting ModelArk when the actor lacks credits', async () => {
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         await expect(
             withTestActor(() => makeProvider().generate({ prompt: 'hi' })),
         ).rejects.toMatchObject({ statusCode: 402 });
@@ -713,7 +713,7 @@ it.each([
             }),
         ),
     ).rejects.toMatchObject({ statusCode: 400, legacyCode: 'bad_request' });
-    expect(hasCreditsSpy).not.toHaveBeenCalled();
+    expect(admitSpy).not.toHaveBeenCalled();
     expect(generateMock).not.toHaveBeenCalled();
 });
 

src/backend/drivers/ai-image/providers/cloudflare/CloudflareImageProvider.test.ts

@@ -73,7 +73,7 @@ let server: PuterServer;
 let globalFetchSpy: MockInstance<typeof fetch>;
 let fetchSpy: Mock<typeof fetch>;
 let modelSearchFetch: Mock<typeof fetch>;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let batchIncrementUsagesSpy: MockInstance<
     MeteringService['batchIncrementUsages']
 >;
@@ -123,7 +123,7 @@ beforeEach(() => {
                 ? modelSearchFetch(input, init)
                 : fetchSpy(input, init),
         );
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     batchIncrementUsagesSpy = vi.spyOn(
         server.services.metering,
         'batchIncrementUsages',
@@ -188,7 +188,7 @@ describe('CloudflareImageProvider.generate test_mode', () => {
         expect(result).toBe(
             'https://puter-sample-data.puter.site/image_example.png',
         );
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(fetchSpy).not.toHaveBeenCalled();
     });
 });
@@ -210,7 +210,7 @@ describe('CloudflareImageProvider.generate argument validation', () => {
 describe('CloudflareImageProvider.generate credit gate', () => {
     it('throws 402 BEFORE hitting Cloudflare when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
 
         await expect(
             withTestActor(() => provider.generate({ prompt: 'hi' })),
@@ -240,7 +240,7 @@ describe('CloudflareImageProvider.generate credit gate', () => {
             actor,
         );
 
-        const [, estimate] = hasCreditsSpy.mock.calls[0]!;
+        const [, estimate] = admitSpy.mock.calls[0]!;
         expect(estimate).toBeGreaterThan(0);
         expect(heldDuringCall).toBe(Math.ceil(estimate));
         // The hold stays until the usage write lands.
@@ -708,7 +708,7 @@ describe('Cloudflare legacy image option', () => {
                 legacyCode: 'bad_request',
             });
             expect(fetchSpy).not.toHaveBeenCalled();
-            expect(hasCreditsSpy).not.toHaveBeenCalled();
+            expect(admitSpy).not.toHaveBeenCalled();
         },
     );
 
@@ -726,7 +726,7 @@ describe('Cloudflare legacy image option', () => {
             message: 'Invalid input image (empty or not base64)',
         });
         expect(fetchSpy).not.toHaveBeenCalled();
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
     });
 });
 
@@ -823,12 +823,12 @@ describe('Cloudflare priced legacy models', () => {
             imageSize: { kind: 'pixels', w: 512, h: 512 },
         }));
         expect(JSON.parse(fetchSpy.mock.calls[0]![1]!.body as string)).toMatchObject({ image: [1, 2, 3], mask: [4, 5, 6], num_steps: 20, width: 512, height: 512 });
-        expect(hasCreditsSpy).toHaveBeenCalledWith(expect.anything(), 0);
+        expect(admitSpy).toHaveBeenCalledWith(expect.anything(), 0, undefined);
     });
 
     it('rejects missing inpainting inputs before credit checks', async () => {
         await expect(withTestActor(() => makeProvider().generate({ model: '@cf/runwayml/stable-diffusion-v1-5-inpainting', prompt: 'A cup' }))).rejects.toMatchObject({ statusCode: 400 });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(fetchSpy).not.toHaveBeenCalled();
     });
 
@@ -844,7 +844,7 @@ describe('Cloudflare priced legacy models', () => {
     it('charges Lucid at the price returned by the model API', async () => {
         fetchSpy.mockResolvedValueOnce(new Response(new Uint8Array([1, 2, 3]), { headers: { 'content-type': 'image/png' } }));
         await withTestActor(() => makeProvider().generate({ model: '@cf/leonardo/lucid-origin', prompt: 'A cup', imageSize: { kind: 'pixels', w: 512, h: 512 }, steps: 1 }));
-        expect(hasCreditsSpy).toHaveBeenCalledWith(expect.anything(), 713200);
+        expect(admitSpy).toHaveBeenCalledWith(expect.anything(), 713200, undefined);
     });
 });
 
@@ -897,9 +897,10 @@ describe('Cloudflare model-search pricing', () => {
         await runPhoenix(provider);
 
         expect(listed?.costs).toEqual({ tile_512: 1_000_000, step: 20_200 });
-        expect(hasCreditsSpy).toHaveBeenCalledWith(
+        expect(admitSpy).toHaveBeenCalledWith(
             expect.anything(),
             4 * 1_000_000 + 10 * 20_200,
+            undefined,
         );
         expect(batchIncrementUsagesSpy.mock.calls[0]![1]).toEqual([
             expect.objectContaining({
@@ -930,9 +931,10 @@ describe('Cloudflare model-search pricing', () => {
         await runPhoenix(makeProvider());
 
         expect(modelSearchFetch).not.toHaveBeenCalled();
-        expect(hasCreditsSpy).toHaveBeenCalledWith(
+        expect(admitSpy).toHaveBeenCalledWith(
             expect.anything(),
             4 * 583_000 + 10 * 11_000,
+            undefined,
         );
     });
 
@@ -955,7 +957,7 @@ describe('Cloudflare model-search pricing', () => {
         await refresh;
         await runPhoenix(provider);
 
-        expect(hasCreditsSpy.mock.calls.map(([, cost]) => cost)).toEqual([
+        expect(admitSpy.mock.calls.map(([, cost]) => cost)).toEqual([
             2_442_000, 4_202_000,
         ]);
     });
@@ -977,9 +979,10 @@ describe('Cloudflare model-search pricing', () => {
         await runPhoenix(provider);
 
         expect(modelSearchFetch).toHaveBeenCalledTimes(2);
-        expect(hasCreditsSpy).toHaveBeenCalledWith(
+        expect(admitSpy).toHaveBeenCalledWith(
             expect.anything(),
             4_202_000,
+            undefined,
         );
     });
 
@@ -1015,7 +1018,7 @@ describe('Cloudflare model-search pricing', () => {
                     steps: 10,
                 } as never),
             );
-            return hasCreditsSpy.mock.lastCall![1];
+            return admitSpy.mock.lastCall![1];
         };
         const settle = () => new Promise((resolve) => setImmediate(resolve));
         const OLD = { tile_512: 583_000, step: 11_000 };
@@ -1069,9 +1072,10 @@ describe('Cloudflare model-search pricing', () => {
             }),
         );
         expect(modelSearchFetch).toHaveBeenCalledOnce();
-        expect(hasCreditsSpy).toHaveBeenCalledWith(
+        expect(admitSpy).toHaveBeenCalledWith(
             expect.anything(),
             4 * 5280 + 4 * 10560,
+            undefined,
         );
     });
 

src/backend/drivers/ai-image/providers/gemini/GeminiImageProvider.test.ts

@@ -82,7 +82,7 @@ vi.mock('../../inputImage.js', async (orig) => ({
 // ── Test harness ────────────────────────────────────────────────────
 
 let server: PuterServer;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 let batchIncrementUsagesSpy: MockInstance<
     MeteringService['batchIncrementUsages']
@@ -111,7 +111,7 @@ beforeEach(() => {
     generateContentMock.mockReset();
     toBase64DataUriMock.mockReset();
     googleAICtor.mockReset();
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
     batchIncrementUsagesSpy = vi.spyOn(
         server.services.metering,
@@ -167,7 +167,7 @@ describe('GeminiImageProvider.generate test_mode', () => {
         expect(result).toBe(
             'https://puter-sample-data.puter.site/image_example.png',
         );
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(generateContentMock).not.toHaveBeenCalled();
     });
 });
@@ -219,7 +219,7 @@ describe('GeminiImageProvider.generate argument validation', () => {
                 legacyCode: 'bad_request',
                 message: expect.stringContaining('Unsupported quality tier'),
             });
-            expect(hasCreditsSpy).not.toHaveBeenCalled();
+            expect(admitSpy).not.toHaveBeenCalled();
             expect(generateContentMock).not.toHaveBeenCalled();
         },
     );
@@ -427,7 +427,7 @@ describe('GeminiImageProvider.generate Flash path (generateContent)', () => {
 
     it('throws 402 BEFORE hitting Gemini when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
 
         await expect(
             withTestActor(() =>
@@ -466,7 +466,7 @@ describe('GeminiImageProvider.generate Flash path (generateContent)', () => {
             actor,
         );
 
-        const [, estimate] = hasCreditsSpy.mock.calls[0]!;
+        const [, estimate] = admitSpy.mock.calls[0]!;
         expect(estimate).toBeGreaterThan(0);
         expect(heldDuringCall).toBe(Math.ceil(estimate));
         // The hold stays until the usage write lands.
@@ -643,7 +643,7 @@ describe('Gemini Nano Banana image generation', () => {
         const request = generateContentMock.mock.calls[0][0];
         expect(request.config.imageConfig.imageSize).toBe('1K');
         // Input + 1120 image tokens at $30/1M + 1000 text tokens at $7.50/1M.
-        expect(hasCreditsSpy).toHaveBeenCalledWith(expect.anything(), 4_194_450);
+        expect(admitSpy).toHaveBeenCalledWith(expect.anything(), 4_194_450, undefined);
         expect(batchIncrementUsagesSpy).toHaveBeenCalledWith(
             expect.anything(),
             expect.arrayContaining([

src/backend/drivers/ai-image/providers/kling/KlingImageProvider.test.ts

@@ -50,7 +50,7 @@ import { KlingImageProvider } from './KlingImageProvider.js';
 
 let server: PuterServer;
 let fetchSpy: MockInstance<typeof fetch>;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
 beforeAll(async () => {
@@ -73,7 +73,7 @@ const makeProvider = () =>
 
 beforeEach(() => {
     fetchSpy = vi.spyOn(globalThis, 'fetch') as MockInstance<typeof fetch>;
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
     incrementUsageSpy.mockResolvedValue({} as never);
 });
@@ -318,7 +318,7 @@ describe('KlingImageProvider billing', () => {
     });
 
     it('rejects before calling Kling when credits run short', async () => {
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         await expect(generate({})).rejects.toMatchObject({ statusCode: 402 });
         expect(fetchSpy).not.toHaveBeenCalled();
     });

src/backend/drivers/ai-image/providers/openai/OpenAiImageProvider.test.ts

@@ -97,7 +97,7 @@ vi.mock('../../inputImage.js', async (orig) => ({
 // ── Test harness ────────────────────────────────────────────────────
 
 let server: PuterServer;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let batchIncrementUsagesSpy: MockInstance<
     MeteringService['batchIncrementUsages']
 >;
@@ -126,7 +126,7 @@ beforeEach(() => {
     editMock.mockReset();
     fetchImageAsBase64Mock.mockReset();
     openAICtor.mockReset();
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     batchIncrementUsagesSpy = vi.spyOn(
         server.services.metering,
         'batchIncrementUsages',
@@ -195,7 +195,7 @@ describe('OpenAiImageProvider.generate test_mode', () => {
         expect(result).toBe(
             'https://puter-sample-data.puter.site/image_example.png',
         );
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(generateMock).not.toHaveBeenCalled();
     });
 });
@@ -253,7 +253,8 @@ describe('OpenAiImageProvider.generate credit gate', () => {
         });
         let blocked = false;
         vi.spyOn(redis, 'meterIncr').mockImplementation(async (...args) => {
-            if (!blocked && args[5].includes(`:actor:${actor.user!.uuid}:`)) {
+            const actorKey = `:actor:${actor.user!.uuid}:`;
+            if (!blocked && args.some((arg) => arg.includes(actorKey))) {
                 blocked = true;
                 await held;
             }
@@ -287,7 +288,7 @@ describe('OpenAiImageProvider.generate credit gate', () => {
 
     it('throws 402 BEFORE hitting OpenAI when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         await expect(
             withTestActor(() =>
                 provider.generate({
@@ -323,7 +324,7 @@ describe('OpenAiImageProvider.generate credit gate', () => {
             actor,
         );
 
-        const [, estimate] = hasCreditsSpy.mock.calls[0]!;
+        const [, estimate] = admitSpy.mock.calls[0]!;
         expect(estimate).toBeGreaterThan(0);
         expect(heldDuringCall).toBe(Math.ceil(estimate));
         // The hold stays until the usage write lands.
@@ -727,8 +728,8 @@ describe('OpenAiImageProvider gpt-image-2.5 models', () => {
         );
 
         // Latent factor 48 vs 24 at the medium tier.
-        const gptImage2Estimate = hasCreditsSpy.mock.calls[0]![1];
-        const gptImage25Estimate = hasCreditsSpy.mock.calls[1]![1];
+        const gptImage2Estimate = admitSpy.mock.calls[0]![1];
+        const gptImage25Estimate = admitSpy.mock.calls[1]![1];
         expect(gptImage25Estimate).toBeLessThan(gptImage2Estimate);
     });
 
@@ -753,8 +754,8 @@ describe('OpenAiImageProvider gpt-image-2.5 models', () => {
             }),
         );
 
-        const highEstimate = hasCreditsSpy.mock.calls[0]![1];
-        const maxEstimate = hasCreditsSpy.mock.calls[1]![1];
+        const highEstimate = admitSpy.mock.calls[0]![1];
+        const maxEstimate = admitSpy.mock.calls[1]![1];
         expect(maxEstimate).toBeGreaterThan(highEstimate);
     });
 });

src/backend/drivers/ai-image/providers/replicate/ReplicateImageGenerationProvider.test.ts

@@ -84,7 +84,7 @@ vi.mock('../../../../util/secureHttp.js', () => ({
 // ── Test harness ────────────────────────────────────────────────────
 
 let server: PuterServer;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 let batchIncrementUsagesSpy: MockInstance<
     MeteringService['batchIncrementUsages']
@@ -115,7 +115,7 @@ beforeEach(() => {
     cancelMock.mockReset();
     replicateCtor.mockReset();
     secureFetchMock.mockReset();
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
     batchIncrementUsagesSpy = vi.spyOn(
         server.services.metering,
@@ -177,7 +177,7 @@ describe('ReplicateImageGenerationProvider.generate test_mode', () => {
         expect(result).toBe(
             'https://puter-sample-data.puter.site/image_example.png',
         );
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(createPredictionMock).not.toHaveBeenCalled();
     });
 });
@@ -199,7 +199,7 @@ describe('ReplicateImageGenerationProvider.generate argument validation', () =>
 describe('ReplicateImageGenerationProvider.generate credit gate', () => {
     it('throws 402 BEFORE hitting Replicate when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         await expect(
             withTestActor(() => provider.generate({ prompt: 'hi' })),
         ).rejects.toMatchObject({ statusCode: 402 });
@@ -255,7 +255,10 @@ describe('ReplicateImageGenerationProvider.generate credit gate', () => {
     it('holds credits for longer than a prediction can run', async () => {
         const provider = makeProvider();
         const actor = holdUser('ttl');
-        const reserveSpy = vi.spyOn(server.services.metering, 'reserveCredits');
+        const reserveSpy = vi.spyOn(
+            server.services.metering,
+            'reserveCreditsIfAffordable',
+        );
         createPredictionMock.mockResolvedValueOnce({
             id: 'ttl',
             status: 'succeeded',
@@ -274,7 +277,9 @@ describe('ReplicateImageGenerationProvider.generate credit gate', () => {
         const provider = makeProvider();
         const actor = holdUser('surcharge');
         secureFetchMock.mockResolvedValueOnce(new Response('not-an-image'));
-        hasCreditsSpy.mockResolvedValueOnce(true).mockResolvedValueOnce(false);
+        admitSpy.mockImplementationOnce((who, amount, opts) =>
+            server.services.metering.reserveCredits(who, amount, opts),
+        ).mockResolvedValueOnce(false);
 
         await expect(
             withTestActor(
@@ -950,7 +955,7 @@ it('does not start a prediction for an already-aborted request', async () => {
         ),
     ).rejects.toMatchObject({ statusCode: 400, legacyCode: 'client_aborted' });
     expect(createPredictionMock).not.toHaveBeenCalled();
-    expect(hasCreditsSpy).not.toHaveBeenCalled();
+    expect(admitSpy).not.toHaveBeenCalled();
 });
 
 it('stops polling and cancels at the prediction deadline', async () => {
@@ -1192,12 +1197,12 @@ it('rejects more than ten input images before fetching or pricing any', async ()
         ),
     ).rejects.toMatchObject({ statusCode: 400, legacyCode: 'bad_request' });
     expect(secureFetchMock).not.toHaveBeenCalled();
-    expect(hasCreditsSpy).not.toHaveBeenCalled();
+    expect(admitSpy).not.toHaveBeenCalled();
     expect(createPredictionMock).not.toHaveBeenCalled();
 });
 
 it('checks credits on the output-only estimate before fetching input images', async () => {
-    hasCreditsSpy.mockResolvedValueOnce(false);
+    admitSpy.mockResolvedValueOnce(null);
     await expect(
         withTestActor(() =>
             makeProvider().generate({

src/backend/drivers/ai-image/providers/together/TogetherImageProvider.test.ts

@@ -71,7 +71,7 @@ vi.mock('together-ai', () => {
 // ── Test harness ────────────────────────────────────────────────────
 
 let server: PuterServer;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
 beforeAll(async () => {
@@ -96,7 +96,7 @@ const makeProvider = () =>
 beforeEach(() => {
     generateMock.mockReset();
     togetherCtor.mockReset();
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
 });
 
@@ -158,7 +158,7 @@ describe('TogetherImageProvider.generate test_mode', () => {
         expect(result).toBe(
             'https://puter-sample-data.puter.site/image_example.png',
         );
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(generateMock).not.toHaveBeenCalled();
     });
 });
@@ -189,7 +189,7 @@ describe('TogetherImageProvider.generate argument validation', () => {
 describe('TogetherImageProvider.generate credit gate', () => {
     it('throws 402 BEFORE hitting Together when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
 
         await expect(
             withTestActor(() =>
@@ -760,7 +760,7 @@ describe('Together supported sizes and references', () => {
             width: 1024,
             height: 1024,
         });
-        expect(Number.isFinite(hasCreditsSpy.mock.calls[0][1])).toBe(true);
+        expect(Number.isFinite(admitSpy.mock.calls[0][1])).toBe(true);
     });
 
     it('resolves provider aliases without changing the selected model', async () => {
@@ -790,7 +790,7 @@ describe('Together supported sizes and references', () => {
                 }),
             ),
         ).rejects.toMatchObject({ statusCode: 400 });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(generateMock).not.toHaveBeenCalled();
     });
 
@@ -859,7 +859,7 @@ it.each(['3K', 'unknown', ' 5k '])(
                 }),
             ),
         ).rejects.toMatchObject({ statusCode: 400, legacyCode: 'bad_request' });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(generateMock).not.toHaveBeenCalled();
     },
 );
@@ -895,7 +895,7 @@ describe('Together current API prices', () => {
                 input_image: 'https://example.com/reference.png',
             }),
         );
-        expect(hasCreditsSpy).toHaveBeenCalledWith(expect.anything(), 4685000);
+        expect(admitSpy).toHaveBeenCalledWith(expect.anything(), 4685000, undefined);
         expect(incrementUsageSpy).toHaveBeenCalledWith(
             expect.anything(),
             'togetherai:google/flash-image-3.1:0.5K',

src/backend/drivers/ai-image/providers/xai/XAIImageProvider.test.ts

@@ -85,7 +85,7 @@ vi.mock('openai', () => {
 // ── Test harness ────────────────────────────────────────────────────
 
 let server: PuterServer;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let batchIncrementUsagesSpy: MockInstance<
     MeteringService['batchIncrementUsages']
 >;
@@ -115,7 +115,7 @@ beforeEach(() => {
     openAICtor.mockReset();
     axiosGetMock.mockReset();
     axiosGetMock.mockRejectedValue(new Error('Catalog unavailable'));
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     batchIncrementUsagesSpy = vi.spyOn(
         server.services.metering,
         'batchIncrementUsages',
@@ -182,7 +182,7 @@ describe('XAIImageProvider.generate test_mode', () => {
         expect(result).toBe(
             'https://puter-sample-data.puter.site/image_example.png',
         );
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(generateMock).not.toHaveBeenCalled();
         expect(batchIncrementUsagesSpy).not.toHaveBeenCalled();
     });
@@ -231,7 +231,7 @@ describe('XAIImageProvider.generate argument validation', () => {
 describe('XAIImageProvider.generate credit gate', () => {
     it('throws 402 BEFORE hitting xAI when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
 
         await expect(
             withTestActor(() =>
@@ -718,7 +718,7 @@ describe('XAIImageProvider quality and reference limits', () => {
                 }),
             ),
         ).rejects.toMatchObject({ statusCode: 400 });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(postMock).not.toHaveBeenCalled();
     });
 });
@@ -746,7 +746,7 @@ describe('xAI option validation', () => {
                 makeProvider().generate({ prompt: 'hi', ...options } as never),
             ),
         ).rejects.toMatchObject({ statusCode: 400 });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(generateMock).not.toHaveBeenCalled();
     });
 });

src/backend/drivers/ai-ocr/OCRDriver.test.ts

@@ -167,7 +167,7 @@ const sentRegions = () =>
 
 let server: PuterServer;
 let driver: OCRDriver;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
 beforeAll(async () => {
@@ -199,7 +199,7 @@ beforeEach(() => {
     mistralCtor.mockReset();
     // Spy on metering — keep the real impl so its recording side runs,
     // but capture calls so per-test assertions can inspect them.
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
 });
 
@@ -335,7 +335,7 @@ describe('OCRDriver.recognize (aws-textract)', () => {
     };
 
     it('throws 402 when the actor does not have enough credits', async () => {
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         const { actor } = await makeUser();
 
         await expect(
@@ -546,7 +546,7 @@ describe('OCRDriver.recognize (aws-textract)', () => {
     describe('regional pricing', () => {
         const onePage = { Blocks: [{ BlockType: 'PAGE' }] };
         const checkedCosts = () =>
-            hasCreditsSpy.mock.calls.map(([, cost]) => cost);
+            admitSpy.mock.calls.map(([, cost]) => cost);
         const textractCharges = () =>
             incrementUsageSpy.mock.calls
                 .filter(([, type]) => type === TEXTRACT_USAGE)
@@ -719,7 +719,7 @@ describe('OCRDriver.recognize (mistral)', () => {
     };
 
     it('throws 402 when the actor does not have enough credits', async () => {
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         const { actor } = await makeUser();
 
         await expect(
@@ -1033,7 +1033,7 @@ describe('OCRDriver.recognize (mistral)', () => {
         const pageType = 'mistral-ocr:mistral-ocr-4-1:page';
         const annotationType = 'mistral-ocr:mistral-ocr-4-1:annotations:page';
         // The pre-flight covers the page plus its annotation.
-        expect(hasCreditsSpy.mock.calls[0]![1]).toBe(
+        expect(admitSpy.mock.calls[0]![1]).toBe(
             OCR_COSTS[pageType] + OCR_COSTS[annotationType],
         );
         const ocrCalls = incrementUsageSpy.mock.calls.filter(
@@ -1080,7 +1080,7 @@ describe('OCRDriver credit pre-flight', () => {
     const pageType = 'mistral-ocr:mistral-ocr-4-1:page';
     const perPage = OCR_COSTS[pageType];
     const pdfSource = (buffer: Buffer) => dataUrl(buffer, 'application/pdf');
-    const checkedCosts = () => hasCreditsSpy.mock.calls.map(([, cost]) => cost);
+    const checkedCosts = () => admitSpy.mock.calls.map(([, cost]) => cost);
     const outstandingHolds = (actor: Actor) =>
         server.stores.creditHold.outstanding(actor.user.uuid!);
 
@@ -1171,8 +1171,8 @@ describe('OCRDriver credit pre-flight', () => {
             ),
         ).rejects.toMatchObject({ statusCode: 402 });
 
-        // The estimate/hold path (which calls hasEnoughCredits) never ran.
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        // The estimate/hold path never ran.
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(mistralOcrProcessMock).not.toHaveBeenCalled();
         noUsage.mockRestore();
     });

src/backend/drivers/ai-speech2speech/VoiceChangerDriver.test.ts

@@ -72,7 +72,7 @@ vi.mock('undici', async (importOriginal) => ({
 let server: PuterServer;
 let driver: VoiceChangerDriver;
 let fetchSpy: MockInstance<typeof fetch>;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
 beforeAll(async () => {
@@ -91,7 +91,7 @@ afterAll(async () => {
 beforeEach(() => {
     webFetchMock.mockReset();
     fetchSpy = vi.spyOn(globalThis, 'fetch') as MockInstance<typeof fetch>;
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
 });
 
@@ -197,7 +197,7 @@ describe('VoiceChangerDriver.convert argument validation', () => {
 describe('VoiceChangerDriver.convert credit gate', () => {
     it('throws 402 BEFORE hitting ElevenLabs when the actor lacks credits', async () => {
         const { actor } = await makeUser();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
 
         await expect(
             withActor(actor, () =>
@@ -387,7 +387,7 @@ describe('VoiceChangerDriver.convert success path', () => {
 
         const [base, factored] = metered();
         expect(factored).toBe(base! * 2);
-        const held = hasCreditsSpy.mock.calls.map(([, amount]) => amount);
+        const held = admitSpy.mock.calls.map(([, amount]) => amount);
         expect(held).toEqual([base, base! * 2]);
     });
 

src/backend/drivers/ai-speech2txt/SpeechToTextDriver.test.ts

@@ -122,7 +122,7 @@ vi.mock('openai', () => {
 
 let server: PuterServer;
 let driver: SpeechToTextDriver;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
 beforeAll(async () => {
@@ -145,7 +145,7 @@ beforeEach(() => {
     toFileMock.mockClear();
     // Spy on metering — keep the real impl so its recording side runs,
     // but capture calls so per-test assertions can inspect them.
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
 });
 
@@ -360,7 +360,7 @@ describe('SpeechToTextDriver argument validation', () => {
 
 describe('SpeechToTextDriver credit gate', () => {
     it('throws 402 BEFORE hitting OpenAI when actor lacks credits', async () => {
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         const { actor } = await makeUser();
 
         await expect(
@@ -686,7 +686,7 @@ describe('SpeechToTextDriver metering', () => {
         expect(cost).toBe(SPEECH_TO_TEXT_COSTS[usageType]);
     });
 
-    it('asks hasEnoughCredits for the same total it later meters', async () => {
+    it('reserves the same total it later meters', async () => {
         const { actor } = await makeUser();
         transcriptionsCreateMock.mockResolvedValueOnce({ text: 'x' });
 
@@ -700,7 +700,7 @@ describe('SpeechToTextDriver metering', () => {
 
         const usageType = 'openai:gpt-transcribe:second';
         const expected = SPEECH_TO_TEXT_COSTS[usageType] * 2;
-        const creditCall = hasCreditsSpy.mock.calls[0]!;
+        const creditCall = admitSpy.mock.calls[0]!;
         expect(creditCall[1]).toBe(expected);
     });
 
@@ -794,7 +794,7 @@ describe('SpeechToTextDriver metering', () => {
 
         const [base, factored] = metered();
         expect(factored).toBe(base! * 2);
-        const held = hasCreditsSpy.mock.calls.map(([, amount]) => amount);
+        const held = admitSpy.mock.calls.map(([, amount]) => amount);
         expect(held).toEqual([base, base! * 2]);
     });
 });

src/backend/drivers/ai-speech2txt/providers/gemini/GeminiSpeechToTextProvider.test.ts

@@ -72,7 +72,7 @@ const HOLD = COUNTED_TOKENS * AUDIO_RATE + 8192 * OUTPUT_RATE;
 let server: PuterServer;
 let sttDriver: SpeechToTextDriver;
 let fetchSpy: MockInstance<typeof fetch>;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let batchSpy: MockInstance<MeteringService['batchIncrementUsages']>;
 
 const driver = {
@@ -102,7 +102,7 @@ beforeEach(() => {
         .mockRejectedValue(new Error('unexpected fetch')) as MockInstance<
         typeof fetch
     >;
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     batchSpy = vi.spyOn(server.services.metering, 'batchIncrementUsages');
 });
 
@@ -220,13 +220,16 @@ const metered = (): UsageInput[] =>
 /** Records each credit hold taken, its options, and its release. */
 const trackHolds = (order: string[] = []) => {
     const metering = server.services.metering;
-    const realReserve = metering.reserveCredits.bind(metering);
+    const realReserve = Object.getPrototypeOf(
+        metering,
+    ).reserveCreditsIfAffordable.bind(metering) as typeof metering.reserveCreditsIfAffordable;
     const held: { amount: number; ttlMs?: number }[] = [];
     const released = vi.fn();
-    vi.spyOn(metering, 'reserveCredits').mockImplementation(
+    vi.spyOn(metering, 'reserveCreditsIfAffordable').mockImplementation(
         async (actor, amount, opts) => {
             held.push({ amount, ttlMs: opts?.ttlMs });
             const hold = await realReserve(actor, amount, opts);
+            if (!hold) return null;
             return {
                 ...hold,
                 release: async () => {
@@ -425,7 +428,7 @@ describe('GeminiSpeechToTextProvider test_mode', () => {
             model: MODEL,
         });
         expect(fetchSpy).not.toHaveBeenCalled();
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(batchSpy).not.toHaveBeenCalled();
     });
 
@@ -1263,12 +1266,12 @@ describe('GeminiSpeechToTextProvider countTokens preflight', () => {
 describe('GeminiSpeechToTextProvider credit hold', () => {
     it('throws 402 after counting tokens, before transcribing', async () => {
         const { actor } = await makeUser();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         fetchSpy.mockResolvedValueOnce(Response.json({ totalTokens: 91 }));
         await expect(
             withActor(actor, () => driver.transcribe({ file: dataUrl() })),
         ).rejects.toMatchObject({ statusCode: 402 });
-        expect(hasCreditsSpy.mock.calls[0]![1]).toBe(HOLD);
+        expect(admitSpy.mock.calls[0]![1]).toBe(HOLD);
         expect(fetchSpy).toHaveBeenCalledTimes(1);
         expect(batchSpy).not.toHaveBeenCalled();
     });

src/backend/drivers/ai-speech2txt/providers/xai/XAISpeechToTextProvider.test.ts

@@ -70,7 +70,7 @@ const driver = {
     getReportedCosts: () => sttDriver.getReportedCosts(),
 };
 let fetchSpy: MockInstance<typeof fetch>;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
 beforeAll(async () => {
@@ -88,7 +88,7 @@ afterAll(async () => {
 
 beforeEach(() => {
     fetchSpy = vi.spyOn(globalThis, 'fetch') as MockInstance<typeof fetch>;
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
 });
 
@@ -193,7 +193,7 @@ describe('XAISpeechToTextProvider test_mode', () => {
         expect(Array.isArray(result.words)).toBe(true);
         // No file required, no actor required, no fetch / metering hit.
         expect(fetchSpy).not.toHaveBeenCalled();
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(incrementUsageSpy).not.toHaveBeenCalled();
     });
 
@@ -258,7 +258,7 @@ describe('XAISpeechToTextProvider missing API key', () => {
 
 describe('XAISpeechToTextProvider credit gate', () => {
     it('throws 402 BEFORE hitting xAI when actor lacks credits', async () => {
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         const { actor } = await makeUser();
 
         await expect(
@@ -273,7 +273,7 @@ describe('XAISpeechToTextProvider credit gate', () => {
         expect(incrementUsageSpy).not.toHaveBeenCalled();
     });
 
-    it('asks hasEnoughCredits for estimated seconds × per-second ucents', async () => {
+    it('reserves estimated seconds × per-second ucents', async () => {
         const { actor } = await makeUser();
         fetchSpy.mockResolvedValueOnce(
             sttResponse({ text: 'ok', duration: 1 }),
@@ -285,7 +285,7 @@ describe('XAISpeechToTextProvider credit gate', () => {
             driver.transcribe({ file: dataUrl(audio, 'audio/mp3') }),
         );
 
-        expect(hasCreditsSpy.mock.calls[0]![1]).toBe(UCENTS_PER_SECOND * 2);
+        expect(admitSpy.mock.calls[0]![1]).toBe(UCENTS_PER_SECOND * 2);
     });
 
     it('estimates 60 seconds for URL inputs (we cannot inspect size locally)', async () => {
@@ -298,7 +298,7 @@ describe('XAISpeechToTextProvider credit gate', () => {
             driver.transcribe({ file: 'https://example.com/clip.mp3' }),
         );
 
-        expect(hasCreditsSpy.mock.calls[0]![1]).toBe(UCENTS_PER_SECOND * 60);
+        expect(admitSpy.mock.calls[0]![1]).toBe(UCENTS_PER_SECOND * 60);
     });
 });
 

src/backend/drivers/ai-tts/TTSDriver.test.ts

@@ -699,9 +699,9 @@ describe('TTSDriver cost factor', () => {
             server.services.metering,
             'incrementUsage',
         );
-        const hasCreditsSpy = vi.spyOn(
+        const admitSpy = vi.spyOn(
             server.services.metering,
-            'hasEnoughCredits',
+            'reserveCreditsIfAffordable',
         );
         const metered = () =>
             incrementUsageSpy.mock.calls.map(([, , , cost]) => cost!);
@@ -725,7 +725,7 @@ describe('TTSDriver cost factor', () => {
         const [base, factored] = metered();
         expect(base).toBeGreaterThan(0);
         expect(factored).toBe(Math.round(base! * 2));
-        const held = hasCreditsSpy.mock.calls.map(([, amount]) => amount);
+        const held = admitSpy.mock.calls.map(([, amount]) => amount);
         expect(held).toEqual([base, Math.round(base! * 2)]);
     });
 });

src/backend/drivers/ai-tts/providers/awsPolly/AWSPollyTTSProvider.test.ts

@@ -84,7 +84,7 @@ vi.mock('@aws-sdk/client-polly', async () => {
 
 let server: PuterServer;
 let priceFetch: MockInstance<typeof fetch>;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
 beforeAll(async () => {
@@ -180,7 +180,7 @@ beforeEach(() => {
     pollyCtor.mockReset();
     priceFetch = vi.fn(servePriceLists());
     vi.stubGlobal('fetch', priceFetch);
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
 });
 
@@ -314,7 +314,7 @@ describe('AWSPollyTTSProvider.synthesize test_mode', () => {
             url: 'https://puter-sample-data.puter.site/tts_example.mp3',
             content_type: 'audio',
         });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(pollySendMock).not.toHaveBeenCalled();
     });
 });
@@ -349,7 +349,7 @@ describe('AWSPollyTTSProvider.synthesize argument validation', () => {
 describe('AWSPollyTTSProvider.synthesize credit gate', () => {
     it('throws 402 BEFORE hitting Polly when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         sendDispatch();
 
         await expect(
@@ -481,7 +481,7 @@ describe('AWSPollyTTSProvider.synthesize metering', () => {
         expect(cost).toBe(expectedCost);
     });
 
-    it('asks for hasEnoughCredits with the same total it later meters', async () => {
+    it('reserves the same total it later meters', async () => {
         const provider = makeProvider();
         sendDispatch();
 
@@ -495,7 +495,7 @@ describe('AWSPollyTTSProvider.synthesize metering', () => {
         );
 
         const expectedCost = AWS_POLLY_COSTS['generative'] * text.length;
-        expect(hasCreditsSpy.mock.calls[0]![1]).toBe(expectedCost);
+        expect(admitSpy.mock.calls[0]![1]).toBe(expectedCost);
     });
 });
 
@@ -599,7 +599,7 @@ describe('AWSPollyTTSProvider published pricing', () => {
             engine: 'generative',
         });
 
-        expect(hasCreditsSpy.mock.calls.map((c) => c[1])).toEqual([
+        expect(admitSpy.mock.calls.map((c) => c[1])).toEqual([
             1700 * 5,
             3000 * 5,
         ]);
@@ -648,7 +648,7 @@ describe('AWSPollyTTSProvider published pricing', () => {
         await readStreamResult(result);
 
         expect(refreshedNeural).toBe(19);
-        expect(hasCreditsSpy.mock.calls[0]![1]).toBe(1700 * 5);
+        expect(admitSpy.mock.calls[0]![1]).toBe(1700 * 5);
         expect(incrementUsageSpy.mock.calls[0]![3]).toBe(1700 * 5);
     });
 

src/backend/drivers/ai-tts/providers/common.test.ts

@@ -86,8 +86,10 @@ describe('meterPerCharacter', () => {
 
     it('holds and then meters exactly text length × price', async () => {
         const check = vi
-            .spyOn(server.services.metering, 'hasEnoughCredits')
-            .mockResolvedValue(true);
+            .spyOn(server.services.metering, 'reserveCreditsIfAffordable')
+            .mockImplementation((who, amount, opts) =>
+                server.services.metering.reserveCredits(who, amount, opts),
+            );
         const record = vi.spyOn(server.services.metering, 'incrementUsage');
 
         await drain(
@@ -101,7 +103,7 @@ describe('meterPerCharacter', () => {
             ),
         );
 
-        expect(check).toHaveBeenCalledWith(actor, 7500);
+        expect(check).toHaveBeenCalledWith(actor, 7500, undefined);
         expect(record).toHaveBeenCalledWith(
             actor,
             'openai:tts-1:character',
@@ -118,8 +120,10 @@ describe('meterPerCharacter', () => {
     it('meters nothing when the upstream breaks before any audio', async () => {
         vi.spyOn(
             server.services.metering,
-            'hasEnoughCredits',
-        ).mockResolvedValue(true);
+            'reserveCreditsIfAffordable',
+        ).mockImplementation((who, amount, opts) =>
+            server.services.metering.reserveCredits(who, amount, opts),
+        );
         const record = vi.spyOn(server.services.metering, 'incrementUsage');
 
         await drain(
@@ -139,8 +143,10 @@ describe('meterPerCharacter', () => {
     it('meters the full text once audio has started, even if the upstream then breaks', async () => {
         vi.spyOn(
             server.services.metering,
-            'hasEnoughCredits',
-        ).mockResolvedValue(true);
+            'reserveCreditsIfAffordable',
+        ).mockImplementation((who, amount, opts) =>
+            server.services.metering.reserveCredits(who, amount, opts),
+        );
         const record = vi.spyOn(server.services.metering, 'incrementUsage');
 
         await drain(
@@ -165,8 +171,10 @@ describe('meterPerCharacter', () => {
     it('meters the full text when the caller hangs up before any audio', async () => {
         vi.spyOn(
             server.services.metering,
-            'hasEnoughCredits',
-        ).mockResolvedValue(true);
+            'reserveCreditsIfAffordable',
+        ).mockImplementation((who, amount, opts) =>
+            server.services.metering.reserveCredits(who, amount, opts),
+        );
         const record = vi.spyOn(server.services.metering, 'incrementUsage');
 
         const result = await meterPerCharacter(
@@ -192,8 +200,10 @@ describe('meterPerCharacter', () => {
     it('meters nothing when synthesis fails', async () => {
         vi.spyOn(
             server.services.metering,
-            'hasEnoughCredits',
-        ).mockResolvedValue(true);
+            'reserveCreditsIfAffordable',
+        ).mockImplementation((who, amount, opts) =>
+            server.services.metering.reserveCredits(who, amount, opts),
+        );
         const record = vi.spyOn(server.services.metering, 'incrementUsage');
         const boom = new Error('vendor down');
 
@@ -213,8 +223,8 @@ describe('meterPerCharacter', () => {
     it('refuses with a 402 before synthesizing when unaffordable', async () => {
         vi.spyOn(
             server.services.metering,
-            'hasEnoughCredits',
-        ).mockResolvedValue(false);
+            'reserveCreditsIfAffordable',
+        ).mockResolvedValue(null);
         const synthesize = vi.fn(audio);
 
         await expect(

src/backend/drivers/ai-tts/providers/elevenlabs/ElevenLabsTTSProvider.test.ts

@@ -57,7 +57,7 @@ import { ELEVENLABS_TTS_COSTS } from './costs.js';
 
 let server: PuterServer;
 let fetchSpy: MockInstance<typeof fetch>;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
 beforeAll(async () => {
@@ -88,7 +88,7 @@ const audioResponse = (body = 'audio-bytes', contentType = 'audio/mpeg') =>
 
 beforeEach(() => {
     fetchSpy = vi.spyOn(globalThis, 'fetch') as MockInstance<typeof fetch>;
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
 });
 
@@ -283,7 +283,7 @@ describe('ElevenLabsTTSProvider.synthesize test_mode', () => {
             url: 'https://puter-sample-data.puter.site/tts_example.mp3',
             content_type: 'audio',
         });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(fetchSpy).not.toHaveBeenCalled();
     });
 });
@@ -323,7 +323,7 @@ describe('ElevenLabsTTSProvider.synthesize argument validation', () => {
         });
 
         expect(fetchSpy).not.toHaveBeenCalled();
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(incrementUsageSpy).not.toHaveBeenCalled();
     });
 
@@ -358,7 +358,7 @@ describe('ElevenLabsTTSProvider.synthesize argument validation', () => {
 describe('ElevenLabsTTSProvider.synthesize credit gate', () => {
     it('throws 402 BEFORE hitting ElevenLabs when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
 
         await expect(
             withTestActor(() => provider.synthesize({ text: 'hi' })),
@@ -510,7 +510,7 @@ describe('ElevenLabsTTSProvider.synthesize request shape', () => {
                 fields: { key, got: value },
             });
             expect(fetchSpy).not.toHaveBeenCalled();
-            expect(hasCreditsSpy).not.toHaveBeenCalled();
+            expect(admitSpy).not.toHaveBeenCalled();
             expect(incrementUsageSpy).not.toHaveBeenCalled();
         },
     );
@@ -893,7 +893,7 @@ describe('ElevenLabsTTSProvider.synthesize metering', () => {
         expect(cost).toBe(expectedCost);
     });
 
-    it('asks for hasEnoughCredits with the same total it later meters', async () => {
+    it('reserves the same total it later meters', async () => {
         const provider = makeProvider();
         fetchSpy.mockResolvedValueOnce(audioResponse());
 
@@ -904,7 +904,7 @@ describe('ElevenLabsTTSProvider.synthesize metering', () => {
 
         const expectedCost =
             ELEVENLABS_TTS_COSTS['eleven_multilingual_v2'] * text.length;
-        expect(hasCreditsSpy.mock.calls[0]![1]).toBe(expectedCost);
+        expect(admitSpy.mock.calls[0]![1]).toBe(expectedCost);
     });
 });
 

src/backend/drivers/ai-tts/providers/gemini/GeminiTTSProvider.test.ts

@@ -61,7 +61,7 @@ const INTERACTIONS_URL =
 
 let server: PuterServer;
 let fetchSpy: MockInstance<typeof fetch>;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let batchIncrementUsagesSpy: MockInstance<
     MeteringService['batchIncrementUsages']
 >;
@@ -183,13 +183,16 @@ const payingUser = () =>
 /** Spies on the credit holds taken, recording the amount and each release. */
 const trackHolds = () => {
     const metering = server.services.metering;
-    const realReserve = metering.reserveCredits.bind(metering);
+    const realReserve = Object.getPrototypeOf(
+        metering,
+    ).reserveCreditsIfAffordable.bind(metering) as typeof metering.reserveCreditsIfAffordable;
     const held: number[] = [];
     const released = vi.fn();
-    vi.spyOn(metering, 'reserveCredits').mockImplementation(
+    vi.spyOn(metering, 'reserveCreditsIfAffordable').mockImplementation(
         async (actor, amount, opts) => {
             held.push(amount);
             const hold = await realReserve(actor, amount, opts);
+            if (!hold) return null;
             return {
                 ...hold,
                 release: async () => {
@@ -214,7 +217,7 @@ beforeEach(() => {
         .mockRejectedValue(new Error('unexpected fetch')) as MockInstance<
         typeof fetch
     >;
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     batchIncrementUsagesSpy = vi.spyOn(
         server.services.metering,
         'batchIncrementUsages',
@@ -332,7 +335,7 @@ describe('GeminiTTSProvider.synthesize test_mode', () => {
             url: 'https://puter-sample-data.puter.site/tts_example.mp3',
             content_type: 'audio',
         });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(fetchSpy).not.toHaveBeenCalled();
     });
 });
@@ -406,7 +409,7 @@ describe('GeminiTTSProvider.synthesize argument validation', () => {
             legacyCode: 'input_too_large',
             fields: { key: 'text', max: 3000 },
         });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(fetchSpy).not.toHaveBeenCalled();
     });
 
@@ -434,7 +437,7 @@ describe('GeminiTTSProvider.synthesize argument validation', () => {
             legacyCode: 'input_too_large',
             fields: { key: 'instructions', max: 3000 },
         });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(fetchSpy).not.toHaveBeenCalled();
     });
 });
@@ -443,7 +446,7 @@ describe('GeminiTTSProvider.synthesize argument validation', () => {
 
 describe('GeminiTTSProvider.synthesize credit gate', () => {
     it('throws 402 before calling Gemini when the actor lacks credits', async () => {
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         await expect(
             withTestActor(() => makeProvider().synthesize({ text: 'hi' })),
         ).rejects.toMatchObject({ statusCode: 402 });
@@ -473,9 +476,10 @@ describe('GeminiTTSProvider.synthesize credit gate', () => {
                         actor,
                     ),
                 ).rejects.toMatchObject({ statusCode: 402 });
-                expect(hasCreditsSpy).toHaveBeenCalledWith(
+                expect(admitSpy).toHaveBeenCalledWith(
                     actor,
                     FLASH_RESERVATION,
+                    undefined,
                 );
                 expect(fetchSpy).not.toHaveBeenCalled();
                 expect(batchIncrementUsagesSpy).not.toHaveBeenCalled();
@@ -1365,9 +1369,10 @@ describe('GeminiTTSProvider.synthesize scheduled rates', () => {
                 server.clients.event.off('ai.cost.factor.*', factor);
             }
 
-            expect(hasCreditsSpy).toHaveBeenCalledWith(
+            expect(admitSpy).toHaveBeenCalledWith(
                 expect.anything(),
                 reserved * 1.5,
+                undefined,
             );
             expect(held).toEqual([reserved * 1.5]);
             expect(metered(':input').costOverride).toBe(7 * inputRate * 1.5);

src/backend/drivers/ai-tts/providers/openai/OpenAITTSProvider.test.ts

@@ -84,7 +84,7 @@ vi.mock('openai', () => {
 // ── Test harness ────────────────────────────────────────────────────
 
 let server: PuterServer;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
 beforeAll(async () => {
@@ -108,7 +108,7 @@ const mockAudioResponse = (bytes = 'opus-audio') => new Response(bytes);
 beforeEach(() => {
     speechCreateMock.mockReset();
     openAICtor.mockReset();
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
 });
 
@@ -205,7 +205,7 @@ describe('OpenAITTSProvider.synthesize test_mode', () => {
             url: 'https://puter-sample-data.puter.site/tts_example.mp3',
             content_type: 'audio',
         });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(speechCreateMock).not.toHaveBeenCalled();
     });
 });
@@ -283,7 +283,7 @@ describe('OpenAITTSProvider.synthesize argument validation', () => {
 describe('OpenAITTSProvider.synthesize credit gate', () => {
     it('throws 402 BEFORE hitting OpenAI when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
 
         await expect(
             withTestActor(() => provider.synthesize({ text: 'hi' })),
@@ -298,7 +298,9 @@ describe('OpenAITTSProvider.synthesize credit gate', () => {
             user: { id: 7, uuid: `tts-hold-${Date.now()}`, username: 'hold' },
         });
         const userId = actor.user!.uuid;
-        hasCreditsSpy.mockResolvedValue(true);
+        admitSpy.mockImplementation((who, amount, opts) =>
+            server.services.metering.reserveCredits(who, amount, opts),
+        );
         let heldDuringCall = -1;
         speechCreateMock.mockImplementationOnce(async () => {
             heldDuringCall = await server.stores.creditHold.outstanding(userId);
@@ -324,7 +326,9 @@ describe('OpenAITTSProvider.synthesize credit gate', () => {
         const actor = makeActor({
             user: { id: 8, uuid: `tts-fail-${Date.now()}`, username: 'fail' },
         });
-        hasCreditsSpy.mockResolvedValue(true);
+        admitSpy.mockImplementation((who, amount, opts) =>
+            server.services.metering.reserveCredits(who, amount, opts),
+        );
         speechCreateMock.mockRejectedValueOnce(new Error('upstream down'));
 
         await expect(
@@ -463,7 +467,7 @@ describe('OpenAITTSProvider.synthesize metering', () => {
         expect(cost).toBe(expectedCost);
     });
 
-    it('asks for hasEnoughCredits with the same total it later meters', async () => {
+    it('reserves the same total it later meters', async () => {
         const provider = makeProvider();
         speechCreateMock.mockResolvedValueOnce(mockAudioResponse());
 
@@ -473,8 +477,8 @@ describe('OpenAITTSProvider.synthesize metering', () => {
         );
 
         const expectedCost = OPENAI_TTS_COSTS['tts-1'] * text.length;
-        // First call to hasEnoughCredits should match the metered cost.
-        const creditCall = hasCreditsSpy.mock.calls[0]!;
+        // The first reservation should match the metered cost.
+        const creditCall = admitSpy.mock.calls[0]!;
         expect(creditCall[1]).toBe(expectedCost);
     });
 });

src/backend/drivers/ai-tts/providers/speechify/SpeechifyTTSProvider.test.ts

@@ -56,7 +56,7 @@ import { SPEECHIFY_TTS_COSTS } from './costs.js';
 
 let server: PuterServer;
 let fetchSpy: MockInstance<typeof fetch>;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
 beforeAll(async () => {
@@ -82,7 +82,7 @@ const audioResponse = (audioData = Buffer.from('audio-bytes').toString('base64')
 
 beforeEach(() => {
     fetchSpy = vi.spyOn(globalThis, 'fetch') as MockInstance<typeof fetch>;
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
 });
 
@@ -159,7 +159,7 @@ describe('SpeechifyTTSProvider.synthesize test_mode', () => {
             url: 'https://puter-sample-data.puter.site/tts_example.mp3',
             content_type: 'audio',
         });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(fetchSpy).not.toHaveBeenCalled();
     });
 });
@@ -206,7 +206,7 @@ describe('SpeechifyTTSProvider.synthesize argument validation', () => {
 describe('SpeechifyTTSProvider.synthesize credit gate', () => {
     it('throws 402 BEFORE hitting Speechify when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
 
         await expect(
             withTestActor(() => provider.synthesize({ text: 'hi' })),
@@ -324,7 +324,7 @@ describe('SpeechifyTTSProvider.synthesize metering', () => {
         expect(cost).toBe(expectedCost);
     });
 
-    it('asks for hasEnoughCredits with the same total it later meters', async () => {
+    it('reserves the same total it later meters', async () => {
         const provider = makeProvider();
         fetchSpy.mockResolvedValueOnce(audioResponse());
 
@@ -332,7 +332,7 @@ describe('SpeechifyTTSProvider.synthesize metering', () => {
         await withTestActor(() => provider.synthesize({ text }));
 
         const expectedCost = SPEECHIFY_TTS_COSTS['simba-3.2'] * text.length;
-        expect(hasCreditsSpy.mock.calls[0]![1]).toBe(expectedCost);
+        expect(admitSpy.mock.calls[0]![1]).toBe(expectedCost);
     });
 });
 

src/backend/drivers/ai-tts/providers/xai/XAITTSProvider.test.ts

@@ -56,7 +56,7 @@ import { XAI_TTS_COSTS } from './costs.js';
 
 let server: PuterServer;
 let fetchSpy: MockInstance<typeof fetch>;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
 beforeAll(async () => {
@@ -79,7 +79,7 @@ const audioResponse = (body = 'audio-bytes', contentType = 'audio/mpeg') =>
 
 beforeEach(() => {
     fetchSpy = vi.spyOn(globalThis, 'fetch') as MockInstance<typeof fetch>;
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
 });
 
@@ -160,7 +160,7 @@ describe('XAITTSProvider.synthesize test_mode', () => {
             url: 'https://puter-sample-data.puter.site/tts_example.mp3',
             content_type: 'audio',
         });
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(fetchSpy).not.toHaveBeenCalled();
     });
 });
@@ -195,7 +195,7 @@ describe('XAITTSProvider.synthesize argument validation', () => {
 describe('XAITTSProvider.synthesize credit gate', () => {
     it('throws 402 BEFORE hitting xAI when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
 
         await expect(
             withTestActor(() => provider.synthesize({ text: 'hi' })),
@@ -327,7 +327,7 @@ describe('XAITTSProvider.synthesize metering', () => {
         expect(cost).toBe(expectedCost);
     });
 
-    it('asks for hasEnoughCredits with the same total it later meters', async () => {
+    it('reserves the same total it later meters', async () => {
         const provider = makeProvider();
         fetchSpy.mockResolvedValueOnce(audioResponse());
 
@@ -335,7 +335,7 @@ describe('XAITTSProvider.synthesize metering', () => {
         await withTestActor(() => provider.synthesize({ text }));
 
         const expectedCost = XAI_TTS_COSTS['xai-tts'] * text.length;
-        expect(hasCreditsSpy.mock.calls[0]![1]).toBe(expectedCost);
+        expect(admitSpy.mock.calls[0]![1]).toBe(expectedCost);
     });
 });
 

src/backend/drivers/ai-video/VideoGenerationDriver.test.ts

@@ -94,7 +94,7 @@ vi.mock('../../util/secureHttp.js', async (importOriginal) => ({
 
 let server: PuterServer;
 let driver: VideoGenerationDriver;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 
 beforeAll(async () => {
     server = await setupTestServer({
@@ -122,8 +122,10 @@ beforeEach(() => {
     togetherVideosCreateMock.mockReset();
     togetherVideosRetrieveMock.mockReset();
     secureFetchMock.mockReset();
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
-    hasCreditsSpy.mockResolvedValue(true);
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
+    admitSpy.mockImplementation((who, amount, opts) =>
+            server.services.metering.reserveCredits(who, amount, opts),
+        );
     vi.spyOn(server.services.metering, 'getRemainingUsage').mockResolvedValue(
         100_000_000_000,
     );

src/backend/drivers/ai-video/providers/gemini/GeminiVideoProvider.test.ts

@@ -106,7 +106,7 @@ const MP4 = Buffer.concat([
 let server: PuterServer;
 let videoDriver: VideoGenerationDriver;
 let fetchSpy: MockInstance<typeof fetch>;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let batchSpy: MockInstance<MeteringService['batchIncrementUsages']>;
 
 const generate = (args: Record<string, unknown> = {}) =>
@@ -139,9 +139,11 @@ beforeEach(() => {
     secureFetchMock.mockReset();
     secureFetchMock.mockRejectedValue(new Error('unexpected secureFetch'));
     // The full hold is more than a new account's allowance.
-    hasCreditsSpy = vi
-        .spyOn(server.services.metering, 'hasEnoughCredits')
-        .mockResolvedValue(true);
+    admitSpy = vi
+        .spyOn(server.services.metering, 'reserveCreditsIfAffordable')
+        .mockImplementation((who, amount, opts) =>
+            server.services.metering.reserveCredits(who, amount, opts),
+        );
     batchSpy = vi.spyOn(server.services.metering, 'batchIncrementUsages');
 });
 
@@ -589,7 +591,7 @@ describe('GeminiVideoProvider test_mode', () => {
     it('returns the sample clip without network, credits or metering', async () => {
         expect(await testMode()).toBe(SAMPLE_URL);
         expect(fetchSpy).not.toHaveBeenCalled();
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(batchSpy).not.toHaveBeenCalled();
     });
 
@@ -1028,7 +1030,7 @@ describe('GeminiVideoProvider source video', () => {
 
         const path = await writeVideo(user, 'dog.mp4', MP4);
         const read = vi.spyOn(server.stores.s3Object, 'getObjectStream');
-        hasCreditsSpy.mockResolvedValue(false);
+        admitSpy.mockResolvedValue(null);
         countOnly();
         await expect(
             withActor(user.actor, () => generate({ input_video: path })),
@@ -1573,12 +1575,12 @@ describe('GeminiVideoProvider countTokens preflight', () => {
 describe('GeminiVideoProvider credit hold', () => {
     it('throws 402 after counting tokens, before generating', async () => {
         const { actor } = await makeUser();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
         fetchSpy.mockResolvedValueOnce(Response.json({ totalTokens: 18 }));
         await expect(withActor(actor, () => generate())).rejects.toMatchObject({
             statusCode: 402,
         });
-        expect(hasCreditsSpy.mock.calls[0]![1]).toBe(HOLD);
+        expect(admitSpy.mock.calls[0]![1]).toBe(HOLD);
         expect(fetchSpy).toHaveBeenCalledTimes(1);
         expect(batchSpy).not.toHaveBeenCalled();
     });

src/backend/drivers/ai-video/providers/together/TogetherVideoProvider.test.ts

@@ -85,7 +85,7 @@ vi.mock('together-ai', () => {
 // ── Test harness ────────────────────────────────────────────────────
 
 let server: PuterServer;
-let hasCreditsSpy: MockInstance<MeteringService['hasEnoughCredits']>;
+let admitSpy: MockInstance<MeteringService['reserveCreditsIfAffordable']>;
 let remainingUsageSpy: MockInstance<MeteringService['getRemainingUsage']>;
 let incrementUsageSpy: MockInstance<MeteringService['incrementUsage']>;
 
@@ -117,11 +117,13 @@ beforeEach(() => {
     videosCreateMock.mockReset();
     videosRetrieveMock.mockReset();
     togetherCtor.mockReset();
-    hasCreditsSpy = vi.spyOn(server.services.metering, 'hasEnoughCredits');
+    admitSpy = vi.spyOn(server.services.metering, 'reserveCreditsIfAffordable');
     // Default to "has credits" so per-test setup only needs to override
     // for the explicit credit-gate scenarios. SYSTEM_ACTOR has no uuid,
     // so the live `getRemainingUsage` path can short-circuit to 0.
-    hasCreditsSpy.mockResolvedValue(true);
+    admitSpy.mockImplementation((who, amount, opts) =>
+            server.services.metering.reserveCredits(who, amount, opts),
+        );
     remainingUsageSpy = vi.spyOn(server.services.metering, 'getRemainingUsage');
     remainingUsageSpy.mockResolvedValue(AMPLE_CREDIT);
     incrementUsageSpy = vi.spyOn(server.services.metering, 'incrementUsage');
@@ -197,7 +199,7 @@ describe('TogetherVideoProvider.generate test_mode', () => {
             provider.generate({ prompt: 'hi', test_mode: true }),
         );
         expect(result).toBe('https://assets.puter.site/txt2vid.mp4');
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
         expect(videosCreateMock).not.toHaveBeenCalled();
     });
 });
@@ -238,7 +240,7 @@ const DIRECTOR = 'togetherai:minimax/video-01-director';
 describe('TogetherVideoProvider.generate credit gate', () => {
     it('throws 402 BEFORE hitting Together when actor lacks credits', async () => {
         const provider = makeProvider();
-        hasCreditsSpy.mockResolvedValueOnce(false);
+        admitSpy.mockResolvedValueOnce(null);
 
         await expect(
             withTestActor(() =>
@@ -715,7 +717,7 @@ describe('TogetherVideoProvider.generate per-second models', () => {
         expect('height' in sent).toBe(false);
         expect('ratio' in sent).toBe(false);
         // The flat-rate gate is not consulted for per-second models.
-        expect(hasCreditsSpy).not.toHaveBeenCalled();
+        expect(admitSpy).not.toHaveBeenCalled();
 
         const seedance = TOGETHER_VIDEO_GENERATION_MODELS.find(
             (m) => m.model === 'ByteDance/Seedance-2.5',

src/backend/drivers/util/aiCostFactor.test.ts

@@ -324,8 +324,10 @@ describe('AI cost factor', () => {
         it('gates and holds at the factored cost', async () => {
             listen(1.3);
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             const reserve = vi.spyOn(metering, 'reserveCredits');
             try {
                 const hold = await scoped.reserveAiCredits(
@@ -334,7 +336,11 @@ describe('AI cost factor', () => {
                     1000,
                 );
                 await hold?.release();
-                expect(check).toHaveBeenCalledWith(actor, 1300);
+                expect(check).toHaveBeenCalledExactlyOnceWith(
+                    actor,
+                    1300,
+                    undefined,
+                );
                 expect(reserve).toHaveBeenCalledWith(actor, 1300, undefined);
             } finally {
                 check.mockRestore();
@@ -342,10 +348,40 @@ describe('AI cost factor', () => {
             }
         });
 
+        it('admits one of two concurrent holds the balance only covers once', async () => {
+            listen(1.3);
+            vi.spyOn(metering, 'getAllowedUsage').mockResolvedValue({
+                remaining: 20_000_000,
+                monthUsageAllowance: 20_000_000,
+                addons: {},
+            });
+            const holds = await Promise.all(
+                [0, 1].map(() =>
+                    scoped.reserveAiCredits(
+                        actor,
+                        'openai:gpt-realtime:realtime:audio_output',
+                        12_500_000,
+                    ),
+                ),
+            );
+            try {
+                expect(holds.filter((h) => h !== null)).toHaveLength(1);
+                // The factor applies once: 12.5M at 1.3 is 16.25M held.
+                expect(await metering.getOutstandingHolds(actor)).toBe(
+                    16_250_000,
+                );
+            } finally {
+                await Promise.all(holds.map((h) => h?.release()));
+                vi.restoreAllMocks();
+            }
+        });
+
         it('passes a hold lifetime through to the service', async () => {
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             const reserve = vi.spyOn(metering, 'reserveCredits');
             try {
                 const hold = await scoped.reserveAiCredits(
@@ -367,8 +403,12 @@ describe('AI cost factor', () => {
         it('takes no hold when the factored cost is unaffordable', async () => {
             listen(1.3);
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockImplementation(async (_actor, amount) => amount <= 1000);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation(async (who, amount, opts) =>
+                    amount <= 1000
+                        ? metering.reserveCredits(who, amount, opts)
+                        : null,
+                );
             const reserve = vi.spyOn(metering, 'reserveCredits');
             try {
                 const hold = await scoped.reserveAiCredits(
@@ -409,8 +449,10 @@ describe('AI cost factor', () => {
         it('holds the factored cost while run executes, then releases it', async () => {
             listen(1.3);
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             try {
                 let heldDuringRun = 0;
                 const result = await withAiCreditHold(
@@ -448,8 +490,10 @@ describe('AI cost factor', () => {
             };
             server.clients.event.on('ai.cost.factor.*', slowFactor as never);
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             try {
                 await runWithContext({ actor }, () =>
                     withAiCreditHold(
@@ -485,8 +529,10 @@ describe('AI cost factor', () => {
 
         it('releases the hold when the usage write fails', async () => {
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             const write = vi
                 .spyOn(metering, 'incrementUsage')
                 .mockRejectedValueOnce(new Error('ledger down'));
@@ -515,8 +561,10 @@ describe('AI cost factor', () => {
 
         it('releases at once when nothing was recorded', async () => {
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             try {
                 await runWithContext({ actor }, () =>
                     withAiCreditHold(
@@ -535,8 +583,10 @@ describe('AI cost factor', () => {
 
         it('releases the hold when run throws', async () => {
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             const boom = new Error('upstream down');
             try {
                 await expect(
@@ -552,8 +602,8 @@ describe('AI cost factor', () => {
 
         it('refuses with a 402 and never runs when the cost is unaffordable', async () => {
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(false);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockResolvedValue(null);
             const run = vi.fn();
             try {
                 await expect(
@@ -575,8 +625,10 @@ describe('AI cost factor', () => {
         const relay = async (source: Readable) => {
             const record = vi.fn();
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             try {
                 const result = await withStreamedAiCharge(
                     scoped,
@@ -624,8 +676,10 @@ describe('AI cost factor', () => {
                 landCharge = resolve;
             });
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             try {
                 const result = await withStreamedAiCharge(
                     scoped,
@@ -655,8 +709,10 @@ describe('AI cost factor', () => {
             vi.useFakeTimers({ toFake: ['setInterval', 'clearInterval'] });
             const refresh = vi.spyOn(server.stores.creditHold, 'refresh');
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             let landCharge = () => {};
             const charge = new Promise<void>((resolve) => {
                 landCharge = resolve;
@@ -697,8 +753,10 @@ describe('AI cost factor', () => {
             vi.useFakeTimers({ toFake: ['setInterval', 'clearInterval'] });
             const refresh = vi.spyOn(server.stores.creditHold, 'refresh');
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             let opened = () => {};
             const opening = new Promise<void>((resolve) => {
                 opened = resolve;
@@ -741,8 +799,10 @@ describe('AI cost factor', () => {
 
         it('releases the hold when recording the charge throws', async () => {
             const check = vi
-                .spyOn(metering, 'hasEnoughCredits')
-                .mockResolvedValue(true);
+                .spyOn(metering, 'reserveCreditsIfAffordable')
+                .mockImplementation((who, amount, opts) =>
+                    metering.reserveCredits(who, amount, opts),
+                );
             try {
                 const result = await withStreamedAiCharge(
                     scoped,

src/backend/drivers/util/aiCostFactor.ts

@@ -354,10 +354,12 @@ export function withAiCostFactor(
             ? await costFactor(actor, aiModelKey(usageType))
             : 1;
         const cost = factor === 1 ? amount : scaleCost(amount, factor);
-        if (!(await metering.hasEnoughCredits(actor, cost))) return null;
-        return outlastingWrites(
-            await metering.reserveCredits(actor, cost, opts),
+        const hold = await metering.reserveCreditsIfAffordable(
+            actor,
+            cost,
+            opts,
         );
+        return hold && outlastingWrites(hold);
     };
 
     const reserveCredits = async (

src/backend/server.test.ts

@@ -336,6 +336,70 @@ describe('PuterServer host header validation', () => {
         expect(api.headers['x-frame-options']).toBeUndefined();
     });
 
+    it('allows deployment subdomains to frame the app browser embed', async () => {
+        const parent = `http://spreadsheet.puter.localhost:${port}`;
+        for (const path of ['/embed/apps', '/embed/apps/']) {
+            const res = await request(path, {
+                host: `puter.localhost:${port}`,
+                referer: `${parent}/sheet/1`,
+            });
+            expect(res.status).toBe(200);
+            expect(res.headers['x-frame-options']).toBeUndefined();
+            expect(res.headers['content-security-policy']).toBe(
+                `frame-ancestors 'self' ${parent}`,
+            );
+            expect(res.headers['cache-control']).toBe('no-store');
+            expect(res.body).toContain('/dist/apps-embed.min.js');
+            expect(res.body).not.toContain('/dist/bundle.min.js');
+            expect(res.body).toContain(`http://api.puter.localhost:${port}`);
+            const config = JSON.parse(
+                res.body.match(/id="app-browser-config">(.*?)<\/script>/)[1],
+            );
+            expect(Array.isArray(config.guestApps)).toBe(true);
+        }
+        const gui = await request('/embed/apps', {
+            host: `puter.localhost:${port}`,
+            referer: `http://puter.localhost:${port}/dashboard`,
+        });
+        expect(gui.headers['content-security-policy']).toBe("frame-ancestors 'self'");
+        const dashboard = await request('/dashboard', { host: `puter.localhost:${port}` });
+        expect(dashboard.headers['x-frame-options']).toBe('SAMEORIGIN');
+        const api = await request('/embed/apps', { host: `api.puter.localhost:${port}` });
+        expect(api.status).toBe(404);
+    });
+
+    it.each([
+        ['a user-hosted site', 'http://someone.site.puter.localhost:PORT/'],
+        ['a private app host', 'http://someone.app.puter.localhost:PORT/'],
+        ['the API host', 'http://api.puter.localhost:PORT/'],
+        ['the WebDAV host', 'http://dav.puter.localhost:PORT/'],
+        ['a native app host', 'http://editor.puter.localhost:PORT/'],
+        ['a nested subdomain', 'http://a.spreadsheet.puter.localhost:PORT/'],
+        ['another site', 'http://evil.test:PORT/'],
+        ['another port', 'http://spreadsheet.puter.localhost:1/'],
+        ['a suppressed referrer', undefined],
+    ])('keeps the embed same-origin only when framed by %s', async (_label, referer) => {
+        const res = await request('/embed/apps', {
+            host: `puter.localhost:${port}`,
+            ...(referer ? { referer: referer.replace('PORT', String(port)) } : {}),
+        });
+        expect(res.status).toBe(200);
+        expect(res.headers['content-security-policy']).toBe("frame-ancestors 'self'");
+    });
+
+    it.each([
+        ['plain', ''],
+        // Past the parser's 1000-pair limit, so `req.query` comes back empty.
+        ['padded', '&'.repeat(1000)],
+    ])('strips %s SDK bootstrap query parameters from the embed route', async (_label, padding) => {
+        const res = await request(`/embed/apps?${padding}puter.app_instance_id=other&puter.api_origin=https://other.test`, {
+            host: `puter.localhost:${port}`,
+        });
+        expect(res.status).toBe(302);
+        expect(res.headers.location).toBe('/embed/apps');
+        expect(res.headers['x-frame-options']).toBe('SAMEORIGIN');
+    });
+
     it('blocks a request the ip.validate listeners veto', async () => {
         const handler = (_key: unknown, data: unknown) => {
             (data as { allow: boolean }).allow = false;

src/backend/services/apps/RecommendedAppsService.ts

@@ -27,17 +27,16 @@ import { PuterService } from '../types.js';
  */
 export const RECOMMENDED_APP_NAMES = [
     'builder',
+    'word-processor',
+    'spreadsheet',
+    'presentation',
+    'whiteboard',
+    'pdf-editor',
     'contacts',
     'calendar',
     'meetings',
     'teamchat',
     'email',
-    'whiteboard',
-    'spreadsheet',
-    'word-processor',
-    'presentation',
-    'pdf-editor',
-    'terminal',
     'dev-center',
     'cap-table',
     'invoices',

src/backend/services/events/registry.ts

@@ -121,7 +121,9 @@ export interface ProjectedNotifEvent extends ProjectedEventBase {
 }
 
 export type ProjectedEvent =
-    ProjectedFsEvent | ProjectedKvEvent | ProjectedNotifEvent;
+    | ProjectedFsEvent
+    | ProjectedKvEvent
+    | ProjectedNotifEvent;
 
 export type GapReason =
     | 'matched_subscription_limit'
@@ -270,7 +272,9 @@ export interface NotifPublicSubject extends SubjectSpec<
 }
 
 export type PublicSubject =
-    FsPublicSubject | KvPublicSubject | NotifPublicSubject;
+    | FsPublicSubject
+    | KvPublicSubject
+    | NotifPublicSubject;
 
 export interface UnpublishedInternalEvent {
     event: EventKey;
@@ -508,6 +512,10 @@ export const UNPUBLISHED_INTERNAL_EVENTS = [
         event: 'notif.delivery.settled',
         reason: 'A node-local report on how the live socket copy of a notification went; the notification itself publishes through `notif.created`.',
     },
+    {
+        event: 'notif.handled',
+        reason: 'A node-local report that a page saw a notification live, for extensions deciding whether to push it; not a change to the notification.',
+    },
 ] as const satisfies readonly UnpublishedInternalEvent[];
 
 // Every `fs.*`/`kv.*`/`notif.*` bus key needs a decision. A new key that is in

src/backend/services/events/userHandlers.worker.integration.test.ts

@@ -144,11 +144,22 @@ const createInline = async (
     return String(response.body.subId);
 };
 
-const delivered = (count = 1): Promise<void> =>
-    vi.waitFor(() => expect(sinkPosts.length).toBeGreaterThanOrEqual(count), {
-        timeout: 30_000,
-        interval: 50,
-    });
+/**
+ * The post each label made, in order, once all have arrived. Matched by label
+ * since earlier tests' subscriptions stay live and post to the same sink.
+ */
+const delivered = async (...labels: string[]): Promise<SinkPost[]> => {
+    await vi.waitFor(
+        () =>
+            expect(sinkPosts.map((post) => post.label)).toEqual(
+                expect.arrayContaining(labels),
+            ),
+        { timeout: 30_000, interval: 50 },
+    );
+    return labels.map((label) =>
+        sinkPosts.find((post) => post.label === label)!,
+    );
+};
 
 beforeAll(async () => {
     env = await setupPuterTestEnv({
@@ -255,14 +266,11 @@ describe('an inline handler in a user scope, end to end', () => {
             await env.server.services.fs.touch(holder.id, {
                 path: `${holder.anchor}/note.txt`,
             });
-            await delivered();
+            const [post] = await delivered('inline');
 
-            expect(sinkPosts[0]).toMatchObject({
-                label: 'inline',
-                path: `${holder.anchor}/note.txt`,
-            });
+            expect(post.path).toBe(`${holder.anchor}/note.txt`);
             const { actor } = await env.server.services.auth.authenticate(
-                sinkPosts[0].token!,
+                post.token!,
             );
             expect(actor!.user?.id).toBe(holder.id);
             expect(actor!.effectiveApp?.uid).toBe(appUid);
@@ -297,14 +305,14 @@ describe('an inline handler in a user scope, end to end', () => {
                 await env.server.services.fs.touch(holder.id, {
                     path: `${holder.anchor}/shared-${i}.txt`,
                 });
-            await delivered(2);
-            const runAs = new Set<number>();
-            for (const post of sinkPosts)
-                runAs.add(
-                    (await env.server.services.auth.authenticate(post.token!))
-                        .actor!.user!.id,
+            // The first test's 'inline' handler also posts for holders[0].
+            const posts = await delivered('first', 'second');
+            for (const [i, post] of posts.entries()) {
+                const { actor } = await env.server.services.auth.authenticate(
+                    post.token!,
                 );
-            expect(runAs).toEqual(new Set(holders.map((h) => h.id)));
+                expect(actor!.user!.id).toBe(holders[i].id);
+            }
             create.mockRestore();
         },
     );

src/backend/services/homepage/PuterHomepageService.test.ts

@@ -49,6 +49,64 @@ const makeReq = (over: Partial<Request> = {}): Request =>
         ...over,
     }) as unknown as Request;
 
+describe('app browser embed shell', () => {
+    it.each([
+        [{ env: 'dev' }, '/sdk/puter.dev.js'],
+        [{ env: 'prod' }, 'https://js.puter.com/v2/'],
+        [{ gui_puterjs_bundle: '/custom-sdk.js' }, '/custom-sdk.js'],
+    ])('uses the deployment SDK configuration %j', (config, sdkUrl) => {
+        const service = makeService({
+            ...config,
+            api_base_url: 'https://api.deployment.test',
+            domain: 'deployment.test',
+        });
+        const html = service.renderAppBrowserEmbed();
+        const configJson = html.match(/id="app-browser-config">(.*?)<\/script>/)![1];
+        expect(JSON.parse(configJson)).toEqual({
+            apiOrigin: 'https://api.deployment.test', sdkUrl, guestApps: [],
+        });
+        expect(html).toContain('/dist/apps-embed.min.js');
+        expect(html).not.toContain('/dist/bundle.min.js');
+        expect(html).not.toContain('gui(');
+    });
+
+    it.each([
+        [{}, '/dist/apps-embed.min.js'],
+        [
+            { gui_bundle: 'https://gui-cdn.deployment.test/bundle.min.js' },
+            'https://gui-cdn.deployment.test/apps-embed.min.js',
+        ],
+        [
+            { gui_bundle: '/static/bundle.min.js?v=42' },
+            '/static/apps-embed.min.js?v=42',
+        ],
+    ])('loads its script from beside the GUI bundle %j', (config, script) => {
+        const html = makeService(config).renderAppBrowserEmbed();
+        expect(html).toContain(`<script defer src="${script}"></script>`);
+    });
+
+    it('escapes configured values without breaking its JSON bootstrap', () => {
+        const sdkUrl = '/sdk.js?value=</script><script>unexpected()</script>';
+        const html = makeService({
+            gui_params: { title: '<script>unexpected()</script>' },
+            gui_puterjs_bundle: sdkUrl,
+        }).renderAppBrowserEmbed();
+        expect(html).not.toContain('<script>unexpected()');
+        const configJson = html.match(/id="app-browser-config">(.*?)<\/script>/)![1];
+        expect(JSON.parse(configJson).sdkUrl).toBe(sdkUrl);
+    });
+
+    it('carries the apps shown without a session, escaped', () => {
+        const guestApps = [
+            { name: 'notes', title: '</script><script>unexpected()</script>' },
+        ];
+        const html = makeService({}).renderAppBrowserEmbed({ guestApps });
+        expect(html).not.toContain('<script>unexpected()');
+        const configJson = html.match(/id="app-browser-config">(.*?)<\/script>/)![1];
+        expect(JSON.parse(configJson).guestApps).toEqual(guestApps);
+    });
+});
+
 /** Capture what the service sends, and return the rendered HTML. */
 const render = async (
     service: PuterHomepageService,

src/backend/services/homepage/PuterHomepageService.ts

@@ -167,6 +167,56 @@ export class PuterHomepageService extends PuterService {
         this.#guiParams[key] = val;
     }
 
+    /** `guestApps` is what the embed shows a visitor without a session. */
+    renderAppBrowserEmbed({
+        guestApps = [],
+    }: { guestApps?: Array<Record<string, unknown>> } = {}): string {
+        const configJson = JSON.stringify({
+            apiOrigin: this.config.api_base_url,
+            sdkUrl:
+                this.config.env === 'dev'
+                    ? '/sdk/puter.dev.js'
+                    : (this.config.gui_puterjs_bundle ??
+                      'https://js.puter.com/v2/'),
+            guestApps,
+        }).replace(/</g, '\\u003c');
+
+        return `<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="utf-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1">
+    <meta name="robots" content="noindex">
+    <title>${encode(String(this.config.gui_params?.title ?? 'Puter'))}</title>
+    <link rel="stylesheet" href="/css/normalize.css">
+    <link rel="stylesheet" href="/css/dashboard.css">
+    <link rel="stylesheet" href="/css/apps-embed.css">
+</head>
+<body>
+    <main id="app-browser" class="dashboard"></main>
+    <script type="application/json" id="app-browser-config">${configJson}</script>
+    <script defer src="${encode(this.#appBrowserEmbedScript())}"></script>
+</body>
+</html>`;
+    }
+
+    /**
+     * The embed's script is built beside the GUI bundle, so it is served from
+     * wherever `gui_bundle` is (a CDN in production), with the same query.
+     */
+    #appBrowserEmbedScript(): string {
+        const base = 'http://relative.invalid';
+        const bundle = new URL(
+            this.config.gui_bundle ?? '/dist/bundle.min.js',
+            base,
+        );
+        const script = new URL('apps-embed.min.js', bundle);
+        script.search = bundle.search;
+        return script.origin === base
+            ? `${script.pathname}${script.search}`
+            : script.href;
+    }
+
     /** `scriptPath` is a same-origin absolute path, like `/ext/push/worker.js`. */
     registerServiceWorkerScript(scriptPath: string): void {
         if (

src/backend/services/localworker/LocalWorkerService.shutdown.test.ts

@@ -78,7 +78,10 @@ describe('LocalWorkerService shutdown disposal', () => {
         disposals[0].mockImplementationOnce(async () => {
             await held;
         });
-        vi.useFakeTimers();
+        vi.useFakeTimers({ toFake: ['setTimeout', 'clearTimeout'] });
+        // The test server keeps arming timers from real callbacks while fakes
+        // are installed, so only what shutdown arms synchronously is checked.
+        const arm = vi.spyOn(globalThis, 'setTimeout');
         try {
             let finished = false;
             const draining = Promise.resolve(service.onServerShutdown()).then(
@@ -86,12 +89,13 @@ describe('LocalWorkerService shutdown disposal', () => {
                     finished = true;
                 },
             );
+            expect(arm.mock.calls.map(([, ms]) => ms)).toEqual([5_000]);
             await vi.advanceTimersByTimeAsync(4_999);
             expect(finished).toBe(false);
             await vi.advanceTimersByTimeAsync(1);
             await draining;
-            expect(vi.getTimerCount()).toBe(0);
         } finally {
+            arm.mockRestore();
             release();
             vi.useRealTimers();
         }

src/backend/services/metering/MeteringService.test.ts

@@ -4290,6 +4290,295 @@ describe('MeteringService', () => {
         });
     });
 
+    describe('reserveCreditsIfAffordable', () => {
+        const BUDGET = 20_000_000;
+        const HOLD = 16_200_000;
+
+        afterEach(() => {
+            vi.restoreAllMocks();
+        });
+
+        /** Spends the fresh actor's allowance down to exactly `balance`. */
+        const withBalance = async (balance: number) => {
+            const { remaining } = await target.getAllowedUsage(actor);
+            if (remaining > balance) {
+                await target.incrementUsage(
+                    actor,
+                    'kv:read',
+                    1,
+                    remaining - balance,
+                );
+            }
+            expect((await target.getAllowedUsage(actor)).remaining).toBe(
+                balance,
+            );
+        };
+
+        it('admits exactly one of two concurrent holds that only fit alone', async () => {
+            await withBalance(BUDGET);
+            const holds = await Promise.all([
+                target.reserveCreditsIfAffordable(actor, HOLD),
+                target.reserveCreditsIfAffordable(actor, HOLD),
+            ]);
+
+            expect(holds.filter((h) => h !== null)).toHaveLength(1);
+            expect(await target.getOutstandingHolds(actor)).toBe(HOLD);
+            await Promise.all(holds.map((h) => h?.release()));
+        });
+
+        it('compares against the gross balance, counting live holds once', async () => {
+            await withBalance(BUDGET);
+            const other = await target.reserveCredits(actor, 5_000_000);
+
+            expect(
+                await target.reserveCreditsIfAffordable(actor, 15_000_001),
+            ).toBeNull();
+            const fits = await target.reserveCreditsIfAffordable(
+                actor,
+                15_000_000,
+            );
+            expect(fits).not.toBeNull();
+            expect(await target.getOutstandingHolds(actor)).toBe(BUDGET);
+
+            // A release gives the headroom back for a retry.
+            await other.release();
+            expect(
+                await target.reserveCreditsIfAffordable(actor, 5_000_000),
+            ).not.toBeNull();
+        });
+
+        it('counts allowance and purchased credit together', async () => {
+            await withBalance(1_000_000);
+            await target.updateAddonCredit(actor.user.uuid!, 2_000_000);
+
+            expect(
+                await target.reserveCreditsIfAffordable(actor, 3_000_001),
+            ).toBeNull();
+            expect(
+                await target.reserveCreditsIfAffordable(actor, 3_000_000),
+            ).not.toBeNull();
+        });
+
+        it('reads the balance again when usage lands and a hold is released after the read', async () => {
+            await withBalance(BUDGET);
+            // A request still running, about to record its usage and release.
+            const running = await target.reserveCredits(actor, HOLD);
+            const settle = async () => {
+                await target.incrementUsage(actor, 'kv:read', 1, HOLD);
+                await running.release();
+            };
+            // The admission's balance read finishes first, then the
+            // settlement lands before its reservation runs.
+            const read = target.getAllowedUsage.bind(target);
+            let settled = false;
+            vi.spyOn(target, 'getAllowedUsage').mockImplementation(
+                async (...args) => {
+                    const balance = await read(...args);
+                    if (!settled) {
+                        settled = true;
+                        await settle();
+                    }
+                    return balance;
+                },
+            );
+
+            expect(await target.reserveCreditsIfAffordable(actor, HOLD)).toBe(
+                null,
+            );
+            expect(await target.getOutstandingHolds(actor)).toBe(0);
+        });
+
+        it('would over-admit that interleaving without the version check', async () => {
+            await withBalance(BUDGET);
+            const running = await target.reserveCredits(actor, HOLD);
+            const read = target.getAllowedUsage.bind(target);
+            let settled = false;
+            vi.spyOn(target, 'getAllowedUsage').mockImplementation(
+                async (...args) => {
+                    const balance = await read(...args);
+                    if (!settled) {
+                        settled = true;
+                        await target.incrementUsage(actor, 'kv:read', 1, HOLD);
+                        await running.release();
+                    }
+                    return balance;
+                },
+            );
+            vi.spyOn(server.stores.creditHold, 'version').mockResolvedValue('');
+
+            const hold = await target.reserveCreditsIfAffordable(actor, HOLD);
+            expect(hold).not.toBeNull();
+            await hold?.release();
+        });
+
+        it('marks the balance changed once a usage write lands', async () => {
+            const before = await server.stores.creditHold.version(
+                actor.user.uuid!,
+            );
+            await target.incrementUsage(actor, 'kv:read', 1, 1_000);
+            expect(
+                await server.stores.creditHold.version(actor.user.uuid!),
+            ).not.toBe(before);
+        });
+
+        it('lets the balance decide, holding nothing, when the cache is unavailable', async () => {
+            await withBalance(BUDGET);
+            vi.spyOn(server.stores.creditHold, 'reserve').mockResolvedValue({
+                outcome: 'unavailable',
+            });
+
+            const hold = await target.reserveCreditsIfAffordable(actor, HOLD);
+            expect(hold).not.toBeNull();
+            expect(await target.getOutstandingHolds(actor)).toBe(0);
+            expect(
+                await target.reserveCreditsIfAffordable(actor, BUDGET + 1),
+            ).toBeNull();
+        });
+
+        it('reads the balance again while it keeps changing, then refuses', async () => {
+            await withBalance(BUDGET);
+            const reserve = vi
+                .spyOn(server.stores.creditHold, 'reserve')
+                .mockResolvedValue({ outcome: 'stale' });
+
+            expect(await target.reserveCreditsIfAffordable(actor, HOLD)).toBe(
+                null,
+            );
+            expect(reserve).toHaveBeenCalledTimes(5);
+            // Every attempt checks the version; none falls back to skipping it.
+            expect(reserve.mock.calls.every(([, , , v]) => v !== '')).toBe(
+                true,
+            );
+            expect(await target.getOutstandingHolds(actor)).toBe(0);
+        });
+
+        it('admits once the balance holds still', async () => {
+            await withBalance(BUDGET);
+            const reserve = vi.spyOn(server.stores.creditHold, 'reserve');
+            reserve
+                .mockResolvedValueOnce({ outcome: 'stale' })
+                .mockResolvedValueOnce({ outcome: 'stale' });
+
+            const hold = await target.reserveCreditsIfAffordable(actor, HOLD);
+            expect(hold).not.toBeNull();
+            expect(reserve).toHaveBeenCalledTimes(3);
+            await hold?.release();
+        });
+
+        it('reads purchased credit consistently only while some is left', async () => {
+            await withBalance(BUDGET);
+            const get = vi.spyOn(server.stores.kv, 'get');
+            const consistentAddonReads = () =>
+                get.mock.calls.filter(
+                    ([args]) =>
+                        args.consistentRead &&
+                        String(args.key).endsWith(':addons'),
+                ).length;
+            // With none, the cached record that says so answers.
+            await target.getAllowedUsage(actor);
+            get.mockClear();
+            await (
+                await target.reserveCreditsIfAffordable(actor, 1000)
+            )?.release();
+            expect(consistentAddonReads()).toBe(0);
+
+            await target.updateAddonCredit(actor.user.uuid!, 5_000);
+            get.mockClear();
+            await (
+                await target.reserveCreditsIfAffordable(actor, 1000)
+            )?.release();
+            expect(consistentAddonReads()).toBe(1);
+        });
+
+        it('marks the balance changed on credit, allowance and manual usage changes', async () => {
+            const uuid = actor.user.uuid!;
+            const versions = [await server.stores.creditHold.version(uuid)];
+            await target.updateAddonCredit(uuid, 1_000);
+            versions.push(await server.stores.creditHold.version(uuid));
+            await target.setActorCurrentMonthUsageTotal(actor, 5_000);
+            versions.push(await server.stores.creditHold.version(uuid));
+            expect(new Set(versions).size).toBe(3);
+        });
+
+        it('treats the system actor, a userless actor and amounts it can’t hold as the plain check does', async () => {
+            for (const [who, amount] of [
+                [SYSTEM_ACTOR, HOLD],
+                [actor, 0],
+                [actor, -5],
+            ] as const) {
+                const hold = await target.reserveCreditsIfAffordable(
+                    who,
+                    amount,
+                );
+                expect(hold).not.toBeNull();
+            }
+            for (const amount of [Number.NaN, Number.POSITIVE_INFINITY]) {
+                expect(
+                    await target.reserveCreditsIfAffordable(actor, amount),
+                ).toBeNull();
+            }
+            await expect(
+                target.reserveCreditsIfAffordable({ user: {} } as Actor, HOLD),
+            ).rejects.toThrow();
+            expect(await target.getOutstandingHolds(actor)).toBe(0);
+        });
+
+        it('passes the hold lifetime through and renews it', async () => {
+            await withBalance(BUDGET);
+            const hold = await target.reserveCreditsIfAffordable(actor, 750, {
+                ttlMs: 500,
+            });
+            await new Promise((r) => setTimeout(r, 620));
+            expect(await target.getOutstandingHolds(actor)).toBe(0);
+            await hold?.extend?.();
+            expect(await target.getOutstandingHolds(actor)).toBe(750);
+            await hold?.release();
+            expect(await target.getOutstandingHolds(actor)).toBe(0);
+        });
+    });
+
+    describe('settling a charge while a flush is out', () => {
+        it('splits a charge over the allowance using usage a flush has claimed', async () => {
+            const uuid = actor.user.uuid!;
+            const { monthUsageAllowance: allowance } =
+                await target.getActorSubscription(actor);
+            await target.updateAddonCredit(uuid, 1_000_000);
+            const Service = target.constructor as typeof MeteringService;
+            const threshold = Service.PRECISION_THRESHOLD;
+            // Keeps the first charge buffered instead of read exactly.
+            Service.PRECISION_THRESHOLD = 2;
+            await target.incrementUsage(actor, 'kv:read', 1, allowance - 5);
+            Service.PRECISION_THRESHOLD = threshold;
+
+            // A flush claims it and stalls before writing it onward.
+            const kv = server.stores.kv;
+            const realIncr = kv.incr.bind(kv);
+            let release: () => void = () => {};
+            const gate = new Promise<void>((r) => (release = r));
+            const write = vi
+                .spyOn(kv, 'incr')
+                .mockImplementation(async (...args) => {
+                    await gate;
+                    return realIncr(...args);
+                });
+            const flushing = server.stores.meteringBuffer.flushCycle();
+            await vi.waitFor(() => expect(write).toHaveBeenCalled());
+            write.mockImplementation(realIncr);
+
+            // 5 µ¢ of this is left in the allowance; the rest is credit.
+            await target.incrementUsage(actor, 'kv:read', 1, 10);
+            release();
+            await flushing;
+            write.mockRestore();
+            await waitFor(async () => {
+                const addons = await target.getActorAddons(actor, {
+                    consistentRead: true,
+                });
+                expect(addons.consumedPurchaseCredits).toBe(5);
+            });
+        });
+    });
+
     // ── Resolver registration ────────────────────────────────────────
 
     describe('resolver registration', () => {

src/backend/services/metering/MeteringService.ts

@@ -698,6 +698,8 @@ export class MeteringService extends PuterService {
                 actorAddons,
                 totalBatchCost,
             );
+            // Before any hold this usage stands for is released.
+            await this.stores.creditHold.bump(userId);
 
             this.maybeAlertOveruse({
                 actor,
@@ -729,6 +731,10 @@ export class MeteringService extends PuterService {
                 ? ({ ...own, ...charged } as UsageByType)
                 : (charged ?? own);
         } catch (e) {
+            // Part of the usage may have landed before the failure.
+            if (actor?.user?.uuid) {
+                await this.stores.creditHold.bump(actor.user.uuid);
+            }
             console.error('[metering] batchIncrementUsages failed', {
                 actor,
                 usages,
@@ -1432,6 +1438,7 @@ export class MeteringService extends PuterService {
         }
 
         this.invalidateActorCredits(userId);
+        await this.stores.creditHold.bump(userId);
         await Promise.all([
             this.stores.meteringBuffer.forgetSummed(actorUsageKey),
             this.stores.meteringBuffer.forgetSummed(
@@ -1545,7 +1552,86 @@ export class MeteringService extends PuterService {
             opts.ttlMs,
         );
         if (!member) return NO_CREDIT_HOLD;
+        return this.#holdHandle(userId, member, opts.ttlMs);
+    }
 
+    /**
+     * Commit part of an actor's budget only if it fits beside what their other
+     * operations hold; null when it doesn't. The check and the hold are one
+     * step in the cache, and a usage write or release that lands after the
+     * balance read sends the reservation back to read it again. When the cache
+     * can't answer, the balance alone decides and nothing is held.
+     */
+    async reserveCreditsIfAffordable(
+        actor: Actor,
+        amount: number,
+        opts: { ttlMs?: number } = {},
+    ): Promise<CreditHold | null> {
+        const userId = actor?.user?.uuid;
+        if (
+            !userId ||
+            isSystemActor(actor) ||
+            !Number.isFinite(amount) ||
+            !(amount > 0)
+        ) {
+            // Nothing to hold; the plain check decides, as it always has.
+            return (await this.hasEnoughCredits(actor, amount))
+                ? NO_CREDIT_HOLD
+                : null;
+        }
+
+        for (let attempt = 1; ; attempt++) {
+            const version = await this.stores.creditHold.version(userId);
+            // Read after the version, so a write it misses bumps the version.
+            const { remaining } = await this.getAllowedUsage(actor, {
+                exact: true,
+            });
+            const budget = remaining || 0;
+            const result =
+                version === null
+                    ? ({ outcome: 'unavailable' } as const)
+                    : await this.stores.creditHold.reserve(
+                          userId,
+                          amount,
+                          budget,
+                          version,
+                          opts.ttlMs,
+                      );
+            switch (result.outcome) {
+                case 'accepted':
+                    return this.#holdHandle(userId, result.member, opts.ttlMs);
+                case 'insufficient':
+                    return null;
+                case 'unavailable':
+                    return budget >= amount ? NO_CREDIT_HOLD : null;
+                case 'stale':
+                    break;
+            }
+            // A balance that never holds still long enough to read is refused
+            // rather than checked against a read that may be out of date.
+            if (attempt >= MeteringService.RESERVE_ATTEMPTS) {
+                console.warn(
+                    `[metering] credit reservation refused for ${userId}: balance kept changing`,
+                );
+                return null;
+            }
+            await new Promise((resolve) =>
+                setTimeout(
+                    resolve,
+                    Math.random() *
+                        MeteringService.RESERVE_BACKOFF_MS *
+                        attempt,
+                ),
+            );
+        }
+    }
+
+    /** Reservations tried while the balance keeps changing under them. */
+    private static readonly RESERVE_ATTEMPTS = 5;
+    /** Upper bound of the randomized wait before each retry, times the attempt. */
+    private static readonly RESERVE_BACKOFF_MS = 20;
+
+    #holdHandle(userId: string, member: string, ttlMs?: number): CreditHold {
         let released = false;
         return {
             release: async () => {
@@ -1555,11 +1641,7 @@ export class MeteringService extends PuterService {
             },
             extend: async () => {
                 if (released) return;
-                await this.stores.creditHold.refresh(
-                    userId,
-                    member,
-                    opts.ttlMs,
-                );
+                await this.stores.creditHold.refresh(userId, member, ttlMs);
             },
         };
     }
@@ -1571,14 +1653,22 @@ export class MeteringService extends PuterService {
         return this.stores.creditHold.outstanding(userId);
     }
 
-    async getAllowedUsage(actor: Actor): Promise<{
+    /**
+     * `exact` reads purchased credit consistently, for a spend decision that
+     * has to see the consumption a write just recorded. A cached record with no
+     * credit left still answers: it can only understate the balance.
+     */
+    async getAllowedUsage(
+        actor: Actor,
+        { exact = false }: { exact?: boolean } = {},
+    ): Promise<{
         remaining: number;
         monthUsageAllowance: number;
         addons: UsageAddons;
     }> {
         const [userSubscription, addons, totals] = await Promise.all([
             this.getActorSubscription(actor),
-            this.#addonsForSettle(actor),
+            this.#addonsForSettle(actor, { consistentRead: exact }),
             this.currentMonthTotals(actor),
         ]);
 
@@ -2245,13 +2335,16 @@ export class MeteringService extends PuterService {
     }
 
     /** `getActorAddons`, through `#spentAddons`. Shared by gates and settles. */
-    async #addonsForSettle(actor: Actor): Promise<UsageAddons> {
+    async #addonsForSettle(
+        actor: Actor,
+        { consistentRead }: { consistentRead?: boolean } = {},
+    ): Promise<UsageAddons> {
         const uuid = actor.user?.uuid;
         const cached = uuid ? this.#spentAddons.get(uuid) : undefined;
         if (cached && cached.expiresAt > Date.now()) return cached.addons;
 
         const drops = this.#spentAddonsDrops;
-        const addons = await this.getActorAddons(actor);
+        const addons = await this.getActorAddons(actor, { consistentRead });
         if (!uuid) return addons;
         this.#spentAddons.delete(uuid);
         const creditLeft =
@@ -2382,6 +2475,7 @@ export class MeteringService extends PuterService {
         // Credit that lands while the account is being turned away has to take
         // effect on the next request, not at the end of the cache window.
         this.invalidateActorCredits(userId);
+        await this.stores.creditHold.bump(userId);
     }
 
     /**
@@ -2429,6 +2523,7 @@ export class MeteringService extends PuterService {
         });
         // Every node drops its cached policy, and the budget answer with it.
         this.invalidateActorSubscription(userUuid);
+        await this.stores.creditHold.bump(userUuid);
     }
 
     /**

src/backend/services/notification/NotificationService.ts

@@ -23,10 +23,14 @@ import { handlerDepthOf, type Actor } from '../../core/actor.js';
 import { Context } from '../../core/context.js';
 import type { NotifEventContext } from '../events/registry.js';
 import { PuterService } from '../types.js';
+import { checkRateLimit } from '../../core/http/middleware/rateLimit.js';
 import {
+    NOTIF_HANDLED_EVENT,
+    NOTIF_HANDLED_PER_MINUTE,
     NotificationSocketAdapter,
     notificationAckTimeoutMs,
     notificationsFoldInEnabled,
+    type NotifSocket,
     type ReceiptScope,
 } from './notificationSocket.js';
 import {
@@ -404,6 +408,11 @@ export class NotificationService extends PuterService {
         return this.#socket().receipt(scope);
     }
 
+    /** Installs the `notif.handled` verb on one connection. */
+    attachSocket(socket: NotifSocket, actor: Actor): void {
+        this.#socket().attachSocket(socket, actor);
+    }
+
     /**
      * Drop notifications past the retention window, in batches, and report how
      * many went. Deleting is all there is to it: nothing is pushed, because a
@@ -439,6 +448,12 @@ export class NotificationService extends PuterService {
             ackTimeoutMs: () => notificationAckTimeoutMs(this.config),
             heldInPeerRegion: (userId, appUid) =>
                 this.#heldInPeerRegion(userId, appUid),
+            allowHandled: (userId) =>
+                checkRateLimit(
+                    `socket:${NOTIF_HANDLED_EVENT}:${userId}`,
+                    NOTIF_HANDLED_PER_MINUTE,
+                    60_000,
+                ),
         });
         return this.#adapter;
     }

src/backend/services/notification/notificationDelivery.test.ts

@@ -32,7 +32,10 @@ import {
 import type { EventMap } from '../../clients/event/types.js';
 import type { PuterServer } from '../../server.js';
 import { accountSocketRoom, SocketService } from '../socket/SocketService.js';
-import { NOTIF_RECEIPT_EVENT } from './notificationSocket.js';
+import {
+    NOTIF_HANDLED_EVENT,
+    NOTIF_RECEIPT_EVENT,
+} from './notificationSocket.js';
 import {
     allocateEphemeralPort,
     createTestUser,
@@ -309,12 +312,13 @@ describe('live notification delivery receipts', () => {
             complete: true,
             peerRegions: false,
             responders: [{ visibilityState: 'hidden' }],
+            handled: false,
         };
         expect(settled).toEqual(expected);
         expect(state.settled).toEqual([expected]);
     });
 
-    it('passes a tab`s focus and own display through, typed strictly', async () => {
+    it('passes a tab`s focus through, typed strictly', async () => {
         const answers = [
             {
                 received: true,
@@ -326,7 +330,6 @@ describe('live notification delivery receipts', () => {
                 received: true,
                 visibilityState: 'nope',
                 focused: 1,
-                showsOwn: 'yes',
             },
         ];
         for (const answer of answers) {
@@ -341,11 +344,39 @@ describe('live notification delivery receipts', () => {
         expect(settled.outcome).toBe('received');
         expect(settled.responders).toEqual(
             expect.arrayContaining([
-                { visibilityState: 'hidden', focused: false, showsOwn: true },
+                { visibilityState: 'hidden', focused: false },
                 { visibilityState: 'unknown' },
             ]),
         );
         expect(settled.responders).toHaveLength(2);
+        expect(settled.handled).toBe(false);
+    });
+
+    it('counts as handled only a visible tab that is focused, or an older one that is visible', async () => {
+        const cases: Array<[Record<string, unknown>, boolean]> = [
+            [{ visibilityState: 'visible', focused: true }, true],
+            [{ visibilityState: 'visible', focused: false }, false],
+            [{ visibilityState: 'hidden', focused: true }, false],
+            [{ visibilityState: 'visible' }, true],
+        ];
+        for (const [answer, handled] of cases) {
+            const tab = await connect({ acks: [NOTIF_RECEIPT_EVENT] });
+            tab.on(NOTIF_RECEIPT_EVENT, (_data: unknown, ack: Ack) =>
+                ack({ received: true, ...answer }),
+            );
+
+            const { settled } = await ask();
+
+            expect(settled.handled, JSON.stringify(answer)).toBe(handled);
+            tab.disconnect();
+            await vi.waitFor(() =>
+                expect(
+                    state.server.services.socket.has({
+                        room: accountSocketRoom(state.userId),
+                    }),
+                ).toBe(false),
+            );
+        }
     });
 
     it('settles as unreceived when every tab answers that it never got it', async () => {
@@ -549,6 +580,7 @@ describe('live receipts from an app’s own sockets', () => {
             complete: true,
             peerRegions: false,
             responders: [{ visibilityState: 'visible' }],
+            handled: true,
         };
         expect(settled).toEqual(expected);
         expect(state.settled).toEqual([expected]);
@@ -719,3 +751,59 @@ describe('app receipts where the adapter cannot count servers', () => {
         });
     });
 });
+
+describe('a page reporting a notification it handled', () => {
+    const { state, connect, appSocket } = withAppReceipts({ cluster: true });
+    const reports: Array<EventMap['notif.handled']> = [];
+    const onHandled = (_key: string, data: unknown) => {
+        reports.push(data as EventMap['notif.handled']);
+    };
+
+    beforeAll(() => {
+        state.server.clients.event.on('notif.handled', onHandled);
+    });
+
+    afterEach(() => {
+        reports.length = 0;
+    });
+
+    afterAll(() => {
+        state.server?.clients.event.off('notif.handled', onHandled);
+    });
+
+    it('relays a desktop tab`s report for its user, without an app', async () => {
+        const tab = await connect({});
+        const uid = uuidv4();
+
+        tab.emit(NOTIF_HANDLED_EVENT, { uid });
+
+        await vi.waitFor(() =>
+            expect(reports).toEqual([{ userId: state.userId, uid }]),
+        );
+    });
+
+    it('relays an app page`s report with the app it acts as', async () => {
+        const page = await appSocket();
+        const uid = uuidv4();
+
+        page.emit(NOTIF_HANDLED_EVENT, { uid });
+
+        await vi.waitFor(() =>
+            expect(reports).toEqual([
+                { userId: state.userId, uid, appUid: OWN_APP },
+            ]),
+        );
+    });
+
+    it('drops a report without a notification uid', async () => {
+        const tab = await connect({});
+
+        tab.emit(NOTIF_HANDLED_EVENT, { uid: 'not-a-uid' });
+        tab.emit(NOTIF_HANDLED_EVENT, null);
+        tab.emit(NOTIF_HANDLED_EVENT, { uid: uuidv4() });
+
+        await vi.waitFor(() => expect(reports).toHaveLength(1));
+        await new Promise((resolve) => setTimeout(resolve, 100));
+        expect(reports).toHaveLength(1);
+    });
+});

src/backend/services/notification/notificationSocket.ts

@@ -17,7 +17,9 @@
  * along with this program.  If not, see <https://www.gnu.org/licenses/>.
  */
 
+import { validate as validateUuid } from 'uuid';
 import type { EventClient } from '../../clients/event/EventClient.js';
+import type { Actor } from '../../core/actor.js';
 import type {
     EventMap,
     NotifDeliveryOutcome,
@@ -53,7 +55,10 @@ import type { IConfig } from '../../types.js';
  */
 
 export type NotifWire =
-    'notif.message' | 'notif.unreads' | 'notif.ack' | 'notif.removed';
+    | 'notif.message'
+    | 'notif.unreads'
+    | 'notif.ack'
+    | 'notif.removed';
 
 export const NOTIF_DELIVERY_EVENT = 'outer.notif.delivery';
 
@@ -67,6 +72,15 @@ export const DEFAULT_NOTIFICATION_ACK_TIMEOUT_MS = 5_000;
  */
 export const NOTIF_RECEIPT_EVENT = 'notif.receipt';
 
+/**
+ * Sent by a page with `{ uid }` when it got that notification live while
+ * visible and focused, so a region that didn't ask the page can still hear it.
+ */
+export const NOTIF_HANDLED_EVENT = 'notif.handled';
+
+/** `notif.handled` reports one user's sockets may send per minute. */
+export const NOTIF_HANDLED_PER_MINUTE = 60;
+
 /** One unread as the replay payload carries it. */
 export interface UnreadNotification {
     uid: string;
@@ -122,12 +136,16 @@ const responderOf = (response: unknown): NotifDeliveryResponder | null => {
         ...(typeof answer.focused === 'boolean'
             ? { focused: answer.focused }
             : {}),
-        ...(typeof answer.showsOwn === 'boolean'
-            ? { showsOwn: answer.showsOwn }
-            : {}),
     };
 };
 
+/**
+ * Visible and focused. An older client doesn't say whether it's focused, so it
+ * counts while visible.
+ */
+export const handledBy = (responder: NotifDeliveryResponder): boolean =>
+    responder.visibilityState === 'visible' && responder.focused !== false;
+
 interface AdapterDeps {
     event: EventClient;
     socket: SocketService;
@@ -138,6 +156,13 @@ interface AdapterDeps {
         userId: number,
         appUid: string | null,
     ) => Promise<boolean>;
+    /** Spends one of the user's `notif.handled` reports; false past the limit. */
+    allowHandled: (userId: number) => Promise<boolean>;
+}
+
+/** The part of a socket `attachSocket` installs its verb on. */
+export interface NotifSocket {
+    on(event: string, listener: (...args: never[]) => void): unknown;
 }
 
 type SettledDelivery = EventMap['notif.delivery.settled'];
@@ -253,6 +278,7 @@ export class NotificationSocketAdapter {
                 complete: asked.complete,
                 peerRegions,
                 responders,
+                handled: responders.some(handledBy),
             };
             this.#deps.event.emit(
                 'notif.delivery.settled',
@@ -266,6 +292,28 @@ export class NotificationSocketAdapter {
         }
     }
 
+    /**
+     * Relays a page's `notif.handled` as the node-local `notif.handled` event,
+     * scoped to the socket's user and app.
+     */
+    attachSocket(socket: NotifSocket, actor: Actor): void {
+        const userId = actor.user?.id;
+        if (userId === undefined) return;
+        const appUid = actor.effectiveApp?.uid;
+        socket.on(NOTIF_HANDLED_EVENT, ((payload: unknown) => {
+            const uid = (payload as { uid?: unknown } | null)?.uid;
+            if (typeof uid !== 'string' || !validateUuid(uid)) return;
+            void this.#deps.allowHandled(userId).then((ok) => {
+                if (!ok) return;
+                this.#deps.event.emit(
+                    'notif.handled',
+                    { userId, uid, ...(appUid ? { appUid } : {}) },
+                    {},
+                );
+            });
+        }) as (...args: never[]) => void);
+    }
+
     /** Whether this region holds a socket for the recipient right now. */
     hasSocket(userId: number): boolean {
         return this.#deps.socket.has({ room: userId });

src/backend/services/realtime/README.md

@@ -11,7 +11,10 @@ Direct clients are documented in the [Realtime WebSocket API reference](../../..
 
 The relay forwards native OpenAI JSON events unchanged after validating the
 event envelope, except that `rate_limits.updated` (Puter's provider account) is
-dropped and provider organization IDs are removed from error messages. An event must be an object with a nonempty string `type` of at
+dropped and provider organization IDs are removed from diagnostic messages in
+`error`, `response.done.status_details.error`, and
+`conversation.item.input_audio_transcription.failed.error`, and from provider
+close reasons. Conversation content is preserved. An event must be an object with a nonempty string `type` of at
 most 128 characters. Binary messages and malformed envelopes close the session.
 Usage fields consumed by metering are validated before use. Event field order
 does not affect billing.
@@ -46,7 +49,8 @@ traffic is not rate limited: OpenAI sends audio faster than real time, and
 backpressure bounds what a slow client can queue. Usage is priced as each event
 arrives, before any forwarding decision, and recorded serially off the
 forwarding path through the AI cost-factor hook. Pricing follows the model the
-session reports, falling back to the connection's; a client `session.update`
+session reports. A missing model uses the last known model, while an explicitly
+unpriced model closes the connection; a client `session.update`
 that changes the model is refused. So are `conversation.item.retrieve` and MCP
 tools in `session.update` or `response.create`, since a retrieved item or MCP
 output can exceed the upstream message cap.
@@ -65,9 +69,55 @@ Each response holds a fixed amount of the account's budget
 until its usage is recorded, so concurrent sessions can't overdraw it; a session
 opens only if one hold fits. Input transcription is priced at the costlier of
 the transcription models in force at commit and at completion. Session attempts
-are limited per credential before authentication runs.
+are limited per credential after signature verification. Local, bounded IP and
+node admission limits protect signature verification and authentication lookups.
+Upgrades use Express's configured proxy trust policy and the `ip.validate` gate.
+Authentication uses `touch: false` on both initial connection and rechecks, so
+opening or maintaining a realtime socket does not update session activity metadata.
+
+Input transcription holds credit per item, at a fixed budget for the selected
+transcription model and that model's cost factor. Audio and the configuration
+that enables transcription are forwarded only while a reserved spare item
+exists. Each provider commit takes one, and the pool refills to two spares in
+the background, up to 64 pending plus spare items. A commit with no spare left
+reserves its own item. A model change, requested or reported, releases the spares
+and reserves new ones at the new model's price. Committed items keep their hold
+and model. Disabling transcription keeps the spares until the connection ends.
+If the account can't cover another item, client reads pause while the session
+still holds items that may free credit, and the session is refused with
+`insufficient_funds` once it holds none.
+
+Audio the provider may commit without the relay seeing the acknowledgement is
+tracked apart from the pool, as a duration at the session's input format. Without
+turn detection that is all audio since the last acknowledged commit or clear.
+With server turn detection it is only speech from `speech_started`, since the
+provider drops silence, plus the buffer after an explicit commit. Each started
+160 seconds of it claims one held item, and an acknowledged commit, a clear, or
+an `input_audio_buffer_commit_empty` error returns those items to the spares.
+The session is never closed for its length. If the upstream closes first, each
+claimed item is billed. Each committed item retains its model
+until completion, failure, or bounded settlement. Missing usage is billed the
+item's fixed budget. Overflow closes the connection without evicting existing
+pending identities.
+
+The post-record continuation check asks whether settled credit remains. It
+excludes outstanding holds and accepts fractional positive credit. Admission of
+new billable work continues to use the atomic hold path, which includes every
+outstanding reservation.
+
+When the host listens for `ai.prompt.validate`, each response that starts runs
+it on the text the client sent since the previous check: session and response
+instructions, message text and transcripts, and function call output, as chat
+messages. The response's cost follows as `ai.prompt.cost-calculated` under the
+same completion id. A refusal closes the session with `permission_denied`; the
+response may already be streaming by then. Pending text past 64 KiB is checked
+on its own. Audio is not checked.
 
 Credential checks run in batches of at most 16 with no overlapping sweeps.
+Session and access-token revocation events close affected sockets and invalidate
+their auth cache entries. Subscription eligibility is checked independently of
+authentication and after subscription invalidations. A failed verification check after a 60-second
+retry grace from the first failed lookup closes with a retryable internal error. Periodic checks remain the missed-event fallback.
 When the drain starts, open sessions get a `puter.reconnect` event and new
 sessions receive `server_restarting` and close with 1012. Open sessions can
 finish an in-flight response for up to five seconds at shutdown, then close

src/backend/services/socket/SocketService.ts

@@ -699,7 +699,8 @@ export class SocketService extends PuterService {
             // `{ auth: { ... } }`, not the query string. puter-js uses
             // `io(url, { auth: { auth_token } })`.
             const handshakeAuth = socket.handshake.auth as
-                Record<string, unknown> | undefined;
+                | Record<string, unknown>
+                | undefined;
             const tokenRaw =
                 typeof handshakeAuth?.auth_token === 'string'
                     ? handshakeAuth.auth_token
@@ -1085,6 +1086,7 @@ export class SocketService extends PuterService {
             // them. Off unless events are enabled, in which case the verbs
             // answer with `events_disabled` rather than going unanswered.
             this.services.events.attachSocket(socket, actor);
+            this.services.notification.attachSocket(socket, actor);
 
             // Everything below is the desktop session's own traffic: two verbs
             // that reach the user's other tabs, and a connect announcement
@@ -1221,7 +1223,8 @@ export class SocketService extends PuterService {
                 // their next poll of /cache/last-change-timestamp.
                 const originalSocketId = (
                     data.response as
-                        { original_client_socket_id?: string } | undefined
+                        | { original_client_socket_id?: string }
+                        | undefined
                 )?.original_client_socket_id;
                 await this.send({ room: userId }, 'cache.updated', {
                     timestamp,
@@ -1235,7 +1238,9 @@ export class SocketService extends PuterService {
     #handleUploadProgress(data: UploadProgressPayload): void {
         const meta = data.meta ?? {};
         const userId = (meta.user_id ?? meta.userId) as
-            number | string | undefined;
+            | number
+            | string
+            | undefined;
         if (!userId) {
             console.warn('[socket] upload-progress missing user_id', { meta });
             return;

src/backend/stores/app/AppStore.test.js

@@ -1542,18 +1542,22 @@ describe('AppStore detailed stats', () => {
     let db;
 
     const APP_UID = `app-stats-${Date.now()}`;
-    const nowSeconds = Math.floor(Date.now() / 1000);
+    // The windows come from the clock, so it is pinned: mid-week and
+    // mid-month, and minutes clear of midnight in any time zone.
+    const NOW = new Date('2026-06-17T12:07:30Z');
+    const nowSeconds = Math.floor(NOW.getTime() / 1000);
 
     beforeAll(async () => {
         server = await setupTestServer();
         appStore = server.stores.app;
         db = server.clients.db;
         await clearAppCache(server.clients.redis);
+        vi.useFakeTimers({ toFake: ['Date'], now: NOW });
 
-        // Two users; the repeat shares a timestamp so it is always same-hour.
+        // Two users in one hour. Windows end at now, exclusive.
         for (const [userId, offset] of [
-            [1, 0],
-            [1, 0],
+            [1, 60],
+            [1, 60],
             [2, 120],
         ]) {
             await db.write(
@@ -1564,6 +1568,7 @@ describe('AppStore detailed stats', () => {
     });
 
     afterAll(async () => {
+        vi.useRealTimers();
         await server?.shutdown();
     });
 

src/backend/stores/metering/CreditHoldStore.test.ts

@@ -21,6 +21,7 @@ import { afterAll, beforeAll, describe, expect, it, vi } from 'vitest';
 
 import { PuterServer } from '../../server.js';
 import { setupTestServer } from '../../testUtil.js';
+import type { IConfig, IRedisConfig } from '../../types.js';
 import type { CreditHoldStore } from './CreditHoldStore.js';
 
 let server: PuterServer;
@@ -142,3 +143,204 @@ describe('CreditHoldStore', () => {
         spy.mockRestore();
     });
 });
+
+// Opt in to a real Redis Cluster with PUTER_TEST_REDIS_CLUSTER=host:port.
+const clusterNode = process.env.PUTER_TEST_REDIS_CLUSTER;
+const backends: { name: string; redis: IRedisConfig }[] = [
+    { name: 'the redis mock', redis: { useMock: true } },
+    ...(clusterNode
+        ? [
+              {
+                  name: `redis cluster ${clusterNode}`,
+                  redis: {
+                      useMock: false,
+                      tls: false,
+                      startupNodes: [
+                          {
+                              host: clusterNode.split(':')[0],
+                              port: Number(clusterNode.split(':')[1]),
+                          },
+                      ],
+                  },
+              },
+          ]
+        : []),
+];
+
+describe.each(backends)('CreditHoldStore.reserve on $name', ({ redis }) => {
+    // Two backends sharing one cache, as separate instances do in production.
+    let servers: PuterServer[];
+    let a: CreditHoldStore;
+    let b: CreditHoldStore;
+    const BUDGET = 20_000_000;
+    const HOLD = 16_200_000;
+
+    beforeAll(async () => {
+        servers = await Promise.all(
+            [0, 1].map(() => setupTestServer({ redis } as IConfig)),
+        );
+        [a, b] = servers.map((s) => s.stores.creditHold as CreditHoldStore);
+    });
+
+    afterAll(async () => {
+        await Promise.all((servers ?? []).map((s) => s.shutdown()));
+    });
+
+    const reserve = async (
+        on: CreditHoldStore,
+        id: string,
+        amount: number,
+        budget = BUDGET,
+        ttlMs?: number,
+    ) => on.reserve(id, amount, budget, (await on.version(id))!, ttlMs);
+
+    it('admits exactly one of two concurrent holds that only fit alone', async () => {
+        const id = user();
+        const results = await Promise.all([
+            reserve(a, id, HOLD),
+            reserve(b, id, HOLD),
+        ]);
+
+        expect(results.map((r) => r.outcome).sort()).toEqual([
+            'accepted',
+            'insufficient',
+        ]);
+        expect(await a.outstanding(id)).toBe(HOLD);
+        expect(await b.outstanding(id)).toBe(HOLD);
+    });
+
+    it('admits exact headroom and refuses one unit more', async () => {
+        const id = user();
+        expect((await reserve(a, id, 15_000_000)).outcome).toBe('accepted');
+        expect((await reserve(b, id, 5_000_001)).outcome).toBe('insufficient');
+        expect((await reserve(b, id, 5_000_000)).outcome).toBe('accepted');
+        expect(await a.outstanding(id)).toBe(BUDGET);
+    });
+
+    it('rounds a fractional amount up once', async () => {
+        const id = user();
+        expect((await reserve(a, id, 9.2, 10)).outcome).toBe('accepted');
+        expect(await a.outstanding(id)).toBe(10);
+        expect((await reserve(a, id, 0.5, 10)).outcome).toBe('insufficient');
+    });
+
+    it('admits several holds that fit together, and keeps actors apart', async () => {
+        const [id, other] = [user(), user()];
+        const results = await Promise.all(
+            [a, b, a, b].map((s) => reserve(s, id, 5_000_000)),
+        );
+        expect(results.every((r) => r.outcome === 'accepted')).toBe(true);
+        expect((await reserve(a, other, HOLD)).outcome).toBe('accepted');
+        expect(await a.outstanding(id)).toBe(BUDGET);
+        expect(await a.outstanding(other)).toBe(HOLD);
+    });
+
+    it('leaves nothing behind when it refuses', async () => {
+        const id = user();
+        await reserve(a, id, HOLD);
+        await reserve(b, id, HOLD);
+        await reserve(b, id, HOLD, 0);
+        expect(await a.outstanding(id)).toBe(HOLD);
+    });
+
+    it('frees headroom on release and on expiry', async () => {
+        const id = user();
+        const first = await reserve(a, id, HOLD);
+        expect((await reserve(b, id, HOLD)).outcome).toBe('insufficient');
+        await a.release(id, first.outcome === 'accepted' ? first.member : null);
+        expect((await reserve(b, id, HOLD, BUDGET, 50)).outcome).toBe(
+            'accepted',
+        );
+        await vi.waitFor(
+            async () =>
+                expect((await reserve(a, id, HOLD)).outcome).toBe('accepted'),
+            { timeout: 2000, interval: 25 },
+        );
+    });
+
+    it('keeps a long hold when a short one is reserved beside it', async () => {
+        const id = user();
+        await reserve(a, id, 1000);
+        await reserve(b, id, 500, BUDGET, 50);
+        const pttl = await (
+            servers[0].clients.redis as unknown as {
+                pttl: (key: string) => Promise<number>;
+            }
+        ).pttl(`meter:holds:{${id}}`);
+        expect(pttl).toBeGreaterThan(500_000);
+    });
+
+    it('refuses a budget read before a usage write landed, but not one before a release', async () => {
+        const id = user();
+        // A release alone leaves the balance as it was.
+        const before = (await a.version(id))!;
+        const held = await reserve(a, id, 1000);
+        await b.release(id, held.outcome === 'accepted' ? held.member : null);
+        const after = await a.reserve(id, 1000, BUDGET, before);
+        expect(after.outcome).toBe('accepted');
+        await a.release(id, after.outcome === 'accepted' ? after.member : null);
+
+        const read = (await a.version(id))!;
+        await b.bump(id);
+        expect((await a.reserve(id, 1000, BUDGET, read)).outcome).toBe('stale');
+        expect(await a.outstanding(id)).toBe(0);
+
+        // Skipping the check still can't overlap another hold.
+        await reserve(a, id, HOLD);
+        expect((await a.reserve(id, HOLD, BUDGET, '')).outcome).toBe(
+            'insufficient',
+        );
+    });
+
+    it('keeps renewal and release consistent when they race', async () => {
+        const id = user();
+        const held = await reserve(a, id, HOLD, BUDGET, 200);
+        const member = held.outcome === 'accepted' ? held.member : null;
+        await Promise.all([a.refresh(id, member), b.release(id, member)]);
+        // Whichever ran last decides; a renewal that lost the race only
+        // brings the hold back until its owner releases it again.
+        await a.release(id, member);
+        expect(await a.outstanding(id)).toBe(0);
+        expect((await reserve(b, id, HOLD)).outcome).toBe('accepted');
+    });
+
+    it('refuses amounts it can’t hold without touching the hold set', async () => {
+        const id = user();
+        await reserve(a, id, 1000);
+        for (const amount of [0, -5_000, Number.NaN, Infinity]) {
+            expect((await reserve(a, id, amount)).outcome).toBe('insufficient');
+        }
+        expect(await a.outstanding(id)).toBe(1000);
+    });
+
+    it('starts a missing version from a random value', async () => {
+        const id = user();
+        expect(await a.version(id)).toBe('0');
+        await b.bump(id);
+        const first = (await a.version(id))!;
+        expect(first).not.toBe('1');
+        await b.bump(id);
+        expect(Number(await a.version(id))).toBe(Number(first) + 1);
+    });
+
+    it('reports an unreachable cache as unavailable, not as insufficient', async () => {
+        const id = user();
+        const client = servers[0].clients.redis as unknown as {
+            creditHoldReserve: () => Promise<unknown>;
+            get: () => Promise<unknown>;
+        };
+        const reserveSpy = vi
+            .spyOn(client, 'creditHoldReserve')
+            .mockRejectedValueOnce(new Error('cache down'));
+        expect((await a.reserve(id, 1000, BUDGET, '0')).outcome).toBe(
+            'unavailable',
+        );
+        const getSpy = vi
+            .spyOn(client, 'get')
+            .mockRejectedValueOnce(new Error('cache down'));
+        expect(await a.version(id)).toBeNull();
+        reserveSpy.mockRestore();
+        getSpy.mockRestore();
+        expect(await a.outstanding(id)).toBe(0);
+    });
+});

src/backend/stores/metering/CreditHoldStore.ts

@@ -48,6 +48,25 @@ const KEY_TTL_SLACK_MS = 60 * 1000;
 
 const holdsKey = (userId: string): string => `meter:holds:{${userId}}`;
 
+/**
+ * Counts changes to an actor's balance and holds, so a reservation can tell
+ * whether the balance it read is still current. Shares the hold key's slot.
+ */
+const versionKey = (userId: string): string =>
+    `meter:holds:{${userId}}:version`;
+
+/** Outlives any read-then-reserve window by orders of magnitude. */
+const VERSION_TTL_MS = 60 * 60 * 1000;
+
+/** What a conditional reservation came to. */
+export type CreditReservation =
+    | { outcome: 'accepted'; member: string }
+    | { outcome: 'insufficient' }
+    // The balance changed since `version` was read; read it again.
+    | { outcome: 'stale' }
+    // The cache couldn't answer; nothing was held.
+    | { outcome: 'unavailable' };
+
 // -- Scripts ----------------------------------------------------------
 
 /**
@@ -79,8 +98,8 @@ return 1
  * name (`<id>:<amount>`) so one sorted set carries both the deadline and the
  * number, and pruning and summing happen in the same pass.
  */
-const SUM_SCRIPT = `
-local dropped = redis.call('ZREMRANGEBYSCORE', KEYS[1], '-inf', ARGV[1])
+const pruneAndSum = (now: string) => `
+local dropped = redis.call('ZREMRANGEBYSCORE', KEYS[1], '-inf', ${now})
 local total = 0
 local members = redis.call('ZRANGE', KEYS[1], 0, -1)
 for i = 1, #members do
@@ -89,13 +108,65 @@ for i = 1, #members do
         total = total + (tonumber(string.sub(members[i], sep + 1)) or 0)
     end
 end
+`;
+
+const SUM_SCRIPT = `${pruneAndSum('ARGV[1]')}
 return { tostring(total), dropped }
 `;
 
+/**
+ * KEYS: the actor's hold set, its version. ARGV: member, deadline, key ttl,
+ * now, amount, budget, expected version ('' skips the check).
+ *
+ * Takes the hold only if it fits in `budget` beside the live holds, in one step
+ * so two reservations can't both fit in the same headroom. The version check
+ * refuses a budget read before a usage write or release that has since landed.
+ */
+const RESERVE_SCRIPT = `
+if ARGV[7] ~= '' and (redis.call('GET', KEYS[2]) or '0') ~= ARGV[7] then
+    return { 'stale', 0 }
+end
+${pruneAndSum('ARGV[4]')}
+if total + tonumber(ARGV[5]) > tonumber(ARGV[6]) then
+    return { 'insufficient', dropped }
+end
+local ttl = redis.call('PTTL', KEYS[1])
+redis.call('ZADD', KEYS[1], ARGV[2], ARGV[1])
+if tonumber(ARGV[3]) > ttl then
+    redis.call('PEXPIRE', KEYS[1], ARGV[3])
+end
+return { 'accepted', dropped }
+`;
+
+/**
+ * Bumps the version in KEYS[n], starting a missing one from a random value so a
+ * version read before the key expired can't match its successor.
+ */
+const bumpVersion = (key: string, seed: string, ttl: string) => `
+if redis.call('EXISTS', ${key}) == 0 then
+    redis.call('SET', ${key}, ${seed})
+end
+redis.call('INCR', ${key})
+redis.call('PEXPIRE', ${key}, ${ttl})
+`;
+
+/** KEYS: the actor's hold version. ARGV: version ttl, seed. */
+const BUMP_SCRIPT = `
+${bumpVersion('KEYS[1]', 'ARGV[2]', 'ARGV[1]')}
+return 1
+`;
+
+/** A random starting version; see `bumpVersion`. */
+const versionSeed = (): string =>
+    String(Math.floor((Math.random() * Number.MAX_SAFE_INTEGER) / 2));
+
 type ScriptRunner = {
     creditHoldTake(...args: string[]): Promise<number>;
     creditHoldSum(...args: string[]): Promise<[string, number]>;
+    creditHoldReserve(...args: string[]): Promise<[string, number]>;
     zrem(key: string, member: string): Promise<number>;
+    creditHoldBump(...args: string[]): Promise<number>;
+    get(key: string): Promise<string | null>;
 };
 
 // -- CreditHoldStore --------------------------------------------------
@@ -135,6 +206,14 @@ export class CreditHoldStore extends PuterStore {
             numberOfKeys: 1,
             lua: SUM_SCRIPT,
         });
+        client.defineCommand('creditHoldReserve', {
+            numberOfKeys: 2,
+            lua: RESERVE_SCRIPT,
+        });
+        client.defineCommand('creditHoldBump', {
+            numberOfKeys: 1,
+            lua: BUMP_SCRIPT,
+        });
     }
 
     /**
@@ -195,7 +274,92 @@ export class CreditHoldStore extends PuterStore {
         }
     }
 
-    /** Give a hold back. Safe to call twice; the second call is a no-op. */
+    /**
+     * The actor's balance version, to pass to `reserve` once the balance has
+     * been read. Null when the cache can't answer.
+     */
+    async version(userId: string): Promise<string | null> {
+        try {
+            return (await this.#redis.get(versionKey(userId))) ?? '0';
+        } catch (e) {
+            console.warn(
+                `[metering] credit hold version unreadable for ${userId}: ${(e as Error).message}`,
+            );
+            return null;
+        }
+    }
+
+    /**
+     * Commit `amount` only if it fits in `budget` beside the holds already
+     * live, as one step. `budget` is the gross balance read after `version`;
+     * pass `version: ''` to skip the staleness check.
+     */
+    async reserve(
+        userId: string,
+        amount: number,
+        budget: number,
+        version: string,
+        ttlMs: number = DEFAULT_HOLD_TTL_MS,
+    ): Promise<CreditReservation> {
+        // Nothing to hold, and a negative hold would lower everyone else's.
+        if (!userId || !Number.isFinite(amount) || amount <= 0) {
+            return { outcome: 'insufficient' };
+        }
+        const units = Math.ceil(amount);
+        const member = `${randomUUID()}:${units}`;
+        const now = Date.now();
+        try {
+            const [outcome, dropped] = await this.#redis.creditHoldReserve(
+                holdsKey(userId),
+                versionKey(userId),
+                member,
+                String(now + ttlMs),
+                String(ttlMs + KEY_TTL_SLACK_MS),
+                String(now),
+                String(units),
+                String(Math.floor(Math.min(budget, Number.MAX_SAFE_INTEGER))),
+                version,
+            );
+            if (dropped > 0) expiredHoldsCounter.add(dropped);
+            if (outcome === 'accepted') return { outcome, member };
+            if (outcome === 'insufficient' || outcome === 'stale') {
+                return { outcome };
+            }
+            throw new Error(`unexpected reservation outcome ${outcome}`);
+        } catch (e) {
+            console.warn(
+                `[metering] credit hold not reserved for ${userId}: ${(e as Error).message}`,
+            );
+            return { outcome: 'unavailable' };
+        }
+    }
+
+    /**
+     * Marks the actor's balance as changed, so a reservation that read it
+     * before the change reads it again. Call once a usage write has landed.
+     */
+    async bump(userId: string): Promise<void> {
+        if (!userId) return;
+        try {
+            await this.#redis.creditHoldBump(
+                versionKey(userId),
+                String(VERSION_TTL_MS),
+                versionSeed(),
+            );
+        } catch (e) {
+            console.warn(
+                `[metering] credit hold version not bumped for ${userId}: ${(e as Error).message}`,
+            );
+        }
+    }
+
+    /**
+     * Give a hold back. Safe to call twice; the second call releases nothing.
+     *
+     * The version isn't bumped here: the usage a hold stands for bumps it when
+     * it's recorded, before the hold is released, and a release with no usage
+     * leaves the balance as it was.
+     */
     async release(userId: string, member: string | null): Promise<void> {
         if (!userId || !member) return;
         try {

src/backend/stores/metering/MeteringBufferStore.test.ts

@@ -420,6 +420,625 @@ describe('MeteringBufferStore', () => {
             expect(await storedTotal(key)).toBe(0);
         });
 
+        /** Holds `method` on `target` until the returned release is called. */
+        const holdCalls = <T extends object>(on: T, method: keyof T) => {
+            let release: () => void = () => {};
+            const gate = new Promise<void>((r) => (release = r));
+            const real = (on[method] as (...a: unknown[]) => unknown).bind(on);
+            const spy = vi
+                .spyOn(on, method as never)
+                .mockImplementation((async (...args: unknown[]) => {
+                    await gate;
+                    return real(...args);
+                }) as never);
+            return { spy, release };
+        };
+        const readTotal = async () =>
+            ((await target.get({ key })).res as { total?: number } | null)
+                ?.total ?? 0;
+
+        it('counts usage a flush has claimed but not yet written onward', async () => {
+            await target.incr({ key, pathAndAmountMap: { total: 10 } });
+            const write = holdCalls(kv, 'incr');
+            try {
+                const flushing = target.flushCycle();
+                await vi.waitFor(() => expect(write.spy).toHaveBeenCalled());
+                expect(await readTotal()).toBe(10);
+                write.release();
+                await flushing;
+            } finally {
+                write.spy.mockRestore();
+            }
+            expect(await readTotal()).toBe(10);
+            expect(await storedTotal(key)).toBe(10);
+        });
+
+        it('counts usage written onward until its settle lands', async () => {
+            await target.incr({ key, pathAndAmountMap: { total: 10 } });
+            await target.flushCycle();
+            await target.incr({ key, pathAndAmountMap: { total: 5 } });
+            const redis = server.clients.redis as unknown as {
+                meterSettle: (...args: string[]) => Promise<number>;
+            };
+            const settle = holdCalls(redis, 'meterSettle');
+            try {
+                const flushing = target.flushCycle();
+                await vi.waitFor(() => expect(settle.spy).toHaveBeenCalled());
+                expect(await storedTotal(key)).toBe(15);
+                expect(await readTotal()).toBe(15);
+                settle.release();
+                await flushing;
+            } finally {
+                settle.spy.mockRestore();
+            }
+            expect(await readTotal()).toBe(15);
+        });
+
+        it('counts a claim in array reads too', async () => {
+            await target.incr({ key, pathAndAmountMap: { total: 4 } });
+            const write = holdCalls(kv, 'incr');
+            try {
+                const flushing = target.flushCycle();
+                await vi.waitFor(() => expect(write.spy).toHaveBeenCalled());
+                const { res } = await target.get({ key: [key] });
+                expect((res as [{ total: number }])[0].total).toBe(4);
+                write.release();
+                await flushing;
+            } finally {
+                write.spy.mockRestore();
+            }
+        });
+
+        it('counts a claim in the running total incr returns', async () => {
+            await target.incr({ key, pathAndAmountMap: { total: 10 } });
+            const write = holdCalls(kv, 'incr');
+            try {
+                const flushing = target.flushCycle();
+                await vi.waitFor(() => expect(write.spy).toHaveBeenCalled());
+                const { res } = await target.incr({
+                    key,
+                    pathAndAmountMap: { total: 5 },
+                });
+                expect((res as { total: number }).total).toBe(15);
+                write.release();
+                await flushing;
+            } finally {
+                write.spy.mockRestore();
+            }
+        });
+
+        it('counts another flush’s claim in an exact read', async () => {
+            await target.incr({ key, pathAndAmountMap: { total: 10 } });
+            const write = holdCalls(kv, 'incr');
+            try {
+                const flushing = target.flushCycle();
+                await vi.waitFor(() => expect(write.spy).toHaveBeenCalled());
+                const { res } = await target.readExact({ key });
+                expect((res as { total: number }).total).toBe(10);
+                write.release();
+                await flushing;
+            } finally {
+                write.spy.mockRestore();
+            }
+            const { res } = await target.readExact({ key });
+            expect((res as { total: number }).total).toBe(10);
+        });
+
+        it('misses nothing in an exact read when a flush lands between its reads', async () => {
+            await target.incr({ key, pathAndAmountMap: { total: 10 } });
+            const write = holdCalls(kv, 'incr');
+            const flushing = target.flushCycle();
+            await vi.waitFor(() => expect(write.spy).toHaveBeenCalled());
+
+            // The store read sees the counter from before the flush's write;
+            // the flush then writes it and leaves its claim before any claim
+            // read that follows the store read.
+            let storeReadStarted = false;
+            let flushed!: Promise<void>;
+            const realGet = kv.get.bind(kv);
+            const get = vi.spyOn(kv, 'get').mockImplementation(async (args) => {
+                if (!args.consistentRead || args.key !== key) {
+                    return realGet(args);
+                }
+                if (storeReadStarted) return realGet(args);
+                storeReadStarted = true;
+                const before = await realGet(args);
+                write.release();
+                flushed = flushing.then(() => {});
+                await flushed;
+                return before;
+            });
+            const redis = server.clients.redis as unknown as {
+                meterRead: (...args: string[]) => Promise<unknown>;
+            };
+            const realRead = redis.meterRead.bind(redis);
+            const read = vi
+                .spyOn(redis, 'meterRead')
+                .mockImplementation(async (...args: string[]) => {
+                    if (storeReadStarted) {
+                        await vi.waitFor(() => expect(flushed).toBeDefined());
+                        await flushed;
+                    }
+                    return realRead(...args);
+                });
+            try {
+                const { res } = await target.readExact({ key });
+                expect((res as { total: number }).total).toBe(10);
+            } finally {
+                get.mockRestore();
+                read.mockRestore();
+                write.spy.mockRestore();
+                await flushing;
+            }
+            expect(await storedTotal(key)).toBe(10);
+        });
+
+        it('counts nothing twice in an exact read when a flush lands during its store read', async () => {
+            await target.incr({ key, pathAndAmountMap: { total: 10 } });
+            const write = holdCalls(kv, 'incr');
+            const flushing = target.flushCycle();
+            await vi.waitFor(() => expect(write.spy).toHaveBeenCalled());
+
+            // The claim was read with the amount still on it; the store read
+            // then sees it written.
+            let landed = false;
+            const realGet = kv.get.bind(kv);
+            const get = vi.spyOn(kv, 'get').mockImplementation(async (args) => {
+                if (args.consistentRead && args.key === key && !landed) {
+                    landed = true;
+                    write.release();
+                    await flushing;
+                }
+                return realGet(args);
+            });
+            try {
+                const { res } = await target.readExact({ key });
+                expect((res as { total: number }).total).toBe(10);
+            } finally {
+                get.mockRestore();
+                write.spy.mockRestore();
+                await flushing;
+            }
+        });
+
+        it('counts usage behind another flush’s claim in an exact read', async () => {
+            await target.incr({ key, pathAndAmountMap: { total: 10 } });
+            const write = holdCalls(kv, 'incr');
+            try {
+                const flushing = target.flushCycle();
+                await vi.waitFor(() => expect(write.spy).toHaveBeenCalled());
+                // Arrives while the claim is out, so it can't be flushed yet.
+                await target.incr({ key, pathAndAmountMap: { total: 5 } });
+                const { res } = await target.readExact({ key });
+                expect((res as { total: number }).total).toBe(15);
+                write.release();
+                await flushing;
+            } finally {
+                write.spy.mockRestore();
+            }
+        });
+
+        /**
+         * Lets a held flush finish during the first consistent store read of
+         * `key`, which then answers with what the store holds afterwards.
+         */
+        const landDuringStoreRead = (
+            release: () => void,
+            flushing: Promise<unknown>,
+        ) => {
+            let landed = false;
+            const realGet = kv.get.bind(kv);
+            return vi.spyOn(kv, 'get').mockImplementation(async (args) => {
+                const ours = Array.isArray(args.key)
+                    ? args.key.includes(key)
+                    : args.key === key;
+                if (args.consistentRead && ours && !landed) {
+                    landed = true;
+                    release();
+                    await flushing;
+                }
+                return realGet(args);
+            });
+        };
+
+        for (const shape of ['single', 'array'] as const) {
+            it(`counts nothing twice in ${shape === 'single' ? 'a single' : 'an array'} read without a base when a flush lands during it`, async () => {
+                await target.incr({ key, pathAndAmountMap: { total: 10 } });
+                // No cached base: the counter is new to the store.
+                await target.forgetBase(key);
+                const write = holdCalls(kv, 'incr');
+                const flushing = target.flushCycle();
+                await vi.waitFor(() => expect(write.spy).toHaveBeenCalled());
+                const get = landDuringStoreRead(write.release, flushing);
+                try {
+                    const { res } = await target.get({
+                        key: shape === 'single' ? key : [key],
+                    });
+                    const value = (
+                        shape === 'single' ? res : (res as unknown[])[0]
+                    ) as { total: number };
+                    expect(value.total).toBe(10);
+                } finally {
+                    get.mockRestore();
+                    write.spy.mockRestore();
+                    await flushing;
+                }
+            });
+        }
+
+        it('reads a written-through increment once the cache is back', async () => {
+            await target.incr({ key, pathAndAmountMap: { total: 10 } });
+            await target.flushCycle();
+            const redis = server.clients.redis as unknown as {
+                meterIncr: (...args: string[]) => Promise<unknown>;
+            };
+            const failing = vi
+                .spyOn(redis, 'meterIncr')
+                .mockRejectedValueOnce(new Error('cache down'));
+            const { res, exact } = await target.incr({
+                key,
+                pathAndAmountMap: { total: 5 },
+            });
+            failing.mockRestore();
+            expect(exact).toBe(true);
+            expect((res as { total: number }).total).toBe(15);
+            expect(await readTotal()).toBe(15);
+        });
+
+        it('reads the store consistently when no base is cached', async () => {
+            await kv.incr({ key, pathAndAmountMap: { total: 3 } });
+            const get = vi.spyOn(kv, 'get');
+            try {
+                expect(await readTotal()).toBe(3);
+                expect(get).toHaveBeenCalledWith(
+                    expect.objectContaining({ key, consistentRead: true }),
+                );
+            } finally {
+                get.mockRestore();
+            }
+        });
+
+        for (const shape of ['single', 'array', 'exact'] as const) {
+            it(`does not retry a ${shape} store read when only new increments arrive`, async () => {
+                await kv.incr({ key, pathAndAmountMap: { total: 100 } });
+                await target.incrAux({ key, pathAndAmountMap: { total: 10 } });
+                const dynamo = server.clients.dynamo;
+                const realGet = dynamo.get.bind(dynamo);
+                const realBatchGet = dynamo.batchGet.bind(dynamo);
+                let storeReads = 0;
+                const increment = async () => {
+                    storeReads++;
+                    await target.incrAux({
+                        key,
+                        pathAndAmountMap: { total: 5 },
+                    });
+                };
+                const get = vi
+                    .spyOn(dynamo, 'get')
+                    .mockImplementation(async (...args) => {
+                        const result = await realGet(...args);
+                        if (args[1].key === key) await increment();
+                        return result;
+                    });
+                const batchGet = vi
+                    .spyOn(dynamo, 'batchGet')
+                    .mockImplementation(async (...args) => {
+                        const result = await realBatchGet(...args);
+                        if (args[0].some(({ items }) => items.key === key))
+                            await increment();
+                        return result;
+                    });
+                const redis = server.clients.redis as unknown as {
+                    meterRead: (...args: string[]) => Promise<unknown>;
+                };
+                const read = vi.spyOn(redis, 'meterRead');
+                try {
+                    const { res } =
+                        shape === 'exact'
+                            ? await target.readExact({ key })
+                            : await target.get({
+                                  key: shape === 'single' ? key : [key],
+                              });
+                    const value = (
+                        shape === 'array' ? (res as unknown[])[0] : res
+                    ) as { total: number };
+                    expect(value.total).toBe(110);
+                    expect(storeReads).toBe(1);
+                    expect(read).toHaveBeenCalledTimes(2);
+                } finally {
+                    get.mockRestore();
+                    batchGet.mockRestore();
+                    read.mockRestore();
+                }
+                expect(await readTotal()).toBe(115);
+            });
+
+            it(`retries a ${shape} read when new increments hide a completed flush`, async () => {
+                await kv.incr({ key, pathAndAmountMap: { total: 0 } });
+                await target.incrAux({ key, pathAndAmountMap: { total: 10 } });
+                const dynamo = server.clients.dynamo;
+                const write = holdCalls(dynamo, 'update');
+                const flushing = target.flushCycle();
+                await vi.waitFor(() => expect(write.spy).toHaveBeenCalled());
+                const realGet = dynamo.get.bind(dynamo);
+                const realBatchGet = dynamo.batchGet.bind(dynamo);
+                let storeReads = 0;
+                const land = async () => {
+                    if (++storeReads !== 1) return;
+                    write.release();
+                    await flushing;
+                    await target.incrAux({
+                        key,
+                        pathAndAmountMap: { total: 20 },
+                    });
+                };
+                const get = vi
+                    .spyOn(dynamo, 'get')
+                    .mockImplementation(async (...args) => {
+                        if (args[1].key === key) await land();
+                        return realGet(...args);
+                    });
+                const batchGet = vi
+                    .spyOn(dynamo, 'batchGet')
+                    .mockImplementation(async (...args) => {
+                        if (args[0].some(({ items }) => items.key === key))
+                            await land();
+                        return realBatchGet(...args);
+                    });
+                try {
+                    const { res } =
+                        shape === 'exact'
+                            ? await target.readExact({ key })
+                            : await target.get({
+                                  key: shape === 'single' ? key : [key],
+                              });
+                    const value = (
+                        shape === 'array' ? (res as unknown[])[0] : res
+                    ) as { total: number };
+                    expect(value.total).toBe(30);
+                    expect(storeReads).toBe(2);
+                } finally {
+                    get.mockRestore();
+                    batchGet.mockRestore();
+                    write.release();
+                    await flushing;
+                    write.spy.mockRestore();
+                }
+            });
+        }
+
+        it('keeps changed counters together in a batch retry', async () => {
+            const keys = [`${key}:a`, `${key}:b`];
+            await Promise.all(
+                keys.map((k) =>
+                    kv.incr({ key: k, pathAndAmountMap: { total: 0 } }),
+                ),
+            );
+            await Promise.all(
+                keys.map((k) =>
+                    target.incrAux({ key: k, pathAndAmountMap: { total: 10 } }),
+                ),
+            );
+            const dynamo = server.clients.dynamo;
+            const realBatchGet = dynamo.batchGet.bind(dynamo);
+            const write = holdCalls(dynamo, 'update');
+            const flushing = target.flushCycle();
+            await vi.waitFor(() => expect(write.spy).toHaveBeenCalledTimes(2));
+            const batches: string[][] = [];
+            const batchGet = vi
+                .spyOn(dynamo, 'batchGet')
+                .mockImplementation(async (...args) => {
+                    batches.push(
+                        args[0].map(({ items }) => items.key as string),
+                    );
+                    if (batches.length === 1) {
+                        write.release();
+                        await flushing;
+                    }
+                    return realBatchGet(...args);
+                });
+            const get = vi.spyOn(dynamo, 'get');
+            try {
+                expect((await target.get({ key: keys })).res).toEqual([
+                    { total: 10 },
+                    { total: 10 },
+                ]);
+                expect(batches).toHaveLength(2);
+                expect(batches.map((batch) => batch.sort())).toEqual([
+                    keys,
+                    keys,
+                ]);
+                expect(get).not.toHaveBeenCalled();
+            } finally {
+                batchGet.mockRestore();
+                get.mockRestore();
+                write.release();
+                await flushing;
+                write.spy.mockRestore();
+            }
+        });
+
+        it('detects a whole flush between snapshots with no outstanding claim', async () => {
+            await kv.incr({ key, pathAndAmountMap: { total: 0 } });
+            await target.incrAux({ key, pathAndAmountMap: { total: 10 } });
+            const dynamo = server.clients.dynamo;
+            const realGet = dynamo.get.bind(dynamo);
+            let reads = 0;
+            const get = vi
+                .spyOn(dynamo, 'get')
+                .mockImplementation(async (...args) => {
+                    if (args[1].key === key && ++reads === 1) {
+                        await target.flushCycle();
+                        await target.incrAux({
+                            key,
+                            pathAndAmountMap: { total: 20 },
+                        });
+                    }
+                    return realGet(...args);
+                });
+            try {
+                expect(await readTotal()).toBe(30);
+                expect(reads).toBe(2);
+            } finally {
+                get.mockRestore();
+            }
+        });
+
+        it('retries a decreasing buffer even when the flush metadata is unchanged', async () => {
+            await kv.incr({ key, pathAndAmountMap: { total: 100 } });
+            await target.incrAux({ key, pathAndAmountMap: { total: 10 } });
+            const dynamo = server.clients.dynamo;
+            const realGet = dynamo.get.bind(dynamo);
+            let reads = 0;
+            const get = vi
+                .spyOn(dynamo, 'get')
+                .mockImplementation(async (...args) => {
+                    const result = await realGet(...args);
+                    if (args[1].key === key && ++reads === 1) {
+                        await target.incrAux({
+                            key,
+                            pathAndAmountMap: { total: -5 },
+                        });
+                    }
+                    return result;
+                });
+            try {
+                expect(await readTotal()).toBe(105);
+                expect(reads).toBe(2);
+            } finally {
+                get.mockRestore();
+            }
+        });
+
+        it('detects applied paths while the claim id and fence token stay unchanged', async () => {
+            await kv.incr({ key, pathAndAmountMap: { total: 0 } });
+            await target.incrAux({ key, pathAndAmountMap: { total: 10 } });
+            const dynamo = server.clients.dynamo;
+            const redis = server.clients.redis as unknown as {
+                meterSettle: (...args: string[]) => Promise<number>;
+            };
+            const write = holdCalls(dynamo, 'update');
+            const settle = holdCalls(redis, 'meterSettle');
+            const flushing = target.flushCycle();
+            await vi.waitFor(() => expect(write.spy).toHaveBeenCalled());
+            const realGet = dynamo.get.bind(dynamo);
+            let reads = 0;
+            const get = vi
+                .spyOn(dynamo, 'get')
+                .mockImplementation(async (...args) => {
+                    if (args[1].key === key && ++reads === 1) {
+                        write.release();
+                        await vi.waitFor(() =>
+                            expect(settle.spy).toHaveBeenCalled(),
+                        );
+                        await target.incrAux({
+                            key,
+                            pathAndAmountMap: { total: 20 },
+                        });
+                    }
+                    return realGet(...args);
+                });
+            try {
+                expect(await readTotal()).toBe(30);
+                expect(reads).toBe(2);
+            } finally {
+                get.mockRestore();
+                write.release();
+                settle.release();
+                await flushing;
+                write.spy.mockRestore();
+                settle.spy.mockRestore();
+            }
+        });
+
+        it('does not retry unchanged amounts when another counter advances the bucket token', async () => {
+            await kv.incr({ key, pathAndAmountMap: { total: 100 } });
+            let other = `${key}:other`;
+            for (let i = 0; bucketTag(other) !== bucketTag(key); i++)
+                other = `${key}:${i}`;
+            await kv.incr({ key: other, pathAndAmountMap: { total: 0 } });
+            await target.incrAux({
+                key: other,
+                pathAndAmountMap: { total: 10 },
+            });
+            const dynamo = server.clients.dynamo;
+            const realGet = dynamo.get.bind(dynamo);
+            let reads = 0;
+            const get = vi
+                .spyOn(dynamo, 'get')
+                .mockImplementation(async (...args) => {
+                    if (args[1].key === key && ++reads === 1)
+                        await target.flushCycle();
+                    return realGet(...args);
+                });
+            try {
+                expect(await readTotal()).toBe(100);
+                expect(reads).toBe(1);
+            } finally {
+                get.mockRestore();
+            }
+        });
+
+        it('recovers a failed initial cache snapshot before retrying the batch', async () => {
+            await kv.incr({ key, pathAndAmountMap: { total: 100 } });
+            await target.incrAux({ key, pathAndAmountMap: { total: 10 } });
+            const redis = server.clients.redis as unknown as {
+                meterRead: (...args: string[]) => Promise<unknown>;
+            };
+            const read = vi
+                .spyOn(redis, 'meterRead')
+                .mockRejectedValueOnce(new Error('cache unavailable'));
+            const batchGet = vi.spyOn(server.clients.dynamo, 'batchGet');
+            try {
+                expect((await target.get({ key: [key] })).res).toEqual([
+                    { total: 110 },
+                ]);
+                expect(batchGet).toHaveBeenCalledTimes(2);
+            } finally {
+                read.mockRestore();
+                batchGet.mockRestore();
+            }
+        });
+
+        it('bounds concurrent cache reads while preserving array order', async () => {
+            const keys = Array.from({ length: 25 }, (_, i) => `${key}:${i}`);
+            await Promise.all(
+                keys.map((k, i) =>
+                    target.incrAux({
+                        key: k,
+                        pathAndAmountMap: { total: i + 1 },
+                    }),
+                ),
+            );
+            const redis = server.clients.redis as unknown as {
+                meterRead: (...args: string[]) => Promise<unknown>;
+            };
+            const realRead = redis.meterRead.bind(redis);
+            let active = 0;
+            let peak = 0;
+            const read = vi
+                .spyOn(redis, 'meterRead')
+                .mockImplementation(async (...args) => {
+                    active++;
+                    peak = Math.max(peak, active);
+                    try {
+                        await new Promise((resolve) => setTimeout(resolve, 1));
+                        return await realRead(...args);
+                    } finally {
+                        active--;
+                    }
+                });
+            try {
+                expect((await target.get({ key: keys })).res).toEqual(
+                    keys.map((_, i) => ({ total: i + 1 })),
+                );
+                expect(peak).toBe(20);
+                expect(active).toBe(0);
+            } finally {
+                read.mockRestore();
+            }
+        });
+
         it('reads an unknown counter as absent', async () => {
             const { res } = await target.get({ key });
             expect(res).toBeNull();
@@ -906,12 +1525,16 @@ describe('MeteringBufferStore', () => {
                 '1000000000',
             );
 
+            // Another deployment's write, which only a replacing settle would
+            // bring into the base.
+            await kv.incr({ key, pathAndAmountMap: { total: 1000 } });
             await target.incr({ key, pathAndAmountMap: { total: 5 } });
             await target.flushCycle();
 
-            expect(await storedTotal(key)).toBe(105);
+            expect(await storedTotal(key)).toBe(1105);
+            // The newer base keeps its view, plus what this flush wrote.
             expect(Number(await server.clients.redis.hget(base, 'total'))).toBe(
-                100,
+                105,
             );
         });
 
@@ -1438,6 +2061,30 @@ describe('MeteringBufferStore', () => {
             ).toEqual({});
         });
 
+        it('reads a re-driven write that already landed, with a warm base', async () => {
+            await target.incr({ key, pathAndAmountMap: { total: 100 } });
+            await target.flushCycle();
+            await target.incr({ key, pathAndAmountMap: { total: 5 } });
+
+            const passThrough = kv.incr.bind(kv);
+            const lost = vi
+                .spyOn(kv, 'incr')
+                .mockImplementationOnce(
+                    async (...args: Parameters<typeof kv.incr>) => {
+                        await passThrough(...args);
+                        throw new Error('request timed out');
+                    },
+                );
+            await target.flushCycle();
+            lost.mockRestore();
+            await ageClaim();
+            await target.flushCycle();
+
+            expect(await storedTotal(key)).toBe(105);
+            const { res } = await target.get({ key });
+            expect((res as { total: number }).total).toBe(105);
+        });
+
         it('applies a re-driven write whose first copy never landed', async () => {
             await target.incr({ key, pathAndAmountMap: { total: 5 } });
             const boom = vi
@@ -1742,6 +2389,17 @@ describe('MeteringBufferStore', () => {
                     order.push('dropFromClaim');
                     return realHdel(...args);
                 });
+            // A written batch leaves the claim through this script.
+            const redis = server.clients.redis as unknown as {
+                meterApply: (...args: string[]) => Promise<number>;
+            };
+            const realApply = redis.meterApply.bind(redis);
+            const applySpy = vi
+                .spyOn(redis, 'meterApply')
+                .mockImplementation((...args: string[]) => {
+                    order.push('dropFromClaim');
+                    return realApply(...args);
+                });
             const alarmSpy = vi
                 .spyOn(server.clients.alarm, 'create')
                 .mockImplementation(() => {
@@ -1751,6 +2409,7 @@ describe('MeteringBufferStore', () => {
             await target.flushCycle();
             incrSpy.mockRestore();
             hdelSpy.mockRestore();
+            applySpy.mockRestore();
             alarmSpy.mockRestore();
 
             expect(order).toEqual(['dropFromClaim', 'alarm']);
@@ -1783,7 +2442,13 @@ describe('MeteringBufferStore', () => {
             // right up until the TTL takes the amounts with it.
             const tag = bucketTag(key);
             await server.clients.redis.del(`meter:dirty:{${tag}}`);
+            // Every few cycles a flush reconciles on its own; keep this one
+            // from doing so, whatever the earlier tests left the count at.
+            const scheduled = vi
+                .spyOn(target, 'reconcile')
+                .mockResolvedValue(undefined);
             expect(await target.flushCycle()).toBe(0);
+            scheduled.mockRestore();
             expect(await storedTotal(key)).toBe(0);
 
             const logged = vi

src/backend/stores/metering/MeteringBufferStore.ts

@@ -83,6 +83,19 @@ interface IncrInput {
 
 type FlatAmounts = Record<string, number>;
 
+interface BufferedRead {
+    delta: FlatAmounts;
+    base: FlatAmounts;
+    claim: FlatAmounts;
+    claimId: string;
+    writeToken: string;
+}
+
+interface StoreAndBufferRead {
+    stored: unknown;
+    buffered: FlatAmounts;
+}
+
 /** A claim's write that was recorded before it was sent; see `INTENT_SCRIPT`. */
 interface SettleIntent {
     writer: string;
@@ -126,8 +139,8 @@ const ORPHAN_AGE_MS = 30_000;
 /** Long enough that a counter for the current month can never expire. */
 const BUFFER_TTL_MS = 40 * 24 * 60 * 60 * 1000;
 
-/** How many counters are written onward at once, to keep the load even. */
-const SETTLE_CONCURRENCY = 20;
+/** Cache reads per call or counter writes per cycle in flight at once. */
+const BUFFER_IO_CONCURRENCY = 20;
 
 /** Roughly a minute between compression reports, so they don't flood the log. */
 const CYCLES_PER_COMPRESSION_REPORT = 12;
@@ -152,6 +165,12 @@ const UNWRITABLE_PATH_MEMO_LIMIT = 1000;
  */
 const EXACT_READ_MEMO_LIMIT = 10_000;
 
+/**
+ * Store reads a consistent read makes while the buffer keeps shrinking under
+ * it; see `#readStoreAndBuffer`.
+ */
+const EXACT_READ_ATTEMPTS = 3;
+
 /** Write units past which settling a counter raises the large-item alarm. */
 const LARGE_SETTLE_WRITE_UNITS = 100;
 
@@ -339,6 +358,15 @@ const toScriptArgs = (amounts: Record<string, number>): string[] => {
     return args;
 };
 
+/** Whether two sets of amounts hold the same paths and values. */
+const sameAmounts = (a: FlatAmounts, b: FlatAmounts): boolean => {
+    const paths = Object.keys(a);
+    return (
+        paths.length === Object.keys(b).length &&
+        paths.every((path) => a[path] === b[path])
+    );
+};
+
 /** The subset of `amounts` at `paths`. */
 const pickAmounts = (amounts: FlatAmounts, paths: string[]): FlatAmounts =>
     Object.fromEntries(paths.map((path) => [path, amounts[path]!]));
@@ -380,11 +408,12 @@ export const isBilledCounter = (key: string): boolean =>
 // -- Scripts ----------------------------------------------------------
 
 /**
- * KEYS: delta, base, dirty set, tracked set. ARGV: ttl, member, then
- * path/amount pairs.
+ * KEYS: delta, base, dirty set, tracked set, claimed. ARGV: ttl, member,
+ * pending key prefix, then path/amount pairs.
  *
- * Returns the delta and the base together so a running total can be served
- * without a second round trip. Amounts are floats — costs are fractional.
+ * Returns the delta, the base and any outstanding claim together, so a running
+ * total can be served without a second round trip (see `READ_SCRIPT` for the
+ * claim). Amounts are floats — costs are fractional.
  *
  * The counter joins both sets here. The dirty set is the working list a flush
  * takes from and hands back; the tracked set is only ever added to here and
@@ -392,7 +421,7 @@ export const isBilledCounter = (key: string): boolean =>
  * working list does not.
  */
 const INCR_SCRIPT = `
-for i = 3, #ARGV, 2 do
+for i = 4, #ARGV, 2 do
     redis.call('HINCRBYFLOAT', KEYS[1], ARGV[i], ARGV[i + 1])
 end
 redis.call('PEXPIRE', KEYS[1], ARGV[1])
@@ -400,12 +429,48 @@ redis.call('SADD', KEYS[3], ARGV[2])
 redis.call('PEXPIRE', KEYS[3], ARGV[1])
 redis.call('SADD', KEYS[4], ARGV[2])
 redis.call('PEXPIRE', KEYS[4], ARGV[1])
-return { redis.call('HGETALL', KEYS[1]), redis.call('HGETALL', KEYS[2]) }
+local claim = {}
+local nonce = redis.call('HGET', KEYS[5], ARGV[2])
+if nonce then claim = redis.call('HGETALL', ARGV[3] .. nonce) end
+return { redis.call('HGETALL', KEYS[1]), redis.call('HGETALL', KEYS[2]), claim }
 `;
 
-/** KEYS: delta, base. Reads both without marking the counter for flushing. */
+/**
+ * KEYS: delta, base, claimed, fence token. ARGV: counter, pending key prefix.
+ *
+ * Reads both without marking the counter for flushing, plus the counter's
+ * outstanding claim: amounts a flush has taken but not yet written onward are
+ * in neither the delta nor the base. The claim's key is built here, which is
+ * safe for the reason given at `RECONCILE_SCRIPT`.
+ */
 const READ_SCRIPT = `
-return { redis.call('HGETALL', KEYS[1]), redis.call('HGETALL', KEYS[2]) }
+local claim = {}
+local nonce = redis.call('HGET', KEYS[3], ARGV[1])
+if nonce then claim = redis.call('HGETALL', ARGV[2] .. nonce) end
+return {
+    redis.call('HGETALL', KEYS[1]), redis.call('HGETALL', KEYS[2]), claim,
+    nonce or '', redis.call('GET', KEYS[4]) or ''
+}
+`;
+
+/**
+ * KEYS: pending, base. ARGV: path/amount pairs just written onward.
+ *
+ * Takes written paths off a claim and, when a base is cached, adds them to it
+ * in the same step, so a read counts them all the way to the settle that
+ * replaces the base with the store's own view. Without a base, reads go to the
+ * store, which already has them.
+ */
+const APPLY_SCRIPT = `
+redis.call('HDEL', KEYS[1], '${INTENT_FIELD}')
+local based = redis.call('EXISTS', KEYS[2]) == 1
+for i = 1, #ARGV, 2 do
+    redis.call('HDEL', KEYS[1], ARGV[i])
+    if based then
+        redis.call('HINCRBYFLOAT', KEYS[2], ARGV[i], ARGV[i + 1])
+    end
+end
+return 1
 `;
 
 /**
@@ -651,9 +716,29 @@ end
 return redis.call('HGETALL', KEYS[1])
 `;
 
+/**
+ * KEYS: pending, base, base sequence. ARGV: paths found already applied.
+ *
+ * Takes paths an earlier copy of the write already applied off a claim, and
+ * drops the cached base in the same step: the base may or may not have them,
+ * and the store does, so the next read seeds a fresh one from the store.
+ */
+const FENCED_SCRIPT = `
+redis.call('HDEL', KEYS[1], '${INTENT_FIELD}')
+for i = 1, #ARGV do
+    redis.call('HDEL', KEYS[1], ARGV[i])
+end
+redis.call('DEL', KEYS[2], KEYS[3])
+return 1
+`;
+
 type ScriptRunner = {
-    meterIncr(...args: string[]): Promise<[string[], string[]]>;
-    meterRead(...args: string[]): Promise<[string[], string[]]>;
+    meterIncr(...args: string[]): Promise<[string[], string[], string[]]>;
+    meterRead(
+        ...args: string[]
+    ): Promise<[string[], string[], string[], string, string]>;
+    meterApply(...args: string[]): Promise<number>;
+    meterFenced(...args: string[]): Promise<number>;
     meterDrain(...args: string[]): Promise<[string[], number]>;
     meterClaim(...args: string[]): Promise<string[] | null>;
     /** `[1, amounts]` when re-claimed, `[0, hadIndexEntry]` when not. */
@@ -770,6 +855,8 @@ export class MeteringBufferStore extends PuterStore {
                 `[metering] buffered incr failed, writing through: ${(e as Error).message}`,
             );
             const { res } = await this.#writeThrough(input);
+            // The cached base doesn't have this write; the next read reseeds it.
+            await this.forgetBase(input.key);
             return { res, exact: true };
         }
     }
@@ -817,43 +904,40 @@ export class MeteringBufferStore extends PuterStore {
 
         const keys = key;
         const results: (unknown | null)[] = new Array(keys.length).fill(null);
-        const deltas: FlatAmounts[] = new Array(keys.length).fill(
-            {} as FlatAmounts,
+        const snapshots: (BufferedRead | null)[] = new Array(keys.length).fill(
+            null,
         );
         const misses: number[] = [];
 
-        await Promise.all(
-            keys.map(async (k, i) => {
-                try {
-                    const { delta, base } = await this.#readBuffered(k);
-                    if (Object.keys(base).length > 0) {
-                        results[i] = mergedOrNull(base, delta);
-                        return;
-                    }
-                    // No cached base yet — the store is the only place left to
-                    // ask, and it's cheaper to ask once for every key like it.
-                    deltas[i] = delta;
-                    misses.push(i);
-                } catch (e) {
-                    console.warn(
-                        `[metering] buffered read failed, reading through: ${(e as Error).message}`,
-                    );
-                    misses.push(i);
-                }
-            }),
+        const reads = await runWithConcurrencyLimitSettled(
+            keys,
+            BUFFER_IO_CONCURRENCY,
+            (k) => this.#readBuffered(k),
         );
+        reads.forEach((read, i) => {
+            if (read.status === 'fulfilled') {
+                const { delta, base } = read.value;
+                if (Object.keys(base).length > 0) {
+                    results[i] = mergedOrNull(base, delta);
+                    return;
+                }
+                snapshots[i] = read.value;
+            } else {
+                console.warn(
+                    `[metering] buffered read failed, reading through: ${(read.reason as Error).message}`,
+                );
+            }
+            misses.push(i);
+        });
 
         if (misses.length > 0) {
-            // Unlike the per-key fallbacks above, a failure here is not
-            // absorbed: a caller deciding on a balance must know its answer
-            // is incomplete rather than silently read a buffered-only view.
-            const { res: fetched } = await this.stores.kv.get({
-                key: misses.map((i) => keys[i]!),
-            });
-            const fetchedValues = fetched as (unknown | null)[];
+            const fetched = await this.#readStoreAndBuffer(
+                misses.map((i) => keys[i]!),
+                misses.map((i) => snapshots[i]!),
+            );
             misses.forEach((i, idx) => {
-                const persisted = flattenAmounts(fetchedValues[idx]);
-                results[i] = mergedOrNull(persisted, deltas[i]!);
+                const { stored, buffered } = fetched[idx]!;
+                results[i] = mergedOrNull(flattenAmounts(stored), buffered);
             });
         }
 
@@ -897,7 +981,11 @@ export class MeteringBufferStore extends PuterStore {
                 `[metering] exact read could not flush ${key}: ${(e as Error).message}`,
             );
         }
-        return this.stores.kv.get({ key, consistentRead: true });
+        // Whatever is still buffered, a claim another deployment holds and
+        // usage that arrived behind it included, isn't in the store yet.
+        const [{ stored, buffered }] = await this.#readStoreAndBuffer(key);
+        if (Object.keys(buffered).length === 0) return { res: stored };
+        return { res: mergedOrNull(flattenAmounts(stored), buffered) };
     }
 
     #rememberExactRead(key: string, now: number): void {
@@ -1115,13 +1203,21 @@ export class MeteringBufferStore extends PuterStore {
         this.#definedScripts = true;
         const client = this.clients.redis;
         client.defineCommand('meterIncr', {
-            numberOfKeys: 4,
+            numberOfKeys: 5,
             lua: INCR_SCRIPT,
         });
         client.defineCommand('meterRead', {
-            numberOfKeys: 2,
+            numberOfKeys: 4,
             lua: READ_SCRIPT,
         });
+        client.defineCommand('meterApply', {
+            numberOfKeys: 2,
+            lua: APPLY_SCRIPT,
+        });
+        client.defineCommand('meterFenced', {
+            numberOfKeys: 3,
+            lua: FENCED_SCRIPT,
+        });
         client.defineCommand('meterDrain', {
             numberOfKeys: 2,
             lua: DRAIN_SCRIPT,
@@ -1172,18 +1268,23 @@ export class MeteringBufferStore extends PuterStore {
         pathAndAmountMap,
     }: IncrInput): Promise<{ delta: FlatAmounts; base: FlatAmounts }> {
         const tag = bucketTag(key);
-        const [delta, base] = await this.#redis.meterIncr(
+        const [delta, base, claim] = await this.#redis.meterIncr(
             deltaKey(tag, key),
             baseKey(tag, key),
             dirtyKey(tag),
             trackedKey(tag),
+            claimedKey(tag),
             String(BUFFER_TTL_MS),
             key,
+            pendingKey(tag, ''),
             ...toScriptArgs(pathAndAmountMap),
         );
         this.#absorbedCount++;
         absorbedCounter.add(1);
-        return { delta: pairsToAmounts(delta), base: pairsToAmounts(base) };
+        return {
+            delta: addAmounts(pairsToAmounts(delta), parseClaim(claim).amounts),
+            base: pairsToAmounts(base),
+        };
     }
 
     /**
@@ -1196,7 +1297,9 @@ export class MeteringBufferStore extends PuterStore {
      * is over within seconds.
      */
     async #seedBase(key: string): Promise<FlatAmounts> {
-        const { res } = await this.stores.kv.get({ key });
+        // Consistent, or a write that just landed could be missing from the
+        // base until the next settle replaces it.
+        const { res } = await this.stores.kv.get({ key, consistentRead: true });
         const persisted = flattenAmounts(res);
         const seeded = await this.#redis.meterSeed(
             baseKey(bucketTag(key), key),
@@ -1212,25 +1315,107 @@ export class MeteringBufferStore extends PuterStore {
      * This deployment's cached delta and base for a counter, straight from
      * redis.
      */
-    async #readBuffered(
-        key: string,
-    ): Promise<{ delta: FlatAmounts; base: FlatAmounts }> {
+    async #readBuffered(key: string): Promise<BufferedRead> {
         const tag = bucketTag(key);
-        const [delta, base] = await this.#redis.meterRead(
-            deltaKey(tag, key),
-            baseKey(tag, key),
-        );
-        return { delta: pairsToAmounts(delta), base: pairsToAmounts(base) };
+        const [delta, base, claim, claimId, writeToken] =
+            await this.#redis.meterRead(
+                deltaKey(tag, key),
+                baseKey(tag, key),
+                claimedKey(tag),
+                tokenKey(tag),
+                key,
+                pendingKey(tag, ''),
+            );
+        const claimed = parseClaim(claim).amounts;
+        // Claimed amounts count like buffered ones until their settle.
+        return {
+            delta: addAmounts(pairsToAmounts(delta), claimed),
+            base: pairsToAmounts(base),
+            claim: claimed,
+            claimId,
+            writeToken,
+        };
     }
 
     async #readThrough(key: string): Promise<unknown | null> {
-        const { delta, base } = await this.#readBuffered(key);
-        const resolved =
-            Object.keys(base).length > 0
-                ? base
-                : flattenAmounts((await this.stores.kv.get({ key })).res);
+        const snapshot = await this.#readBuffered(key);
+        const { delta, base } = snapshot;
+        if (Object.keys(base).length > 0) return mergedOrNull(base, delta);
+        const [{ stored, buffered }] = await this.#readStoreAndBuffer(key, [
+            snapshot,
+        ]);
+        return mergedOrNull(flattenAmounts(stored), buffered);
+    }
 
-        return mergedOrNull(resolved, delta);
+    /**
+     * Order buffer snapshots around consistent store reads to avoid losing
+     * flushed usage. Retry changed claims together; new increments alone do not
+     * move amounts to the store and do not invalidate the first snapshot.
+     */
+    async #readStoreAndBuffer(
+        key: string | string[],
+        snapshots?: (BufferedRead | null)[],
+    ): Promise<StoreAndBufferRead[]> {
+        const multi = Array.isArray(key);
+        const keys = multi ? key : [key];
+        const initial = snapshots ?? [await this.#readBuffered(keys[0]!)];
+        let pending = keys.map((k, i) => ({
+            key: k,
+            index: i,
+            before: initial[i] ?? null,
+        }));
+        const results: StoreAndBufferRead[] = new Array(keys.length);
+        for (let attempt = 1; pending.length > 0; attempt++) {
+            const { res } = await this.stores.kv.get({
+                key: multi ? pending.map((item) => item.key) : pending[0]!.key,
+                consistentRead: true,
+            });
+            const stored = multi ? (res as unknown[]) : [res];
+            const checks =
+                attempt < EXACT_READ_ATTEMPTS
+                    ? await runWithConcurrencyLimitSettled(
+                          pending,
+                          BUFFER_IO_CONCURRENCY,
+                          (item) => this.#readBuffered(item.key),
+                      )
+                    : [];
+            const retry: typeof pending = [];
+            pending.forEach((item, i) => {
+                const check = checks[i];
+                if (check?.status === 'fulfilled') {
+                    const after = check.value;
+                    const before = item.before;
+                    // The token catches a whole claim starting and finishing
+                    // between snapshots, even when new increments hide it.
+                    const unchanged =
+                        before &&
+                        (sameAmounts(before.delta, after.delta) ||
+                            (before.writeToken === after.writeToken &&
+                                before.claimId === after.claimId &&
+                                sameAmounts(before.claim, after.claim) &&
+                                Object.entries(before.delta).every(
+                                    ([path, amount]) =>
+                                        (after.delta[path] ?? 0) >= amount,
+                                ) &&
+                                Object.entries(after.delta).every(
+                                    ([path, amount]) =>
+                                        amount >= (before.delta[path] ?? 0),
+                                )));
+                    if (!unchanged) {
+                        retry.push({ ...item, before: after });
+                        return;
+                    }
+                } else if (check?.status === 'rejected' && !multi) {
+                    throw check.reason;
+                }
+                results[item.index] = {
+                    stored: stored[i],
+                    buffered: item.before?.delta ?? {},
+                };
+            });
+            pending = retry;
+        }
+        return results;
     }
 
     // -- Internals: flush ---------------------------------------------
@@ -1302,7 +1487,7 @@ export class MeteringBufferStore extends PuterStore {
 
         const outcomes = await runWithConcurrencyLimitSettled(
             work,
-            SETTLE_CONCURRENCY,
+            BUFFER_IO_CONCURRENCY,
             (run) => run(),
         );
         for (const outcome of outcomes) {
@@ -1590,7 +1775,12 @@ export class MeteringBufferStore extends PuterStore {
                 if (err instanceof KvFencedWriteError) {
                     // Applied already, by an earlier copy of this write.
                     fencedCounter.add(1);
-                    await this.#dropFromClaim(tag, nonce, paths);
+                    await this.#redis.meterFenced(
+                        pendingKey(tag, nonce),
+                        baseKey(tag, key),
+                        baseSeqKey(tag, key),
+                        ...paths,
+                    );
                     continue;
                 }
                 if (!isPermanentSettleError(err)) throw e;
@@ -1618,7 +1808,11 @@ export class MeteringBufferStore extends PuterStore {
             // before anything else can fail: a re-drive then picks up only what
             // is still outstanding. The alarm below is the "anything else" — it
             // comes after.
-            await this.#dropFromClaim(tag, nonce, paths);
+            await this.#redis.meterApply(
+                pendingKey(tag, nonce),
+                baseKey(tag, key),
+                ...toScriptArgs(batch),
+            );
 
             if (usage.write >= LARGE_SETTLE_WRITE_UNITS) {
                 this.#recordLargeSettle(key, usage.write);

src/backend/testUtil.ts

@@ -154,6 +154,9 @@ export const setupTestServer = async (
             redis: { useMock: true },
             // Suites run system jobs themselves, not on the poll timer.
             jobs: { enabled: false },
+            // The health loop's timed `SELECT 1` would land in tests that
+            // count queries.
+            server_health: { disabled_checks: ['database-liveness'] },
             s3: { localConfig: { inMemory: true } },
             no_default_user: true,
             no_devwatch: true,

src/backend/types.ts

@@ -256,12 +256,7 @@ export interface IPreludeConfig {
      * an RCS agent provisioned in the Prelude account to actually use RCS.
      */
     preferredChannel?:
-        | 'sms'
-        | 'rcs'
-        | 'whatsapp'
-        | 'viber'
-        | 'zalo'
-        | 'telegram';
+        'sms' | 'rcs' | 'whatsapp' | 'viber' | 'zalo' | 'telegram';
 }
 
 /**
@@ -1106,7 +1101,10 @@ interface IConfigOptional {
     builtin_apps: Record<string, string>;
     /** Force the bundled GUI even in dev. Default: false. */
     use_bundled_gui: boolean;
-    /** Override the GUI bundle JS path. Default: `/dist/bundle.min.js`. */
+    /**
+     * Override the GUI bundle JS path. Default: `/dist/bundle.min.js`. The app
+     * browser embed loads `apps-embed.min.js` from the same directory.
+     */
     gui_bundle: string;
     /** Override the GUI CSS path when bundled. Default: `/dist/bundle.min.css`. */
     gui_css: string;
@@ -1393,7 +1391,8 @@ export interface ServerDrainInfo {
 }
 
 export interface WithCostsReporting extends WithLifecycle {
-    getReportedCosts?: () => // eslint-disable-next-line @typescript-eslint/no-explicit-any
+    getReportedCosts?: () =>
+        // eslint-disable-next-line @typescript-eslint/no-explicit-any
         | Promise<Record<string, any>[]>
         // eslint-disable-next-line @typescript-eslint/no-explicit-any
         | Record<string, any>[];

src/backend/vitest.config.ts

@@ -37,6 +37,9 @@ const repoRoot = path.resolve(backendDir, '../..');
 const isPgmockMode =
     (process.env.PUTER_TEST_DB_ENGINE ?? '').toLowerCase() === 'postgres';
 const pgmockTimeoutMs = 600_000;
+// Loaded CI runners stall timers well past the 5 s / 10 s defaults; a real
+// hang still fails, just later.
+const ciTimeouts = { testTimeout: 20_000, hookTimeout: 60_000 };
 
 // Tests that always boot pgmock regardless of PUTER_TEST_DB_ENGINE. Too slow
 // for CI; run locally or via `npm run test:backend:postgres`.
@@ -103,9 +106,11 @@ export const lowerDecoratorsPlugin = {
             loader: 'ts',
             target: 'es2024',
             sourcefile: id,
-            sourcemap: 'inline',
+            sourcemap: 'external',
         });
-        return { code: result.code, map: null };
+        // Without the map, coverage and stack traces report esbuild's
+        // output lines instead of the source's.
+        return { code: result.code, map: result.map };
     },
 };
 
@@ -145,7 +150,9 @@ export default defineConfig(({ mode }) => ({
                   hookTimeout: pgmockTimeoutMs,
                   fileParallelism: false,
               }
-            : {}),
+            : isCi
+              ? ciTimeouts
+              : {}),
         coverage: {
             provider: 'v8',
             reporter: isCi

src/docs/build.js

@@ -247,10 +247,10 @@ function generateTableOfContentsHTML (htmlContent, title) {
     html += '<nav class="toc-nav">';
 
     const cleanTitle = removeTags(title);
-    html += `<a href="#" class="toc-link toc-level-1" data-level="1">${cleanTitle}</a>`;
+    html += `<a href="#" class="toc-link toc-level-1" data-level="1">${breakAfterDots(cleanTitle)}</a>`;
 
     for ( const heading of headings ) {
-        html += `<a href="#${heading.slug}" class="toc-link toc-level-${heading.level}" data-level="${heading.level}">${heading.text}</a>`;
+        html += `<a href="#${heading.slug}" class="toc-link toc-level-${heading.level}" data-level="${heading.level}">${breakAfterDots(heading.text)}</a>`;
     }
 
     html += '</nav>';
@@ -457,10 +457,12 @@ function generateDocsHTML (filePath, rootDir, page, isIndex = false) {
     html += '</div>';
     // content
     html += `<div id="docs-content-${page.slug ?? ''}" class="docs-content col-xl-7 col-lg-7 col-md-12 col-sm-12 col-xs-12">`;
-    // context menu
+    // title, with the context menu beside it
+    const pageTitle = breakAfterDots(page.page_title ?? page.title);
+    html += '<div class="page-title-row">';
+    html += `<h1>${page.icon ? `<img src="/${baseURL}${page.icon}" style="opacity:0.5; width: 24px; height: 24px; margin-right: 10px;">` : '' }${pageTitle}</h1>`;
     html += generateMenuHTML();
-
-    html += `<h1>${page.icon ? `<img src="/${baseURL}${page.icon}" style="opacity:0.5; width: 24px; height: 24px; margin-right: 10px;">` : '' }${page.page_title ?? page.title}</h1>`;
+    html += '</div>';
     // Platform compatibility badges
     html += generatePlatformCompatibilityHTML(frontMatter);
     html += '<hr class="hr-inset">';
@@ -852,6 +854,12 @@ function removeTags (html) {
     return html.replace(/<[^>]*>?/gm, '');
 }
 
+// Let a long title wrap after a dot, so `puter.auth.getDetailedAppUsage()`
+// never splits inside the method name.
+function breakAfterDots (html) {
+    return html.replace(/\.(?=\w)/g, '.<wbr>');
+}
+
 // Replace '/path/to/root-directory' with the path to your root directory
 generateDocumentation('./src');
 generateRedirects();

src/docs/src/AI.md

@@ -194,18 +194,18 @@ Choose a model and compare provider rates in the [AI model directory](https://de
 These AI features are supported out of the box when using Puter.js:
 
 - **[`puter.ai.chat()`](/AI/chat/)** - Chat with AI models like Claude, GPT, and others
+- **[`puter.ai.txt2img()`](/AI/txt2img/)** - Generate images from text descriptions
+- **[`puter.ai.txt2speech()`](/AI/txt2speech/)** - Convert text to speech
+- **[`puter.ai.speech2txt()`](/AI/speech2txt/)** - Transcribe audio recordings into text
+- **[`puter.ai.img2txt()`](/AI/img2txt/)** - Extract text from images (OCR)
+- **[`puter.ai.txt2vid()`](/AI/txt2vid/)** - Generate short video clips from text or a reference image with Wan, Seedance, Veo, Kling, MiniMax H3 and other models, or edit a video you supply with Gemini Omni, FLUX 3 or Kling Omni
+- **[`puter.ai.speech2speech()`](/AI/speech2speech/)** - Convert speech in one voice to another voice
+- **[`puter.ai.convo()`](/AI/convo/)** - Prepare a beta realtime conversation with optional text, microphone, playback, and interruption helpers
+- **[Realtime WebSocket API](/AI/realtime/)** - Connect directly without puter.js. This API is in beta and may change abruptly.
 - **[`puter.ai.listModels()`](/AI/listModels/)** - List available AI chat models (and providers) that Puter currently exposes.
 - **[`puter.ai.listModelProviders()`](/AI/listModelProviders/)** - List the AI providers that Puter currently exposes.
-- **[`puter.ai.txt2img()`](/AI/txt2img/)** - Generate images from text descriptions
-- **[`puter.ai.img2txt()`](/AI/img2txt/)** - Extract text from images (OCR)
-- **[`puter.ai.txt2speech()`](/AI/txt2speech/)** - Convert text to speech
 - **[`puter.ai.txt2speech.listEngines()`](/AI/txt2speech.listEngines/)** - List available TTS engines/models
 - **[`puter.ai.txt2speech.listVoices()`](/AI/txt2speech.listVoices/)** - List available TTS voices
-- **[`puter.ai.speech2speech()`](/AI/speech2speech/)** - Convert speech in one voice to another voice
-- **[`puter.ai.txt2vid()`](/AI/txt2vid/)** - Generate short video clips from text or a reference image with Wan, Seedance, Veo, Kling, MiniMax H3 and other models, or edit a video you supply with Gemini Omni, FLUX 3 or Kling Omni
-- **[`puter.ai.speech2txt()`](/AI/speech2txt/)** - Transcribe audio recordings into text
-- **[Realtime WebSocket API](/AI/realtime/)** - Connect directly without puter.js. This API is in beta and may change abruptly.
-- **[`puter.ai.convo()`](/AI/convo/)** - Prepare a beta realtime conversation with optional text, microphone, playback, and interruption helpers
 
 ## Examples
 

src/docs/src/AI/chat.md

@@ -29,24 +29,24 @@ An object containing the following properties:
 
 | Option | Type | Description |
 |--------|------|-------------|
-| `model` | String | The model you want to use for the completion. If not specified, defaults to `gpt-5.4-nano`. More than 500 models are available from providers including OpenAI, Anthropic, Google, Alibaba Cloud, xAI, Mistral, OpenRouter, Infron, and others. For a full list, see the [AI models list](https://developer.puter.com/ai/models/) page. |
-| `stream` | Boolean | A boolean indicating whether you want to stream the completion. Defaults to `false`. |
-| `max_tokens` | Number | The maximum number of tokens to generate in the completion. By default, the specific model's maximum is used. |
-| `temperature` | Number | A number between 0 and 2 indicating the randomness of the completion. Lower values make the output more focused and deterministic, while higher values make it more random. By default, the model's own temperature is used. See [Model-specific features](#model-specific-features). |
-| `reasoning_effort` / `reasoning.effort` | String | Controls how much effort reasoning models spend thinking. Supported values: `none`, `minimal`, `low`, `medium`, `high`, `xhigh`, and `max`. Lower values give faster responses with less reasoning. OpenAI models, Moonshot's Kimi K3.1 Pro models, Meta's Muse Spark models, `gemini-4-argon`, `gemini-3.1-flash-lite` and `gemini-3-flash-preview` only. See [Model-specific features](#model-specific-features). |
-| `thinking` | Object \| null | Extended-thinking control: `{ type: "adaptive" \| "enabled" \| "disabled" \| "between_tools", budgetTokens?, display? }`. Pass `null` to send no thinking configuration at all (the model's own default); leaving it unset falls back to the legacy derivation from `reasoning_effort`. `budgetTokens` only applies to `type: "enabled"`. |
-| `verbosity` / `text.verbosity` | String | Controls how long or short responses are. Supported values: `low`, `medium`, and `high`. Lower values give shorter responses. OpenAI models only. |
-| `tools` | Array | Function definitions the AI can call. See [Function Calling](#function-calling) for details. |
-| `toolChoice` | String \| Object | Which tool, if any, the model must call. Pass `"auto"` (the model decides), `"none"` (never call a tool), `"required"` (call at least one), or an object naming a specific tool: `{ type: "tool", name: "..." }`. Requires `tools`. |
-| `parallelToolCalls` | Boolean | Whether the model may call more than one tool in a single turn. Left unset, tool calls run one at a time. |
-| `streamToolInput` | Boolean | Streaming only. When `true`, a tool call's arguments arrive incrementally as `tool_use_start` / `tool_input_delta` chunks instead of a single `tool_use` chunk once the call completes. |
-| `normalize` | Boolean | Controls the format of the non-streaming response. See [Response normalization](#response-normalization). |
-| `compaction` | Boolean \| Object | Opt into inline context compaction for long conversations. See [Compaction](#compaction). |
-| `signal` | AbortSignal | Abort the request, including while waiting for the next streamed chunk. Cancellation rejects with the signal reason (normally an `AbortError`). |
-| `provider` | String | Pin the request to a specific provider, for example `openrouter` or `infron`. Without it, Puter selects a provider for the requested model. Call [`puter.ai.listModelProviders()`](/AI/listModelProviders) for the available values, and [`puter.ai.listModels(provider)`](/AI/listModels) for the models a given provider serves. See [Providers](#providers). |
-| `fallback` | Boolean | Set to `false` to fail the request when the first provider tried fails, instead of retrying it on another provider that serves the same model. Defaults to `true`. See [Providers](#providers). |
-| `outputFormat` | Object | Constrains the response to a JSON schema: `{ type: "json_schema", schema, name?, strict? }`. |
-| `stopSequences` | Array of String | Up to the model's limit of strings; generation stops the moment one of them appears in the output. |
+| `model` | `String` | The model you want to use for the completion. If not specified, defaults to `gpt-5.4-nano`. More than 500 models are available from providers including OpenAI, Anthropic, Google, Alibaba Cloud, xAI, Mistral, OpenRouter, Infron, and others. For a full list, see the [AI models list](https://developer.puter.com/ai/models/) page. |
+| `stream` | `Boolean` | A boolean indicating whether you want to stream the completion. Defaults to `false`. |
+| `max_tokens` | `Number` | The maximum number of tokens to generate in the completion. By default, the specific model's maximum is used. |
+| `temperature` | `Number` | A number between 0 and 2 indicating the randomness of the completion. Lower values make the output more focused and deterministic, while higher values make it more random. By default, the model's own temperature is used. See [Model-specific features](#model-specific-features). |
+| `reasoning_effort` / `reasoning.effort` | `String` | Controls how much effort reasoning models spend thinking. Supported values: `none`, `minimal`, `low`, `medium`, `high`, `xhigh`, and `max`. Lower values give faster responses with less reasoning. OpenAI models, Moonshot's Kimi K3.1 Pro models, Meta's Muse Spark models, `gemini-4-argon`, `gemini-3.1-flash-lite` and `gemini-3-flash-preview` only. See [Model-specific features](#model-specific-features). |
+| `thinking` | `Object \| null` | Extended-thinking control: `{ type: "adaptive" \| "enabled" \| "disabled" \| "between_tools", budgetTokens?, display? }`. Pass `null` to send no thinking configuration at all (the model's own default); leaving it unset falls back to the legacy derivation from `reasoning_effort`. `budgetTokens` only applies to `type: "enabled"`. |
+| `verbosity` / `text.verbosity` | `String` | Controls how long or short responses are. Supported values: `low`, `medium`, and `high`. Lower values give shorter responses. OpenAI models only. |
+| `tools` | `Array` | Function definitions the AI can call. See [Function Calling](#function-calling) for details. |
+| `toolChoice` | `String \| Object` | Which tool, if any, the model must call. Pass `"auto"` (the model decides), `"none"` (never call a tool), `"required"` (call at least one), or an object naming a specific tool: `{ type: "tool", name: "..." }`. Requires `tools`. |
+| `parallelToolCalls` | `Boolean` | Whether the model may call more than one tool in a single turn. Left unset, tool calls run one at a time. |
+| `streamToolInput` | `Boolean` | Streaming only. When `true`, a tool call's arguments arrive incrementally as `tool_use_start` / `tool_input_delta` chunks instead of a single `tool_use` chunk once the call completes. |
+| `normalize` | `Boolean` | Controls the format of the non-streaming response. See [Response normalization](#response-normalization). |
+| `compaction` | `Boolean \| Object` | Opt into inline context compaction for long conversations. See [Compaction](#compaction). |
+| `signal` | `AbortSignal` | Abort the request, including while waiting for the next streamed chunk. Cancellation rejects with the signal reason (normally an `AbortError`). |
+| `provider` | `String` | Pin the request to a specific provider, for example `openrouter` or `infron`. Without it, Puter selects a provider for the requested model. Call [`puter.ai.listModelProviders()`](/AI/listModelProviders) for the available values, and [`puter.ai.listModels(provider)`](/AI/listModels) for the models a given provider serves. See [Providers](#providers). |
+| `fallback` | `Boolean` | Set to `false` to fail the request when the first provider tried fails, instead of retrying it on another provider that serves the same model. Defaults to `true`. See [Providers](#providers). |
+| `outputFormat` | `Object` | Constrains the response to a JSON schema: `{ type: "json_schema", schema, name?, strict? }`. |
+| `stopSequences` | `Array<String>` | Up to the model's limit of strings; generation stops the moment one of them appears in the output. |
 
 Some models accept additional options or handle these options differently. See [Model-specific features](#model-specific-features).
 
@@ -399,6 +399,23 @@ The code implementation is available in our [web search example](/playground/ai-
 
 List of OpenAI models that support the web search can be found in their [API compatibility documentation](https://platform.openai.com/docs/guides/tools-web-search#api-compatibility).
 
+#### Prompt Caching
+
+OpenAI models cache prompt prefixes on their own. Requests from the same app and user share one cache by default, and you can pick the cache with `prompt_cache_key`. Their `usage` reports the tokens read from the cache as `cached_tokens`.
+
+GPT-5.6 and later models charge for cache writes and let you choose how long an entry lives. Pass `prompt_cache_options` with a `ttl` of `"30m"` (the default) or `"12h"`. The `usage` reports the tokens written with the 30 minute lifetime as `cache_write_tokens` and the tokens written with the 12 hour lifetime as `cache_write_12h_tokens`. Writing a 30 minute entry costs 1.25 times the model's input rate, and writing a 12 hour entry costs 2.5 times the input rate. Reads cost the same for both lifetimes. To mark where a reusable prefix ends, add `prompt_cache_breakpoint: { mode: "explicit" }` to a content part. A breakpoint can carry its own `ttl`, which overrides the request's. OpenAI only takes a 12 hour lifetime, or a lifetime on a breakpoint, through its Responses API, so a request that uses either is sent there.
+
+```js
+puter.ai.chat('<your message>', {
+    model: 'gpt-6-luna',
+    prompt_cache_options: { ttl: '12h' }
+});
+```
+
+Older OpenAI models charge nothing for cache writes. Pass `prompt_cache_retention` with `"in_memory"` for an entry that lives a few minutes or `"24h"` for one that can live up to a day. The 24 hour option is the default, and GPT-5.5 and later models only support it.
+
+The `usageDetails` object on every result carries the same counts under provider-neutral names, with one field per lifetime, such as `cacheWrite30mTokens` and `cacheWrite12hTokens`. See the [ChatResponse](/Objects/chatresponse/) object for the full list.
+
 ### Anthropic models
 
 #### Prompt Caching

src/docs/src/AI/convo.md

@@ -4,7 +4,9 @@ description: Prepare an OpenAI realtime conversation with optional audio and tex
 platforms: [websites, apps, nodejs, workers]
 ---
 
-The conversation API prepares a realtime session, verifies Puter sign-in, and lets you connect when your app is ready. Realtime conversations require a paid Puter plan; usage is billed to the signed-in user. Native OpenAI events and convenience methods use the same session. Microphone capture and speaker playback start when your app requests them.
+Prepares a realtime OpenAI conversation that your app can connect to when it's ready.
+
+Preparing the conversation verifies Puter sign-in. Realtime conversations require a paid Puter plan, and usage is billed to the signed-in user. Native OpenAI events and convenience methods use the same session. Microphone capture and speaker playback start when your app requests them.
 
 In Node.js, the SDK requires a global `WebSocket`, available in Node.js 22 or later.
 
@@ -21,15 +23,17 @@ await conversation.connect();
 
 ## Parameters
 
+#### `options` (Object) (Optional)
+
 The options object accepts these fields.
 
-| Field | Type | Default | Description |
-| --- | --- | --- | --- |
-| `model` | String | `gpt-realtime-2.1-mini` | Also accepts `gpt-realtime-2.1`, `gpt-realtime-2`, `gpt-realtime-1.5`, `gpt-realtime`, and `gpt-realtime-mini`. |
-| `instructions` | String | Provider default | Session instructions, up to 8192 characters. |
-| `output` | Array of Strings | `['audio']` | Accepts `['text']`, `['audio']`, or both. Audio includes a transcript. Requesting both uses audio output and emits transcript deltas through the `text` event. |
-| `voice` | String | Provider default | OpenAI voice identifier, up to 128 characters. |
-| `turnDetection` | Boolean | `true` | Whether the provider detects the end of spoken turns automatically. Set to `false` for manual audio commits. |
+| Option | Type | Description |
+|--------|------|-------------|
+| `model` | `String` | The realtime model. Defaults to `gpt-realtime-2.1-mini`. Also accepts `gpt-realtime-2.1`, `gpt-realtime-2`, `gpt-realtime-1.5`, `gpt-realtime`, and `gpt-realtime-mini`. |
+| `instructions` | `String` | Session instructions, up to 8192 characters. Defaults to the provider's default. |
+| `output` | `Array<String>` | Accepts `['text']`, `['audio']`, or both. Defaults to `['audio']`. Audio includes a transcript. Requesting both uses audio output and emits transcript deltas through the `text` event. |
+| `voice` | `String` | OpenAI voice identifier, up to 128 characters. Defaults to the provider's default. |
+| `turnDetection` | `Boolean` | Whether the provider detects the end of spoken turns automatically. Defaults to `true`. Set to `false` for manual audio commits. |
 
 ## Return value
 
@@ -41,12 +45,12 @@ The conversation exposes these properties and methods.
 | --- | --- |
 | `state` | `prepared`, `connecting`, `connected`, or `closed`. |
 | `model` | Resolved model name. |
-| `connect()` | Opens the connection and applies settings. Resolves after the provider acknowledges session configuration. Concurrent calls share one attempt. Calling it while connected opens no additional connection. During a reconnection, it resolves when the reconnection succeeds and rejects if the conversation closes instead. |
+| `connect()` | Opens the connection and applies settings. Resolves after the provider acknowledges session configuration. Concurrent calls share one connection Promise, including bounded retries for transient setup failures. Calling it while connected opens no additional connection. During a reconnection, it resolves when the reconnection succeeds and rejects if the conversation closes instead. |
 | `configure(options)` | Updates the given session settings and resolves after the provider acknowledges this update. Settings it isn't given, including ones set through native events, are left alone. Accepts the factory options except `model`. One update can be pending at a time. Provider rules apply to changes such as switching a voice after audio has been generated. |
 | `sendText(text)` | Adds a user message and requests a response. Text must be nonempty and at most 8192 characters. |
 | `appendAudio(pcm)` | Appends an `Int16Array` containing PCM16, 24 kHz mono audio. A chunk contains at most 12000 samples. |
 | `commitAudio()` | Commits buffered audio and requests a response. Requires turn detection to be off, through `turnDetection: false` or a native session update. The provider requires at least 100 ms of buffered audio. |
-| `interrupt(position?)` | Stops built-in playback immediately and waits for cancellation and playback-history synchronization. A reply requested but not yet started is cancelled as soon as it starts. Custom players provide `{ itemId, contentIndex, playedMs }`. An idle call resolves immediately. |
+| `interrupt(position?)` | Stops built-in playback immediately and waits for cancellation and playback-history synchronization. A reply requested but not yet started is cancelled as soon as it starts. Custom players provide `{ itemId, contentIndex, playedMs }`. An idle call resolves immediately. Readiness failures reject the returned Promise with `session_not_ready` or `session_closed`. |
 | `send(event)` | Sends a native OpenAI JSON event. |
 | `on(type, handler)` | Registers a listener and returns an unsubscribe function. Use `'*'` to receive every native and convenience event. |
 | `off(type, handler)` | Removes a listener. |
@@ -54,9 +58,9 @@ The conversation exposes these properties and methods.
 
 Attach listeners before connecting to receive initial session events. Closing during connection rejects the pending attempt.
 
-If the connection drops because of network loss, a server restart or a provider error, the conversation reconnects on its own. While it reconnects, `state` is `connecting`, a `reconnecting` event fires, and pending operations reject with `connection_lost`. The new session gets every setting the provider acknowledged, from the factory options, `configure()` and native `session.update` events. Then `ready` fires again with `reconnected: true`. Listeners, the microphone and the speaker carry over. Microphone audio captured while reconnecting is dropped.
+If the connection drops because of network loss, a server restart or a provider error, the conversation reconnects on its own. Normal closes with code 1000 end the conversation. While it reconnects, `state` is `connecting`, a `reconnecting` event fires, and pending operations reject with `connection_lost`. The new session gets every setting the provider acknowledged, from the factory options, `configure()` and native `session.update` events. Then `ready` fires again with `reconnected: true`. Listeners, the microphone and the speaker carry over. Microphone audio captured while reconnecting is dropped.
 
-The conversation also restores recent context. Before `ready` fires, it adds back up to 40 of the most recent conversation items, 48 KiB in total. Restored items are user text, user speech that has a transcript, assistant replies as text, and function calls with their outputs. User speech has a transcript only when input transcription is configured through native session settings. An assistant reply that was interrupted, or still playing when the connection dropped, is cut to what the user saw or heard. The cut is estimated from how much of the audio played, so it can be off by a few words. Images, speech without a transcript, deleted items, and replies still generating when the connection dropped are left out. Restored items get new item IDs, don't fire item events, and count as input tokens on the next response. Item IDs from before the reconnect don't work in later events. If the provider refuses a restored item, an `error` event fires with code `context_restore_failed` and the conversation stays connected.
+The conversation also restores recent context. Before `ready` fires, it adds back up to 40 of the most recent conversation items, 48 KiB in total. Restored items are user text, user speech that has a transcript, assistant replies as text, and function calls with their outputs. User speech has a transcript only when input transcription is configured through native session settings. An assistant reply that was interrupted, or still playing when the connection dropped, is cut to what the user saw or heard. The cut is estimated from how much of the audio played, so it can be off by a few words. Images, speech without a transcript, deleted items, and replies still generating when the connection dropped are left out. Restored items get new item IDs, don't fire item events, and count as input tokens on the next response. Item IDs from before the reconnect don't work in later events. Restoration uses a contiguous suffix of restorable items. An oversized newest item leaves that suffix empty. Older items left out to fit the window fire a `context_trimmed` event with the count in `omittedItems`; this is routine and not an error. A provider refusal or a failure to send the restored items fires one `error` event with code `context_restore_failed` per restoration. In both cases the conversation stays connected. Items are preserved without truncating text, function arguments or function results.
 
 When Puter knows ahead of time that the server a session runs on will restart, it sends a `puter.reconnect` event, which also reaches your listeners. Without that notice, a restart drops the connection and the conversation reconnects as above. Within a few seconds, the conversation opens a replacement session in the background while the current one keeps working. It switches once no reply is generating or playing, no user turn is being spoken or transcribed, no function call is waiting for its output, and nothing you sent is waiting on the provider. The switch goes through the same steps as a reconnect, and the `reconnecting` event has code 1012. If the replacement can't open yet, the conversation keeps the current session and tries again later. A refusal that retrying can't fix, such as `insufficient_funds`, stops the attempts. If the server closes the session first, the conversation switches to the replacement, or reconnects if none is open.
 
@@ -64,6 +68,14 @@ The conversation makes up to ten attempts over about a minute. In a browser that
 
 Native event definitions are listed in the [OpenAI realtime reference](https://developers.openai.com/api/reference/resources/realtime/server-events). Native events retain their provider fields. Sending native events preserves direct control of the session. Puter refuses `conversation.item.retrieve` events and MCP tools (`"type": "mcp"`) in `session.update` and `response.create` events. Sending one fires an `error` event with code `invalid_event`, and the conversation closes with code 1008. Audio helpers expect the session to use PCM16 at 24 kHz.
 
+## Errors and limits
+
+Failures use objects containing `code` and `message`. Preparation can reject with `invalid_argument`, `model_not_found`, `auth_required`, `auth_canceled`, or `auth_failed`. Connection can reject with `subscription_required` or `insufficient_funds`, which also show the Puter upgrade prompt. Connection and asynchronous operations can reject with provider errors, `session_closed`, `connection_lost`, `operation_timeout`, or `operation_in_progress`. Sending before readiness, or while reconnecting, throws `session_not_ready`.
+
+Invalid input throws `invalid_argument`, `invalid_event`, or `input_too_large`. A full SDK send queue throws `queue_full`. Media helpers report `unsupported_environment`, `permission_denied`, `media_error`, `media_timeout`, or `queue_full`. Runtime media errors and `context_restore_failed` arrive through the `error` event. Provider errors also arrive as native `error` events with a nested `error` object.
+
+Each native event is capped at 64 KiB. The SDK bounds JSON traversal, outgoing queues, acknowledgment waits, microphone work, and playback. All numeric limits are listed on [rate limits and quotas](/rate-limits-and-quotas/).
+
 ## Convenience events
 
 The conversation emits these events alongside native events.
@@ -72,6 +84,7 @@ The conversation emits these events alongside native events.
 | --- | --- |
 | `ready` | `model`, `provider`, and `reconnected: true` after a reconnection |
 | `reconnecting` | `code`, `reason` of the dropped connection. After a `puter.reconnect` event, `code` is 1012. |
+| `context_trimmed` | `omittedItems`, the number of restorable items left out of the restored context after a reconnection. Fires before `ready`. |
 | `text` | `delta`, `responseId` |
 | `transcript` | `role`, `text`, `final`. Assistant deltas use `final: false`. Completed assistant and user transcripts use `final: true`. User transcripts require transcription to be configured through native session settings. |
 | `audio` | `pcm`, `sampleRate`, `responseId`, `itemId`, `contentIndex`. The PCM chunk is an `Int16Array`. |
@@ -96,16 +109,10 @@ The optional microphone and speaker helpers require a browser with Web Audio sup
 
 Both start methods require a connected session. Capture and playback are independent, so custom audio input can use the built-in speaker, and built-in capture can use custom playback.
 
-## Errors and limits
-
-Failures use objects containing `code` and `message`. Preparation can reject with `invalid_argument`, `model_not_found`, `auth_required`, `auth_canceled`, or `auth_failed`. Connection can reject with `subscription_required` or `insufficient_funds`, which also show the Puter upgrade prompt. Connection and asynchronous operations can reject with provider errors, `session_closed`, `connection_lost`, `operation_timeout`, or `operation_in_progress`. Sending before readiness, or while reconnecting, throws `session_not_ready`.
-
-Invalid input throws `invalid_argument`, `invalid_event`, or `input_too_large`. A full SDK send queue throws `queue_full`. Media helpers report `unsupported_environment`, `permission_denied`, `media_error`, `media_timeout`, or `queue_full`. Runtime media errors and `context_restore_failed` arrive through the `error` event. Provider errors also arrive as native `error` events with a nested `error` object.
-
-Each native event is capped at 64 KiB. The SDK bounds JSON traversal, outgoing queues, acknowledgment waits, microphone work, and playback. All numeric limits are listed on [rate limits and quotas](/rate-limits-and-quotas/).
-
 ## Examples
 
+### Have a text conversation
+
 A text conversation uses convenience methods for input and streamed output.
 
 ```js
@@ -122,6 +129,8 @@ await conversation.connect();
 conversation.sendText('Say hello in three words.');
 ```
 
+### Have a voice conversation
+
 A voice conversation enables the microphone and speaker after connection.
 
 ```js

src/docs/src/AI/listModelProviders.md

@@ -22,6 +22,8 @@ A `Promise` that will resolve to an array of string containing each AI providers
 
 ## Examples
 
+### List all model providers
+
 ```html;ai-list-model-providers
 <html>
 <body>

src/docs/src/AI/listModels.md

@@ -43,13 +43,15 @@ Example model entry:
 ]
 ```
 
+<div class="info">Successful results are cached for 60 seconds per provider.</div>
+
+## Errors
+
 Network and server failures reject the promise; they do not resolve to an empty
 catalogue. Network failures use `{ message, code: "network_error" }`, invalid
 catalogue responses use `invalid_response`, idle timeouts use `request_timeout`,
 and server errors pass through unchanged.
 
-<div class="info">Successful results are cached for 60 seconds per provider.</div>
-
 ## Model variants
 
 Some models have variants. A variant is the same model with a different price, rate limit, or speed. The variant is a suffix on the model ID, separated by a colon.
@@ -64,6 +66,8 @@ The model ID without a suffix is the standard variant. Not every model has varia
 
 ## Examples
 
+### List all models
+
 ```html;ai-list-models
 <html>
 <body>

src/docs/src/AI/realtime.md

@@ -4,7 +4,9 @@ description: Connect to OpenAI realtime conversations through Puter without pute
 platforms: [websites, apps, nodejs, workers]
 ---
 
-The Realtime WebSocket API lets your client send and receive native OpenAI realtime JSON events through Puter. Puter authenticates the user and charges their account for usage.
+The Realtime WebSocket API lets your client send and receive native OpenAI realtime JSON events through Puter.
+
+Puter authenticates the user and charges their account for usage.
 
 **Beta:** This API is incomplete. Its methods, options, events, and behavior may change abruptly, including breaking changes without notice.
 
@@ -30,6 +32,8 @@ Browser WebSocket clients pass `realtime` and `openai-insecure-api-key.<PUTER_TO
 
 The socket can open before authentication and provider setup finish. Listen for errors and closure immediately. Wait for `session.created`, send your session settings, then wait for `session.updated` before submitting a turn.
 
+To open a replacement while another connection remains healthy, set the optional `purpose=handover` query parameter. These attempts share a smaller per-credential budget, checked before the general connection-attempt budget. Initial connections and dropped-connection recovery use the URL without this parameter. The limits are listed in [rate limits and quotas](/rate-limits-and-quotas/).
+
 ## Text example
 
 Install the `ws` package and set `PUTER_TOKEN` to your Puter token. Run this example as a Node.js ES module:
@@ -109,8 +113,14 @@ A new connection starts an empty conversation. To keep context, add earlier item
 
 ## Errors, billing, and limits
 
-API errors arrive as `{ "type": "error", "error": { "code": "...", "message": "..." } }`. Handle socket errors and close frames too. Authentication failures use `invalid_api_key`. Other relay codes include `model_not_found`, `insufficient_funds`, `subscription_required`, `rate_limit_exceeded`, `invalid_event`, `connection_timeout`, `provider_unavailable`, `upstream_error`, and `server_restarting`.
+API errors arrive as `{ "type": "error", "error": { "code": "...", "message": "..." } }`. Handle socket errors and close frames too. Invalid, missing, or revoked credentials use `invalid_api_key`. If Puter can't check a credential or the account's plan, the error is `internal_error` and the session closes with 1011, so try again after a short, randomized delay. Other relay codes include `permission_denied`, `model_not_found`, `insufficient_funds`, `subscription_required`, `rate_limit_exceeded`, `invalid_event`, `connection_timeout`, `provider_unavailable`, `upstream_error`, and `server_restarting`.
 
-Policy failures generally close with 1008, oversized client messages with 1009, upstream or setup failures with 1011, and server restarts with 1012. Authentication and model failures can arrive after the WebSocket handshake succeeds.
+Policy failures generally close with 1008, oversized client messages with 1009, upstream or setup failures with 1011, server restarts with 1012, and temporary server overload with 1013. Reconnect after 1011, 1012, or 1013. Authentication and model failures can arrive after the WebSocket handshake succeeds.
 
 The endpoint requires a paid Puter plan. Usage is charged after completed responses and input transcriptions. While a response generates, part of your balance is held for it, so a session opens only if your balance covers one response. If the provider connection closes before a response reports its usage, the response is charged the full amount held for it. Each account can open 20 sessions at once by default. Each JSON message you send is limited to 64 KiB, including reassembled fragments. Messages from the provider can be up to 4 MiB. Stream audio in small chunks. All message rates, queue caps, timeouts, and other bounds are published in [rate limits and quotas](/rate-limits-and-quotas/).
+
+Input transcription holds a small amount of your balance per transcribed item, priced at the transcription model you select. Two items are reserved ahead of your audio and replenished as the provider commits them, so a session holds only its spare items plus the items still awaiting a transcript. Changing the transcription model reprices the spare items. If your balance can't cover another item, audio waits while earlier items settle; once none are left to settle, the session closes with `insufficient_funds`. Failed transcriptions, items without final usage, and audio whose commit the provider never acknowledges before the connection closes are each charged that model's fixed per-item amount. Unacknowledged audio is charged one item per started 160 seconds. With server turn detection, only detected speech counts, so silence is never charged. The amounts, timeout and audio limits are listed in [rate limits and quotas](/rate-limits-and-quotas/).
+
+Provider-reported realtime models must have a known price. A missing model uses the last known model. An explicitly unpriced model closes the connection with `upstream_error`.
+
+After recording usage, the connection can continue while settled credit is positive, including a fractional balance. Each new billable operation still needs its full reservation after accounting for other outstanding holds.

src/docs/src/AI/txt2speech.listEngines.md

@@ -54,7 +54,7 @@ Example response:
 
 ## Examples
 
-<strong class="example-title">List engines for a specific provider</strong>
+### List engines for a specific provider
 
 ```html;ai-txt2speech-list-engines
 <html>
@@ -73,7 +73,7 @@ Example response:
 </html>
 ```
 
-<strong class="example-title">List engines using options object</strong>
+### List engines using options object
 
 ```js
 const engines = await puter.ai.txt2speech.listEngines({ provider: 'elevenlabs' });

src/docs/src/AI/txt2speech.listVoices.md

@@ -4,7 +4,9 @@ description: List available TTS voices, optionally filtered by provider.
 platforms: [websites, apps, nodejs, workers]
 ---
 
-Returns the voices available from a TTS provider. Each voice entry includes metadata such as language, category, and supported models.
+Returns the voices available from a TTS provider.
+
+Each voice entry includes metadata such as language, category, and supported models.
 
 ## Syntax
 
@@ -52,7 +54,7 @@ Example response (with `provider: 'all'`):
 
 ## Examples
 
-<strong class="example-title">List voices for a provider</strong>
+### List voices for a provider
 
 ```html;ai-txt2speech-list-voices
 <html>
@@ -71,7 +73,7 @@ Example response (with `provider: 'all'`):
 </html>
 ```
 
-<strong class="example-title">List all default (AWS Polly) voices</strong>
+### List all default (AWS Polly) voices
 
 ```js
 const voices = await puter.ai.txt2speech.listVoices();
@@ -81,7 +83,7 @@ for (const voice of voices) {
 }
 ```
 
-<strong class="example-title">List Gemini voices</strong>
+### List Gemini voices
 
 ```js
 const voices = await puter.ai.txt2speech.listVoices({ provider: 'gemini' });

src/docs/src/AI/txt2speech.md

@@ -4,7 +4,9 @@ description: Convert text to speech with AI using multiple languages, voices, an
 platforms: [websites, apps, nodejs, workers]
 ---
 
-Converts text into speech using AI. Supports multiple languages and voices.
+Converts text into speech using AI.
+
+It supports multiple languages and voices.
 
 ## Syntax
 

src/docs/src/Apps.md

@@ -183,9 +183,9 @@ These Apps API are supported out of the box when using Puter.js:
 
 - **[`puter.apps.create()`](/Apps/create/)** - Create a new application
 - **[`puter.apps.list()`](/Apps/list/)** - List all applications
-- **[`puter.apps.delete()`](/Apps/delete/)** - Delete an application
-- **[`puter.apps.update()`](/Apps/update/)** - Update application settings
 - **[`puter.apps.get()`](/Apps/get/)** - Get information about a specific application
+- **[`puter.apps.update()`](/Apps/update/)** - Update application settings
+- **[`puter.apps.delete()`](/Apps/delete/)** - Delete an application
 - **[`puter.apps.checkName()`](/Apps/checkName/)** - Check whether an app name is available
 
 ## Examples

src/docs/src/Apps/checkName.md

@@ -4,7 +4,9 @@ description: Check whether an app name is still available.
 platforms: [websites, apps, nodejs, workers]
 ---
 
-Checks whether an app name is free, without creating anything. App names are unique across all Puter accounts, so a name is free only when no app on any account uses it. Useful before calling [`puter.apps.create()`](/Apps/create/), which rejects when the name is already taken.
+Checks whether an app name is free, without creating anything.
+
+App names are unique across all Puter accounts, so a name is free only when no app on any account uses it. Useful before calling [`puter.apps.create()`](/Apps/create/), which rejects when the name is already taken.
 
 ## Syntax
 
@@ -24,7 +26,7 @@ A `Promise` that will resolve to an object describing the name's availability.
 
 ## Examples
 
-<strong class="example-title">Check a name before creating the app</strong>
+### Check a name before creating the app
 
 ```html
 <html>

src/docs/src/Apps/create.md

@@ -4,7 +4,9 @@ description: Create apps in the Puter desktop environment.
 platforms: [websites, apps, nodejs, workers]
 ---
 
-Creates a Puter app with the given name. The app will be created in the user's apps, and will be accessible to this app. The app will be created with no permissions, and will not be able to access any data until permissions are granted to it.
+Creates a Puter app with the given name.
+
+The app will be created in the user's apps, and will be accessible to this app. The app will be created with no permissions, and will not be able to access any data until permissions are granted to it.
 
 ## Syntax
 
@@ -39,17 +41,19 @@ The title of the app. If this parameter is not provided, the app will be created
 
 An object containing the options for the app to create. The object can contain the following properties:
 
-- `name` (String) (required): The name of the app to create. App names are unique across all Puter accounts. If any app already has this name, the promise will be rejected.
-- `indexURL` (String) (required): The URL of the app's index page. This URL must be accessible to the user. If this parameter is not provided, the promise will be rejected.
-- `title` (String) (optional): The human-readable title of the app. If this parameter is not provided, the app will be created with `name` as its title.
-- `description` (String) (optional): The description of the app aimed at the end user.
-- `icon` (String) (optional): The new icon of the app.
-- `maximizeOnStart` (Boolean) (optional): Whether the app should be maximized when it is started. Defaults to `false`.
-- `filetypeAssociations` (Array<String>) (optional): An array of strings representing the filetypes that the app can open. Defaults to `[]`. File extentions and MIME types are supported; For example, `[".txt", ".md", "application/pdf"]` would allow the app to open `.txt`, `.md`, and PDF files. At most 200 entries, each at most 60 characters.
-- `dedupeName` (Boolean) (optional) - Whether to deduplicate the app name if it already exists. Defaults to `false`.
-- `background` (Boolean) (optional) - Whether the app should run in the background. Defaults to `false`.
-- `feedbackEnabled` (Boolean) (optional) - Whether users can send feedback to you through [`puter.ui.showFeedbackDialog()`](/UI/showFeedbackDialog/). Defaults to `false`.
-- `metadata` (Object) (optional) - An object containing custom metadata for the app. This can be used to store arbitrary key-value pairs associated with the app. At most 16 KiB when serialized as JSON.
+| Option | Type | Description |
+|--------|------|-------------|
+| `name` | `String` | Required. The name of the app to create. App names are unique across all Puter accounts. If any app already has this name, the promise will be rejected. |
+| `indexURL` | `String` | Required. The URL of the app's index page. This URL must be accessible to the user. If it's missing, the promise will be rejected. |
+| `title` | `String` | The human-readable title of the app. If it's missing, the app will be created with `name` as its title. |
+| `description` | `String` | The description of the app aimed at the end user. |
+| `icon` | `String` | The new icon of the app. |
+| `filetypeAssociations` | `Array<String>` | An array of strings representing the filetypes that the app can open. Defaults to `[]`. File extensions and MIME types are supported. For example, `[".txt", ".md", "application/pdf"]` would allow the app to open `.txt`, `.md`, and PDF files. At most 200 entries, each at most 60 characters. |
+| `maximizeOnStart` | `Boolean` | Whether the app should be maximized when it is started. Defaults to `false`. |
+| `background` | `Boolean` | Whether the app should run in the background. Defaults to `false`. |
+| `feedbackEnabled` | `Boolean` | Whether users can send feedback to you through [`puter.ui.showFeedbackDialog()`](/UI/showFeedbackDialog/). Defaults to `false`. |
+| `metadata` | `Object` | An object containing custom metadata for the app. This can be used to store arbitrary key-value pairs associated with the app. At most 16 KiB when serialized as JSON. |
+| `dedupeName` | `Boolean` | Whether to deduplicate the app name if it already exists. Defaults to `false`. |
 
 ## Return value
 
@@ -57,7 +61,7 @@ A `Promise` that will resolve to the [`CreateAppResult`](/Objects/createappresul
 
 ## Examples
 
-<strong class="example-title">Create an app pointing to example.com</strong>
+### Create an app pointing to example.com
 
 ```html;app-create
 <html>

src/docs/src/Apps/delete.md

@@ -24,7 +24,7 @@ A `Promise` that will resolve to an object `{ success: true, uid: <app uid> }` i
 
 ## Examples
 
-<strong class="example-title">Create a random app then delete it</strong>
+### Create a random app then delete it
 
 ```html;app-delete
 <html>

src/docs/src/Apps/get.md

@@ -4,7 +4,9 @@ description: Retrieve details of your Puter app.
 platforms: [websites, apps, nodejs, workers]
 ---
 
-Returns an app with the given name. If the app does not exist, the promise will be rejected.
+Returns an app with the given name.
+
+If the app does not exist, the promise will be rejected.
 
 ## Syntax
 
@@ -19,13 +21,14 @@ puter.apps.get(name, options)
 
 The name of the app to get.
 
-### options (optional)
+#### `options` (Object) (optional)
 
 An object containing the following properties:
 
-- `stats_period` (optional): A string representing the period for which to get the user and open count. Possible values are `today`, `yesterday`, `7d`, `30d`, `this_month`, `last_month`, `this_year`, `last_year`, `month_to_date`, `year_to_date`, `last_12_months`. Default is `all` (all time).
-
-- `icon_size` (optional): An integer representing the size of the icons to return. Possible values are `null`, `16`, `32`, `64`, `128`, `256`, and `512`. Default is `null` (the original size).
+| Option | Type | Description |
+|--------|------|-------------|
+| `icon_size` | `Number \| null` | An integer representing the size of the icons to return. Possible values are `null`, `16`, `32`, `64`, `128`, `256`, and `512`. Default is `null` (the original size). |
+| `stats_period` | `String` | The period for which to get the user and open count. Possible values are `today`, `yesterday`, `7d`, `30d`, `this_month`, `last_month`, `this_year`, `last_year`, `month_to_date`, `year_to_date`, `last_12_months`. Default is `all` (all time). |
 
 ## Return value
 
@@ -33,7 +36,7 @@ A `Promise` that will resolve to the [`App`](/Objects/app/) object with the give
 
 ## Examples
 
-<strong class="example-title">Create a random app then get it</strong>
+### Create a random app then get it
 
 ```html;app-get
 <html>

src/docs/src/Apps/list.md

@@ -4,7 +4,9 @@ description: List all apps in your Puter account.
 platforms: [websites, apps, nodejs, workers]
 ---
 
-Returns an array of all apps belonging to the user and that this app has access to. If the user has no apps, the array will be empty.
+Returns an array of all apps belonging to the user and that this app has access to.
+
+If the user has no apps, the array will be empty.
 
 ## Syntax
 
@@ -19,19 +21,15 @@ puter.apps.list(options)
 
 An object containing the following properties:
 
-- `stats_period` (optional): A string representing the period for which to get the user and open count. Possible values are `today`, `yesterday`, `7d`, `30d`, `this_month`, `last_month`, `this_year`, `last_year`, `month_to_date`, `year_to_date`, `last_12_months`. Default is `all` (all time).
-
-- `icon_size` (optional): An integer representing the size of the icons to return. Possible values are `null`, `16`, `32`, `64`, `128`, `256`, and `512`. Default is `null` (the original size).
-
-- `limit` (optional): Maximum number of apps to return in a single call.
-
-- `offset` (optional): Skips the given number of apps. Prefer `cursor` for paging through large lists.
-
-- `cursor` (optional): Opts into paginated results. Pass `null` for the first page, then the `cursor` from each page to fetch the next one.
-
-- `includeTotal` (optional): If `true`, the paginated result includes a `total` count of the user's apps.
-
-- `stream` (optional): If `true`, the method returns an async iterator of page objects instead of a promise, for use with `for await ... of`. Combine with `limit` to control the page size, or `cursor` to resume from a previous page. Cannot be combined with `offset`. With `includeTotal`, only the first page carries `total`.
+| Option | Type | Description |
+|--------|------|-------------|
+| `limit` | `Number` | Maximum number of apps to return in a single call. |
+| `cursor` | `String \| null` | Opts into paginated results. Pass `null` for the first page, then the `cursor` from each page to fetch the next one. |
+| `stream` | `Boolean` | If `true`, the method returns an async iterator of page objects instead of a promise, for use with `for await ... of`. Combine with `limit` to control the page size, or `cursor` to resume from a previous page. Cannot be combined with `offset`. With `includeTotal`, only the first page carries `total`. |
+| `includeTotal` | `Boolean` | If `true`, the paginated result includes a `total` count of the user's apps. |
+| `offset` | `Number` | Skips the given number of apps. Prefer `cursor` for paging through large lists. |
+| `icon_size` | `Number \| null` | An integer representing the size of the icons to return. Possible values are `null`, `16`, `32`, `64`, `128`, `256`, and `512`. Default is `null` (the original size). |
+| `stats_period` | `String` | The period for which to get the user and open count. Possible values are `today`, `yesterday`, `7d`, `30d`, `this_month`, `last_month`, `this_year`, `last_year`, `month_to_date`, `year_to_date`, `last_12_months`. Default is `all` (all time). |
 
 ## Return value
 
@@ -57,7 +55,7 @@ for await (const page of puter.apps.list({ stream: true })) {
 
 ## Examples
 
-<strong class="example-title">Create 3 random apps and then list them</strong>
+### Create 3 random apps and then list them
 
 ```html;app-list
 <html>

src/docs/src/Apps/update.md

@@ -17,16 +17,19 @@ The name of the app to update.
 
 #### `attributes` (required)
 An object containing the attributes to update. The object can contain the following properties:
-- `name` (optional): The new name of the app. App names are unique across all Puter accounts. If any app already has this name, the promise will be rejected.
-- `indexURL` (optional): The new URL of the app's index page. This URL must be accessible to the user.
-- `title` (optional): The new title of the app.
-- `description` (optional): The new description of the app aimed at the end user.
-- `icon` (optional): The new icon of the app.
-- `maximizeOnStart` (optional): Whether the app should be maximized when it is started. Defaults to `false`.
-- `background` (optional): Whether the app should run in the background. Defaults to `false`.
-- `feedbackEnabled` (optional): Whether users can send feedback to you through [`puter.ui.showFeedbackDialog()`](/UI/showFeedbackDialog/). Omitted leaves the app's current value unchanged.
-- `filetypeAssociations` (optional): An array of strings representing the filetypes that the app can open. Defaults to `[]`. File extentions and MIME types are supported; For example, `[".txt", ".md", "application/pdf"]` would allow the app to open `.txt`, `.md`, and PDF files. At most 200 entries, each at most 60 characters.
-- `metadata` (optional): An object containing custom metadata for the app. This can be used to store arbitrary key-value pairs associated with the app. At most 16 KiB when serialized as JSON.
+
+| Option | Type | Description |
+|--------|------|-------------|
+| `title` | `String` | The new title of the app. |
+| `indexURL` | `String` | The new URL of the app's index page. This URL must be accessible to the user. |
+| `description` | `String` | The new description of the app aimed at the end user. |
+| `icon` | `String` | The new icon of the app. |
+| `name` | `String` | The new name of the app. App names are unique across all Puter accounts. If any app already has this name, the promise will be rejected. |
+| `filetypeAssociations` | `Array<String>` | An array of strings representing the filetypes that the app can open. Defaults to `[]`. File extensions and MIME types are supported. For example, `[".txt", ".md", "application/pdf"]` would allow the app to open `.txt`, `.md`, and PDF files. At most 200 entries, each at most 60 characters. |
+| `maximizeOnStart` | `Boolean` | Whether the app should be maximized when it is started. Defaults to `false`. |
+| `background` | `Boolean` | Whether the app should run in the background. Defaults to `false`. |
+| `feedbackEnabled` | `Boolean` | Whether users can send feedback to you through [`puter.ui.showFeedbackDialog()`](/UI/showFeedbackDialog/). Omitted leaves the app's current value unchanged. |
+| `metadata` | `Object` | An object containing custom metadata for the app. This can be used to store arbitrary key-value pairs associated with the app. At most 16 KiB when serialized as JSON. |
 
 Changing `indexURL` can merge the app into an existing app for that hosted origin. The returned app can have a different UID. The deleted source app's KV data and KV shares are removed. Data already belonging to the surviving app is retained.
 
@@ -35,7 +38,7 @@ A `Promise` that will resolve to the [`App`](/Objects/app/) object that was upda
 
 ## Examples
 
-<strong class="example-title">Create a random app then change its title</strong>
+### Create a random app then change its title
 
 ```html;app-update
 <html>

src/docs/src/Auth.md

@@ -98,8 +98,8 @@ This is essential for users to access the various Puter.js APIs integrated into
 These authentication features are supported out of the box when using Puter.js:
 
 - **[`puter.auth.signIn()`](/Auth/signIn/)** - Sign in a user
-- **[`puter.auth.signOut()`](/Auth/signOut/)** - Sign out the current user
 - **[`puter.auth.isSignedIn()`](/Auth/isSignedIn/)** - Check if a user is signed in
+- **[`puter.auth.signOut()`](/Auth/signOut/)** - Sign out the current user
 - **[`puter.auth.getUser()`](/Auth/getUser/)** - Get information about the current user
 - **[`puter.auth.getProfile()`](/Auth/getProfile/)** - Get a user's profile, when available
 - **[`puter.auth.getProfilePicture()`](/Auth/getProfilePicture/)** - Get a user's profile picture, when available

src/docs/src/Auth/getDetailedAppUsage.md

@@ -29,7 +29,9 @@ The id of the application.
 
 A `Promise` that resolves to a [`DetailedAppUsage`](/Objects/detailedappusage) object containing resource usage statistics for the given application.
 
-## Example
+## Examples
+
+### Get an app's usage
 
 ```html
 <html>

src/docs/src/Auth/getMonthlyUsage.md

@@ -26,7 +26,9 @@ None
 
 A `Promise` that resolves to a [`MonthlyUsage`](/Objects/monthlyusage) object containing the user's monthly usage information.
 
-## Example
+## Examples
+
+### Get the user's monthly usage
 
 ```html;auth-get-monthly-usage
 <html>

src/docs/src/Auth/getProfile.md

@@ -4,7 +4,9 @@ description: Read a user's profile, returning null when it is not available.
 platforms: [websites, apps, nodejs, workers]
 ---
 
-Returns a user's profile: their picture, display name, and bio. Users set these on the [Account page](https://puter.com/dashboard#account) of their Puter dashboard. A user can always read their own profile. Another user's profile is available only while that user is on a paid plan; otherwise the method resolves to `null`, the same as for a user who does not exist.
+Returns a user's profile, which holds their picture, display name, and bio.
+
+Users set these on the [Account page](https://puter.com/dashboard#account) of their Puter dashboard. A user can always read their own profile. Another user's profile is available only while that user is on a paid plan; otherwise the method resolves to `null`, the same as for a user who does not exist.
 
 ## Syntax
 
@@ -15,7 +17,7 @@ puter.auth.getProfile(username);
 
 ## Parameters
 
-### `username` (optional)
+#### `username` (optional)
 
 The username to look up. Defaults to the signed-in user.
 
@@ -25,7 +27,9 @@ A promise that resolves to a [`UserProfile`](/Objects/userprofile) object, or `n
 
 When signed out with no `username`, it resolves to `null` without opening a sign-in prompt.
 
-## Example
+## Examples
+
+### Show the user's profile
 
 ```html
 <html>

src/docs/src/Auth/getProfilePicture.md

@@ -4,7 +4,9 @@ description: Read a user's profile picture, returning null when unavailable.
 platforms: [websites, apps, nodejs, workers]
 ---
 
-Returns the `picture` field of a user's profile (see [`puter.auth.getProfile()`](/Auth/getProfile/)). A user can always read their own picture. Another user's picture is available only while that user is on a paid plan.
+Returns the `picture` field of a user's profile (see [`puter.auth.getProfile()`](/Auth/getProfile/)).
+
+A user can always read their own picture. Another user's picture is available only while that user is on a paid plan.
 
 ## Syntax
 
@@ -15,7 +17,7 @@ puter.auth.getProfilePicture(username);
 
 ## Parameters
 
-### `username` (optional)
+#### `username` (optional)
 
 The username to look up. Defaults to the signed-in user.
 
@@ -27,7 +29,9 @@ Missing users, profiles that are not public to the caller, profiles with no pict
 
 When signed out with no `username`, it resolves to `null` without opening a sign-in prompt.
 
-## Example
+## Examples
+
+### Show the user's profile picture
 
 ```html
 <html>

src/docs/src/Auth/getUser.md

@@ -20,7 +20,9 @@ None
 
 A promise that resolves to a [`User`](/Objects/user) object containing the user's basic information.
 
-## Example
+## Examples
+
+### Get the signed-in user
 
 ```html;auth-get-user
 <html>

src/docs/src/Auth/isSignedIn.md

@@ -20,7 +20,9 @@ None
 
 Returns `true` if the user is signed in, `false` otherwise.
 
-## Example
+## Examples
+
+### Check whether the user is signed in
 
 ```html;auth-is-signed-in
 <html>

src/docs/src/Auth/signIn.md

@@ -4,7 +4,9 @@ description: Initiate sign in process in your application with user's Puter acco
 platforms: [websites]
 ---
 
-Initiates the sign in process for the user. This will open a popup window with the appropriate authentication method. Puter automatically handles the authentication process and will resolve the promise when the user has signed in.
+Initiates the sign in process for the user.
+
+This will open a popup window with the appropriate authentication method. Puter automatically handles the authentication process and will resolve the promise when the user has signed in.
 
 It is important to note that all essential methods in Puter handle authentication automatically. This method is only necessary if you want to handle authentication manually, for example if you want to build your own custom authentication flow.
 
@@ -27,11 +29,12 @@ puter.auth.signIn(options)
 
 #### `options` (optional)
 
-`options` is an object with the following properties:
+An object with the following properties:
 
-- `attempt_temp_user_creation`: A boolean value that indicates whether to Puter should automatically create a temporary user. This is useful if you want to quickly onboard a user without requiring them to sign up. They can always sign up later if they want to.
-
-- `request_auth`: A boolean value that asks the popup to let the user re-pick their account, even when your site already holds a token for them. Puter otherwise skips that prompt for a site it has seen before. Useful for an explicit "switch account" button.
+| Option | Type | Description |
+|--------|------|-------------|
+| `attempt_temp_user_creation` | `Boolean` | Whether Puter should automatically create a temporary user. This is useful if you want to quickly onboard a user without requiring them to sign up. They can always sign up later if they want to. |
+| `request_auth` | `Boolean` | Asks the popup to let the user re-pick their account, even when your site already holds a token for them. Puter otherwise skips that prompt for a site it has seen before. Useful for an explicit "switch account" button. |
 
 Each call completes from its own sign-in window, including when multiple calls run at once.
 
@@ -39,7 +42,7 @@ Each call completes from its own sign-in window, including when multiple calls r
 
 A `Promise` that will resolve to a [`SignInResult`](/Objects/signinresult/) object when the user has signed in.
 
-## Rejection
+## Errors
 
 The promise will reject with an object containing an `error` code and a human-readable `msg` in the following cases:
 
@@ -53,7 +56,9 @@ The promise will reject with an object containing an `error` code and a human-re
 
 The promise may also reject with the failure response returned by the authentication window itself.
 
-## Example
+## Examples
+
+### Sign in from a button click
 
 ```html;auth-sign-in
 <html>

src/docs/src/Auth/signOut.md

@@ -22,7 +22,9 @@ None
 
 A `Promise` that resolves with nothing once the user is signed out and, on a website, this browser's push subscription for your app is ended. It doesn't reject.
 
-## Example
+## Examples
+
+### Sign the user out
 
 ```html;auth-sign-out
 <html>