arcane-os 0.7.0 → 0.7.2

package/browser-runtime/ai/speech-worker-runtime.mjs

@@ -1592,6 +1592,52 @@ export function createSpeechWorkerRuntime({ role, scope = globalThis, send } = {
     if (engine) return status();
     lifecycleReason = `${role}-load-started`;
     let loadFailureReason = `${role}-worker-runtime-configuration-rejected`;
+    let reporting = true;
+    const completedFiles = new Set();
+    const loadingFiles = new Set();
+    function publishProgress(progress) {
+        if (!reporting || signal?.aborted) return;
+        send(
+            {
+                protocol: SPEECH_WORKER_PROTOCOL,
+                id: request.id,
+                type: 'progress',
+                progress,
+            }
+        );
+    }
+    function reportModelProgress(detail) {
+        if (!reporting || signal?.aborted) return;
+        const file = detail?.file ?? null;
+        if (file !== null) {
+            const fileKey = `${detail.name ?? ''}/${file}`;
+            if (detail?.status === 'done') {
+                completedFiles.add(fileKey);
+                loadingFiles.delete(fileKey);
+            } else {
+                loadingFiles.add(fileKey);
+            }
+        }
+        const initializing = detail?.status === 'ready'
+            || (detail?.status === 'done' && loadingFiles.size === 0);
+        const message = initializing
+            ? 'Preparing the speech model session'
+            : detail?.status === 'done'
+                ? 'Loaded model file'
+                : 'Loading model file';
+        publishProgress(
+            {
+                phase: initializing ? 'initialize' : 'download',
+                stage: 'model',
+                message,
+                file,
+                completed: completedFiles.size,
+                total: null,
+                unit: 'files',
+                detail,
+            }
+        );
+    }
     try {
       configuration = validateConfiguration(request.payload?.configuration, role);
       const entry = configuration.runtime.files.find((file) =>
@@ -1605,6 +1651,15 @@ export function createSpeechWorkerRuntime({ role, scope = globalThis, send } = {
         });
       }
       loadFailureReason = `${role}-worker-runtime-import-rejected`;
+      publishProgress(
+          {
+              phase: 'initialize',
+              stage: 'runtime',
+              message: 'Opening the speech runtime',
+              file: entry.path,
+              total: null,
+          }
+      );
       const namespace = await import(entry.moduleUrl);
       throwIfAborted(signal, `${role}-load-cancelled`);
       restoreNamespace = configureRuntimeNamespace(
@@ -1614,10 +1669,17 @@ export function createSpeechWorkerRuntime({ role, scope = globalThis, send } = {
         environment.cache,
       );
       loadFailureReason = `${role}-worker-model-load-rejected`;
-      const report = () => undefined;
+      publishProgress(
+          {
+              phase: 'initialize',
+              stage: 'model',
+              message: `Initializing the speech model on ${configuration.execution?.device ?? 'wasm'}`,
+              total: null,
+          }
+      );
       engine = role === "stt"
-        ? await createWhisperEngine(namespace, configuration, signal, report)
-        : await createKokoroEngine(namespace, configuration, signal, report);
+        ? await createWhisperEngine(namespace, configuration, signal, reportModelProgress)
+        : await createKokoroEngine(namespace, configuration, signal, reportModelProgress);
       lifecycleReason = `${role}-load-completed`;
       return status();
     } catch (error) {
@@ -1645,6 +1707,8 @@ export function createSpeechWorkerRuntime({ role, scope = globalThis, send } = {
       configuration = null;
       lifecycleReason = failure.reason;
       throw failure;
+    } finally {
+        reporting = false;
     }
   }
 

package/browser-runtime/ai/wllama/index.mjs

@@ -3490,7 +3490,9 @@ var Wllama = class {
   }
 
   const adapterAnchor = "if(adapter){WebGPU.Internals.jsObjectInsert(adapterPtr,adapter)";
-  const adapterProjection = `if(adapter){(function arcaneRecordSelectedWebgpuAdapter(){const key="__arcaneWllamaWebgpuTelemetry";const previous=globalThis[key]??${emptyWorkerTelemetry};const info=adapter.info??{};const text=value=>typeof value==="string"?value.slice(0,256):"";const next={selected:true,vendorId:null,vendor:text(info.vendor),architecture:text(info.architecture),deviceId:null,name:text(info.device),description:text(info.description)};const conflicts=previous.adapter!==null&&JSON.stringify(previous.adapter)!==JSON.stringify(next);globalThis[key]={protocol:"${protocol}",adapter:previous.adapter??next,bufferCount:previous.bufferCount,bufferBytes:previous.bufferBytes,queueSubmissions:previous.queueSubmissions,commandBuffers:previous.commandBuffers,queueFenceRequests:previous.queueFenceRequests,queueFenceCompletions:previous.queueFenceCompletions,invalid:previous.invalid||conflicts}})();WebGPU.Internals.jsObjectInsert(adapterPtr,adapter)`;
+  const adapterProjection = `if(adapter){(function arcaneRecordSelectedWebgpuAdapter(){const key="__arcaneWllamaWebgpuTelemetry";const previous=globalThis[key]??${emptyWorkerTelemetry};const info=adapter.info??{};const text = function preserveAdapterText(value) {
+    return typeof value === 'string' ? value : '';
+};const next={selected:true,vendorId:null,vendor:text(info.vendor),architecture:text(info.architecture),deviceId:null,name:text(info.device),description:text(info.description),type:typeof info.type==="string"&&info.type?info.type:null,isFallbackAdapter:typeof info.isFallbackAdapter==="boolean"?info.isFallbackAdapter:typeof adapter.isFallbackAdapter==="boolean"?adapter.isFallbackAdapter:null};const conflicts=previous.adapter!==null&&JSON.stringify(previous.adapter)!==JSON.stringify(next);globalThis[key]={protocol:"${protocol}",adapter:previous.adapter??next,bufferCount:previous.bufferCount,bufferBytes:previous.bufferBytes,queueSubmissions:previous.queueSubmissions,commandBuffers:previous.commandBuffers,queueFenceRequests:previous.queueFenceRequests,queueFenceCompletions:previous.queueFenceCompletions,invalid:previous.invalid||conflicts}})();WebGPU.Internals.jsObjectInsert(adapterPtr,adapter)`;
   WLLAMA_EMSCRIPTEN_CODE = replaceSingle(
     WLLAMA_EMSCRIPTEN_CODE,
     adapterAnchor,
@@ -3526,7 +3528,7 @@ var Wllama = class {
   );
 
   const workerAnchor = "  if (verb === 'module.init') {";
-  const workerProjection = `  if (verb === 'arcane.telemetry') {\n    const observed = globalThis.__arcaneWllamaWebgpuTelemetry;\n    const adapter = observed?.adapter?.selected === true ? {\n      selected: true,\n      vendorId: null,\n      vendor: typeof observed.adapter.vendor === 'string' ? observed.adapter.vendor.slice(0, 256) : '',\n      architecture: typeof observed.adapter.architecture === 'string' ? observed.adapter.architecture.slice(0, 256) : '',\n      deviceId: null,\n      name: typeof observed.adapter.name === 'string' ? observed.adapter.name.slice(0, 256) : '',\n      description: typeof observed.adapter.description === 'string' ? observed.adapter.description.slice(0, 256) : '',\n    } : null;\n    msg({\n      callbackId,\n      result: {\n        protocol: '${protocol}',\n        adapter,\n        bufferCount: Number.isSafeInteger(observed?.bufferCount) ? observed.bufferCount : 0,\n        bufferBytes: Number.isSafeInteger(observed?.bufferBytes) ? observed.bufferBytes : 0,\n        queueSubmissions: Number.isSafeInteger(observed?.queueSubmissions) ? observed.queueSubmissions : 0,\n        commandBuffers: Number.isSafeInteger(observed?.commandBuffers) ? observed.commandBuffers : 0,\n        queueFenceRequests: Number.isSafeInteger(observed?.queueFenceRequests) ? observed.queueFenceRequests : 0,\n        queueFenceCompletions: Number.isSafeInteger(observed?.queueFenceCompletions) ? observed.queueFenceCompletions : 0,\n        invalid: observed?.invalid === true,\n      },\n    });\n    return;\n  }\n\n${workerAnchor}`;
+  const workerProjection = `  if (verb === 'arcane.telemetry') {\n    const observed = globalThis.__arcaneWllamaWebgpuTelemetry;\n    const adapter = observed?.adapter?.selected === true ? {\n      selected: true,\n      vendorId: null,\n      vendor: typeof observed.adapter.vendor === 'string' ? observed.adapter.vendor : '',\n      architecture: typeof observed.adapter.architecture === 'string' ? observed.adapter.architecture : '',\n      deviceId: null,\n      name: typeof observed.adapter.name === 'string' ? observed.adapter.name : '',\n      description: typeof observed.adapter.description === 'string' ? observed.adapter.description : '',\n      type: typeof observed.adapter.type === 'string' ? observed.adapter.type : null,\n      isFallbackAdapter: typeof observed.adapter.isFallbackAdapter === 'boolean' ? observed.adapter.isFallbackAdapter : null,\n    } : null;\n    msg({\n      callbackId,\n      result: {\n        protocol: '${protocol}',\n        adapter,\n        bufferCount: Number.isSafeInteger(observed?.bufferCount) ? observed.bufferCount : 0,\n        bufferBytes: Number.isSafeInteger(observed?.bufferBytes) ? observed.bufferBytes : 0,\n        queueSubmissions: Number.isSafeInteger(observed?.queueSubmissions) ? observed.queueSubmissions : 0,\n        commandBuffers: Number.isSafeInteger(observed?.commandBuffers) ? observed.commandBuffers : 0,\n        queueFenceRequests: Number.isSafeInteger(observed?.queueFenceRequests) ? observed.queueFenceRequests : 0,\n        queueFenceCompletions: Number.isSafeInteger(observed?.queueFenceCompletions) ? observed.queueFenceCompletions : 0,\n        invalid: observed?.invalid === true,\n      },\n    });\n    return;\n  }\n\n${workerAnchor}`;
   LLAMA_CPP_WORKER_CODE = replaceSingle(
     LLAMA_CPP_WORKER_CODE,
     workerAnchor,
@@ -3539,7 +3541,7 @@ var Wllama = class {
       return Number.isSafeInteger(candidate) && candidate >= 0 ? candidate : 0;
     }
     function adapterText(candidate) {
-      return typeof candidate === "string" && candidate.length <= 256 ? candidate : "";
+      return typeof candidate === "string" ? candidate : "";
     }
     const rawAdapter = value?.adapter;
     const adapterInvalid = rawAdapter !== undefined && rawAdapter !== null && (
@@ -3547,16 +3549,12 @@ var Wllama = class {
       || rawAdapter.selected !== true
       || rawAdapter.vendorId !== null
       || typeof rawAdapter.vendor !== "string"
-      || rawAdapter.vendor.length > 256
       || typeof rawAdapter.architecture !== "string"
-      || rawAdapter.architecture.length > 256
       || rawAdapter.deviceId !== null
       || typeof rawAdapter.name !== "string"
-      || rawAdapter.name.length > 256
       || typeof rawAdapter.description !== "string"
-      || rawAdapter.description.length > 256
     );
-    const adapter = rawAdapter?.selected === true ? Object.freeze({
+    const adapter = rawAdapter?.selected === true ? {
       selected: true,
       vendorId: null,
       vendor: adapterText(rawAdapter.vendor),
@@ -3564,7 +3562,9 @@ var Wllama = class {
       deviceId: null,
       name: adapterText(rawAdapter.name),
       description: adapterText(rawAdapter.description),
-    }) : null;
+      type: typeof rawAdapter.type === "string" ? rawAdapter.type : null,
+      isFallbackAdapter: typeof rawAdapter.isFallbackAdapter === "boolean" ? rawAdapter.isFallbackAdapter : null,
+    } : null;
     return Object.freeze({
       protocol,
       adapter,

package/docs/architecture.md

@@ -279,7 +279,7 @@ paths are withheld from the native provider. The provider copies the complete
 selected release rather than accepting an unrelated source path. Verification
 is a separate explicit operation for a selected release artifact.
 
-The SDK `0.7.0` runtime requires Arcane `0.8.12` or newer. Compatibility
+The SDK `0.7.2` runtime requires Arcane `0.8.12` or newer. Compatibility
 is contractual rather than exact-version pinning: the prepared Core must meet
 the highest minimum declared by the runtime, selected app, and bundled app
 dependencies; keep each app's Arcane protocol generation; and provide every

package/docs/reference/ai/browser-speech.md

@@ -144,28 +144,34 @@ name. It does not load a model or create a GPU device or WebNN context:
 </html-import>
 ```
 
-The **NPU setup** section provides real browser-specific flags links. Chrome uses
+**NPU setup** provides a **Copy NPU flag address** button regardless of WebNN
+availability, with one short explanation. Chrome uses
 `chrome://flags/#web-machine-learning-neural-network`; Edge uses
 `edge://flags/#web-machine-learning-neural-network`. Unrecognized browsers
-receive explicit Chrome and Edge choices. Full addresses and copy controls
-remain available if the browser blocks internal-page links.
+receive explicit Chrome and Edge choices. A failed copy shows the complete
+address as selectable text; the component does not open internal browser pages.
 The [ONNX Runtime WebNN guide](https://onnxruntime.ai/docs/tutorials/web/ep-webnn.html)
 documents the **Enables WebNN API** flag and model/operator requirements.
 
-The **GPU performance** section shows adapter detection and, on desktop Windows
-Chromium browsers, the matching Force High Performance GPU flags link. Identified
+The **GPU performance** section shows a short adapter status and a **Copy GPU
+flag address** button on desktop Windows Chromium browsers. The button remains
+available for every adapter class, unavailable API, or detection failure. One
+shared instruction explains pasting, enabling, saving work, and relaunching. Identified
 Chrome uses `chrome://flags/#force-high-performance-gpu`; other recognized
-Chromium browsers use their corresponding internal scheme. **Detect GPU** or
-**Refresh browser availability** requests a new availability result; concurrent
+Chromium browsers use their corresponding internal scheme. **Refresh**
+requests a new availability result; concurrent
 requests share the pending operation. The public `checkGpu()` method provides
 the same promise, while `refresh()` remains a synchronous API-presence and
 settings update. Component readiness does not wait for adapter detection.
 
 Adapter selection uses `powerPreference: "high-performance"` as a hint, not proof
-of the selected GPU's performance or the flag's current state. A returned
-adapter establishes availability only; its fallback classification is reported
-when exposed. The component cannot read the browser flag or prove that a model
-is executing on the adapter. See the
+of the selected GPU's performance or the flag's current state. When the browser
+explicitly reports a discrete GPU, Profile says **Already using the performance
+GPU.** Explicit integrated or software/fallback
+metadata is reported as such. Missing type metadata shows **GPU available.**
+and leaves performance selection unconfirmed; a vendor name is not a GPU
+classification. The component cannot read the browser flag or prove that a
+model is executing on the adapter. See the
 [component contract](../runtime-components.md#browser-ai-setuphtml) for result
 fields, failure handling, and disposal behavior.
 
@@ -1204,10 +1210,22 @@ that accelerator kernels overlap, or that generated speech is correct. WebNN
 may execute unsupported operations through WASM. A security field is absent
 in ordinary mode.
 
-The provider/2 load context accepts an optional progress callback for interface
-compatibility, but the current browser-speech artifact and Worker transport
-publishes no progress records. Consumers present the explicit lifecycle states
-instead of inventing a numeric total.
+The provider/2 load context accepts an optional `progress` callback. Speech
+artifact preparation and upstream model loading publish records through that
+callback and the shared sticky AI runtime state. Records include `phase`,
+`stage`, `message`, the current `file` when available, and `elapsedMs`.
+Known artifact inventories report `completed`, `total`, and `unit:'files'`.
+Upstream model files are discovered during loading, so their total remains
+`null` while completed files are counted. Remaining TTS pool initialization
+reports completed model sessions. Consumers use an indeterminate progress bar
+when no final total is known; transfer quantities do not determine this display.
+
+The Worker sends intermediate `{protocol,id,type:'progress',progress}` messages
+for its active load request. They do not settle the request; the ordinary final
+response still owns completion or failure. Complete upstream callback content
+remains in the adjacent `detail` field for diagnostics. Compatible observers
+receive the latest progress when joining an active load, and cancelled,
+superseded, or settled loads stop publishing progress.
 
 Compatible concurrent loads share one underlying preparation and pool load
 while each caller retains its own cancellation signal. One observer cannot

package/docs/reference/ai/browser-wasm.md

@@ -187,34 +187,42 @@ Applications remain responsible for model selection and license compliance.
 cannot download. Source downloads use CORS, omit credentials and referrer,
 disable HTTP caching, and honor `AbortSignal`.
 
-On Windows desktop Chromium browsers, observing a WebGPU adapter may open one
-flags page and show one advisory per page session. The notice names only
-**Force High Performance GPU** (`#force-high-performance-gpu`), conditional on
-the computer having multiple GPUs, and asks the user to completely close and
-reopen their browser after enabling it. It applies to any observed GPU vendor;
-an Intel identity alone does not establish that an adapter is integrated or
-slower, and the browser does not expose a reliable machine-wide GPU count.
-
-An identified Microsoft Edge, Brave, Opera, or Vivaldi receives its own
-`edge://`, `brave://`, `opera://`, or `vivaldi://` flags address. Chrome,
-Chromium, and browsers that conceal their brand receive
-`about://flags/#force-high-performance-gpu`, which the browser rewrites to its
-own flags page, with neutral browser wording. Only the current browser's one
-address appears. Firefox, Safari, non-Windows platforms, and mobile browsers
-receive no flag notice. The notice does not suggest unrelated WebGPU, ANGLE,
-Vulkan, rasterization, or blocklist flags.
-
-The Windows selection workaround is documented by
-[Chrome](https://developer.chrome.com/docs/web-platform/webgpu/troubleshooting-tips#Windows-specific_limitations).
-Other Chromium browsers inherit that flag; the per-vendor addresses and generic
-`about://` rewrite follow the
-[Chromium browser-flags guide](https://developer.chrome.com/blog/browser-flags/).
-[Vivaldi masks its browser identity by default](https://help.vivaldi.com/developers/web/vivaldi-user-agent-and-client-hints-user-agent/),
-so a Chrome user-agent token is not proof that the current browser is Chrome.
-This source-based guidance is not a live GPU-selection test of every browser
-release. The advisory does not change settings, prove another or faster adapter
-exists, inspect the flag's current value, or make a failed load succeed. If the
-browser blocks opening an internal page, the alert retains the copyable address.
+After a model loads, its existing Worker reports the adapter it selected.
+The runtime retains that adapter in `evidence().webgpu.adapter`, including
+optional browser-reported `type` and `isFallbackAdapter` fields. This reads
+the loaded Worker's metadata once; it does not request another adapter.
+Missing fields remain unknown. Cancellation also cancels the metadata wait;
+an unavailable observation is logged without rejecting an otherwise loaded model.
+
+On Windows desktop Chromium browsers, an explicitly reported integrated GPU,
+CPU, or fallback adapter triggers one GPU-only browser alert per page session.
+The alert names the selected adapter and **Force High Performance GPU**
+(`#force-high-performance-gpu`), conditional on the computer also having a
+discrete GPU. It contains the address to paste manually and asks the user to
+save their work and completely close and reopen the browser after changing
+the flag. It does not open a window or navigate to an internal browser page.
+Discrete adapters and unknown adapter classes do not trigger the alert.
+
+An identified Chrome, Microsoft Edge, Brave, Opera, or Vivaldi receives its own
+`chrome://`, `edge://`, `brave://`, `opera://`, or `vivaldi://` flags address.
+Other detected Chromium browsers receive
+`about://flags/#force-high-performance-gpu` with neutral browser wording.
+Only the current browser's address appears. Firefox, Safari, non-Windows
+platforms, and mobile browsers receive no flag notice. The alert contains no
+NPU guidance or unrelated flags. The Profile `browser-ai-setup` component
+provides Copy controls for its browser addresses.
+
+An adapter vendor or name does not establish its GPU class, and
+`powerPreference: "high-performance"` is a request hint rather than proof of
+the selected hardware. Chromium's optional adapter `type` is a
+[developer feature](https://developer.chrome.com/docs/web-platform/webgpu/developer-features);
+the standard API does not always expose integrated or discrete classification.
+The Profile says **Already using the performance GPU.** for an explicitly
+reported discrete GPU and hides performance flag advice. Missing classification
+shows **GPU available.** without flag advice and remains unconfirmed.
+The browser does not expose the flag's current value or a reliable machine-wide
+GPU count, and a discrete adapter does not prove it is the fastest available.
+See [Chrome's GPU selection guidance](https://developer.chrome.com/docs/web-platform/webgpu/troubleshooting-tips#webgpu_is_slower_than_webgl).
 
 ## Streaming, cancellation, and tools
 

package/docs/reference/inventory/runtime-components.json

@@ -5,7 +5,7 @@
     "repository": "https://github.com/TheWizardNexus/arcane-os-sdk.git",
     "branch": "main",
     "path": "runtime/arcane/components",
-    "sdkVersion": "0.6.2"
+    "sdkVersion": "0.7.2"
   },
   "componentCount": 40,
   "loader": "/arcane/modules/HTMLImport.js",
@@ -79,7 +79,7 @@
     {
       "file": "runtime/arcane/components/browser-ai-setup.html",
       "name": "browser-ai-setup.html",
-      "purpose": "Displays browser API and WebGPU adapter availability with browser-specific NPU and GPU flags links.",
+      "purpose": "Displays browser API availability, browser-reported GPU class, and copyable NPU/GPU flags addresses.",
       "methods": [
         "refresh()",
         "checkGpu()",
@@ -97,8 +97,8 @@
         "arcane-os/event-manager"
       ],
       "availability": "Browser and supported native WebViews",
-      "transport": "HTMLImport + DOM; browser flags links, Clipboard API, WebGPU adapter request, and preserved open() navigation attempt with alert",
-      "normalization": "API presence, adapter availability, and setup instructions normalized. Mounting starts one adapter request; checkGpu() coalesces concurrent calls, while refresh() remains synchronous. No GPU device, model, or WebNN context is created. Adapter availability does not establish browser flag state, high-performance selection, or model execution."
+      "transport": "HTMLImport + DOM; Clipboard API and WebGPU adapter request. Settings addresses use Copy controls and appear as text if copying fails; open() copies the NPU address without navigation or an alert.",
+      "normalization": "API presence, adapter availability, explicit browser GPU class, and copy instructions normalized. Supported browser NPU and GPU flag copy controls remain available regardless of API availability, adapter class, or detection failure. A reported discrete GPU shows Already using the performance GPU. Absent type metadata shows GPU available and leaves performance selection unconfirmed. Mounting starts one adapter request; checkGpu() coalesces concurrent calls, while refresh() remains synchronous. No GPU device, model, or WebNN context is created. Adapter metadata does not establish browser flag state, the fastest physical GPU, or another page's model execution."
     },
     {
       "file": "runtime/arcane/components/calculator.html",

package/docs/reference/protocols.md

@@ -380,8 +380,13 @@ the module declares a shadowing binding with the same name. Explicit
 `globalThis` or `self` calls continue through the materialized-file router.
 
 The speech Worker uses its ordinary global Worker message boundary directly for
-requests, results, errors, and cancellation. It creates no private
-`MessageChannel` and publishes no Worker progress transport.
+requests, progress, results, errors, and cancellation. It creates no private
+`MessageChannel`. During loading, `{protocol,id,type:'progress',progress}`
+messages retain the active request id without settling its pending result.
+Progress records describe the current phase, stage, message, file, and completed
+file count with `total:null` for dynamically discovered model files. Complete
+upstream callback content remains in the adjacent `detail` field. The ordinary
+final success or failure envelope ends the request and its progress stream.
 
 Worker operations use `arcane-ai-speech-worker/1`. The public Worker client
 supports `load`, `use`, `status`, `unload`, and `dispose`; the transport host
@@ -459,9 +464,11 @@ while the loaded provider remains ready.
 Speech failure neither disables text chat nor retries through another local,
 native, or cloud provider. The provider/Worker layer is event-neutral: it
 exposes promises, `AbortSignal`, and precise lifecycle/status records, but owns
-no event bus or listener registry. The provider/2 load context accepts an
-optional progress callback for interface compatibility; the current
-browser-speech artifact and Worker transport publishes no progress records.
+no event bus. The provider/2 load context accepts an optional `progress`
+callback that receives artifact preparation, upstream model loading, and session
+initialization records. Compatible concurrent observers share one load and
+receive its current progress; cancelling one observer leaves the others active.
+The shared AI runtime projects those records into sticky role state.
 
 ### Persistent chat and document context
 

package/docs/reference/runtime-components.md

@@ -60,7 +60,7 @@ appropriate.
 | --- | --- | --- | --- | --- |
 | [`app-bar.html`](#app-barhtml) | Responsive application navigation, route state, status, and trailing actions. | `setNavigation()`<br>`setActiveRoute()`<br>`setStatus()`<br>`refresh()`<br>`destroy()` | `app-bar-ready` | DOM-normalized |
 | [`assistant-panel.html`](#assistant-panelhtml) | Reusable assistant drawer, message area, composer, pending/streaming/empty/error state, and actions. | `open()`<br>`close()`<br>`toggle()`<br>`send()`<br>`clear()`<br>`setState()`<br>`focusComposer()`<br>`scrollToEnd()`<br>`destroy()` | `assistant-ready`<br>`assistant-opened`<br>`assistant-closed`<br>`assistant-send`<br>`assistant-clear` | DOM-normalized; caller/provider results remain external |
-| [`browser-ai-setup.html`](#browser-ai-setuphtml) | Browser API and WebGPU adapter availability, with browser-specific NPU and GPU flags links. | `refresh()`<br>`checkGpu()`<br>`open()`<br>`destroy()`<br>`ready` | `browser-ai-setup-ready` | API and adapter availability; browser flag state and model execution remain browser-owned |
+| [`browser-ai-setup.html`](#browser-ai-setuphtml) | Browser API availability, reported GPU class, and copyable NPU/GPU settings addresses. | `refresh()`<br>`checkGpu()`<br>`open()`<br>`destroy()`<br>`ready` | `browser-ai-setup-ready` | Browser-reported adapter class; browser flag state and model execution remain browser-owned |
 | [`calculator.html`](#calculatorhtml) | Calculator keypad and result/error event surface backed by CalculatorEngine. | `calculate()`<br>`destroy()` | `calculator-ready`<br>`calculation-complete`<br>`calculation-error` | Normalized Calculation/error events |
 | [`chart.html`](#charthtml) | Accessible uPlot line, area, or point chart with normalized options and rows. | `configure()`<br>`populate()`<br>`setData()`<br>`addData()`<br>`update()`<br>`destroy()` | `chart-ready`<br>`chart-remove` | Options/rows normalized; uPlot rendering is vendor-native |
 | [`chat.html`](#chathtml) | Shared chat, visible selected-model activation request, file upload, streaming, structural tool settlement, speech, language, availability, and conversation-timebox surface. | `streamMessage()`<br>`setMessageProgress()`<br>`setAIAvailability()`<br>`setInitialSpeechMuted()`<br>`setConversationComplete()`<br>`bindConversationTimebox()`<br>`bindSession()`<br>`submitMessage()`<br>`submitToolResult()`<br>`submitToolResults()`<br>`sendMessage()`<br>`languageChanged()`<br>`requestAIActivation()`<br>`destroy()` | `chat-ready`<br>`chat-session-bound`<br>`chat-session-message`<br>`chat-session-error`<br>`chat-send-message`<br>`chat-send-error`<br>`chat-file-uploaded`<br>`chat-file-upload-error`<br>`chat-language-changed`<br>`chat-language-change-error`<br>`chat-ai-activation-request`<br>`chat-ai-activation-error`<br>`chat-speech-synthesis-error`<br>`conversation-timebox-error` | UI/runtime state, explicit user activation intent, and honest structural-call settlement normalized; AI/storage/media behavior mixed |
@@ -158,14 +158,16 @@ Slots: `title`, `subtitle`, `identity`, `messages/message`, `composer`, `actions
 
 ### Overview
 
-Displays WebNN and WebGPU API availability for the current page, detects an
-available WebGPU adapter, and provides browser-specific NPU and GPU flags links.
-The WebNN addresses are real links: Chrome and Edge receive their own addresses;
-other or unidentified browsers show both explicit choices. Desktop Windows
-Chromium browsers also receive the matching Force High Performance GPU link,
+Displays a short WebNN availability status and the detected WebGPU adapter.
+Chrome and Edge receive a **Copy NPU flag address** button for their own browser;
+unidentified browsers show both choices. Desktop Windows Chromium browsers
+receive a **Copy GPU flag address** button for the matching browser,
 including `chrome://flags/#force-high-performance-gpu` for identified Chrome.
-Copy controls and complete addresses remain available when the browser blocks
-navigation to an internal page.
+Both controls remain available regardless of API availability, adapter class,
+or detection failure. Each flag has one short explanation, followed by a shared
+paste, enable, save, and relaunch instruction.
+Settings addresses use Copy controls and appear as selectable text if copying
+fails. The component does not attempt navigation to internal browser pages.
 
 ### Public surface
 
@@ -184,26 +186,31 @@ calls share the pending promise. The component requests an adapter with the
 `"high-performance"` power-preference hint; it does not create a GPU device or
 load a model. An unavailable result contains `available: false` and a `reason`
 of `"api-unavailable"` or `"adapter-unavailable"`. An available result contains
-`available: true`, the reported adapter `name`, and `isFallbackAdapter` as a
-boolean or `null` when the browser exposes no classification. Adapter selection
-does not prove that a browser flag is enabled, that the fastest GPU was selected,
-or that a model is using it. Failure is displayed and logged, and the promise
+`available: true`, the reported adapter `name`, `isFallbackAdapter` as a boolean
+or `null`, `adapterType` as the browser's string or `null`, and
+`performanceStatus`: `"discrete"`, `"integrated"`, `"fallback"`, or `"unknown"`.
+Only the explicit browser type or fallback field establishes that status;
+vendor names and the requested power preference do not. A discrete result
+shows **Already using the performance GPU.** An unknown class shows **GPU available.**
+and leaves performance selection unconfirmed. Adapter selection does not
+prove that a browser flag is enabled, that the fastest GPU was selected, or that
+a model is using it. Failure is displayed and logged, and the promise
 resolves to `false`; destruction also yields `false` and suppresses late UI
 updates.
 
-Mounting starts one `checkGpu()` operation. **Detect GPU** repeats it explicitly;
-**Refresh browser availability** calls both `refresh()` and `checkGpu()`. Calling
+Mounting starts one `checkGpu()` operation. **Refresh** calls both `refresh()`
+and `checkGpu()`. Calling
 `refresh()` directly remains synchronous and does not request an adapter.
 `ready` and `browser-ai-setup-ready` describe component setup, which can complete
 while adapter detection is still pending.
 
-`open()` refreshes the display, attempts to open the detected Chrome or Edge flags
-page, and presents an alert containing the full address and enable/relaunch
-instructions. Pass `"chrome"` or `"edge"` to choose explicitly. Invoke it from a
-user action. It returns `false` when destroyed or when no supported target is
-selected, and otherwise returns `undefined`; it never reports that navigation
-succeeded. Browsers may block internal-page navigation, so the full addresses
-also remain visible and copyable in the component.
+The retained `open(browserId?)` method now copies the matching NPU flags address
+instead of navigating or showing an alert. Pass `"chrome"` or `"edge"` to choose
+explicitly and invoke it from a user action. It returns a promise resolving to
+`true` only after clipboard success, or `false` for destruction, an unsupported
+target, or copy failure. The complete address remains selectable if clipboard
+access fails. On-screen instructions explain pasting it into the address bar,
+enabling WebNN, and relaunching.
 
 `destroy()` aborts owned listeners, disposes the event source, marks `ready`
 false, and suppresses UI updates from pending clipboard and adapter operations.
@@ -1497,7 +1504,11 @@ For a selected `unloaded`, `loading`, `unloading`, or `error` role, the componen
 keeps recording Start disabled and presents the same keyboard-operable Start
 transcription/Try again or Cancel loading control as `speech.html`. Both use the
 shared `createSTTActivationController()` contract. User operation emits the
-cancelable `speech-stt-activation-request` event with mutable complete `{intent,state}`;
+cancelable `speech-stt-activation-request` event with mutable complete `{intent,state}`.
+During loading, both controls show a native progress bar and the current
+preparation, download, or initialization message with the current file and
+reported elapsed time. Unknown totals remain indeterminate. In `speech.html`,
+prior local status stays visible alongside the loading status. The activation event's
 `preventDefault()` suppresses the callback. The default
 `requestSTTActivation(intent)` publishes the mutable
 `{role:'stt',action:'load'|'unload',reason:'user'}` intent. Callback failure emits

package/docs/reference/runtime-modules.md

@@ -1417,7 +1417,7 @@ behavior for chart, dashboard, Markdown, and voice components.
 ### Public surface
 
 Constant sets plus normalization, formatting, and explicit STT activation
-helpers. `createSTTActivationController({host,button,onChange,EventClass=CustomEvent})`
+helpers. `createSTTActivationController({host,button,progress=null,onChange,EventClass=CustomEvent})`
 consumes only normalized
 [`AIRuntimeState`](#airuntimestatejs) `stt` role records. Its mutable controller
 exposes `action`, `error`, `label`, `pending`, `selected`, `status`, `title`, and
@@ -1428,6 +1428,12 @@ exposes `action`, `error`, `label`, `pending`, `selected`, `status`, `title`, an
 callers use the default `CustomEvent`; non-DOM callers must inject a compatible
 `EventClass` constructor.
 
+An optional native `progress` element presents the active loading phase. Known
+file or session totals produce determinate progress; an unknown total remains
+indeterminate. The shared status includes the complete loading message, current
+file, completion count, and reported elapsed time. Terminal state and destruction
+clear the loading display without starting any provider work.
+
 `request('load'|'unload')` emits the cancelable
 `speech-stt-activation-request` event with mutable `{intent,state}` before it
 invokes `host.requestSTTActivation(intent)`. Callback failure emits

package/docs/reference/sdk-api.md

@@ -6395,9 +6395,11 @@ and materializes object URLs, then returns
 `{cache,runtime,model,release}`. Call `release()` when the Worker no longer
 needs those URLs. `offline:true` never fetches and rejects a miss with
 `ARCANE_AI_ARTIFACT_OFFLINE_MISS`. `remove(authority)` deletes that exact
-authority's files and selection metadata. `onProgress` remains an optional
-provider-interface callback, but the current store publishes no progress
-records.
+authority's files and selection metadata. Optional `onProgress` receives
+preparation and download records with `phase`, `stage`, `message`, and the
+current `file` when available. File operations report actual `completed` and
+`total` inventory counts with `unit:'files'`; module preparation uses
+`total:null`. Cancellation stops progress along with the owning operation.
 
 ### Availability and normalization
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "arcane-os",
-  "version": "0.7.0",
+  "version": "0.7.2",
   "description": "Arcane OS JavaScript SDK, project-local CLI, browser runtime, and repository-portable application packager.",
   "type": "module",
   "main": "./src/index.mjs",