arcane-os 0.6.1 → 0.7.0

package/CHANGELOG.md

@@ -1,5 +1,28 @@
 # Changelog
 
+## 0.7.0
+
+- Add `arcane dev --public`, including forwarding through
+  `npm run dev -- --app <id> --public`, to bind the selected app's existing
+  source server to all IPv4 interfaces. Keep the ordinary localhost default
+  and let an explicit `--host` select the bind address.
+- Allow explicit network hosts in the shared development server. Report usable
+  local and network application URLs, including `networkUrls` in server results
+  and lifecycle events, while preserving selected routes and cancellation.
+- Document LAN use and the browser HTTPS requirement for features that need a
+  secure context. No app-specific server, firewall rule, or tunnel is added.
+
+## 0.6.2
+
+- Replace browser NPU setup buttons with direct browser-specific flags links
+  and retain copy-address controls and the public `open()` method.
+- Add the missing GPU performance flags link, setup instructions, and GPU
+  detection control. Chrome uses its own `chrome://` flags address.
+- Show adapter availability and its reported name after one shared detection
+  request. Keep fallback and failed detection visible without claiming that
+  the browser performance flag is enabled or a model is running on that GPU.
+  Detection creates no device, loads no model, and changes no browser settings.
+
 ## 0.6.1
 
 - Add a reusable profile component for browser NPU setup and WebNN/WebGPU API

package/README.md

@@ -19,7 +19,7 @@ version-locked SDK runtime, while an integrated Arcane checkout uses its live
 `arcane/` runtime. Both profiles preserve the same app URLs, theme, packaging,
 event, cancellation, and browser run contracts.
 
-This checkout defines the `0.6.1` SDK contract. Applications pin one exact npm
+This checkout defines the `0.7.0` SDK contract. Applications pin one exact npm
 version and lockfile; registry state is deliberately not baked into application
 artifacts.
 
@@ -35,12 +35,27 @@ Create one browser application, install its pinned SDK, and start its source
 server:
 
 ```bash
-npx arcane-os@0.6.1 new hello-speech --path ./hello-speech --target browser
+npx arcane-os@0.7.0 new hello-speech --path ./hello-speech --target browser
 cd hello-speech
 npm install
 npm run dev
 ```
 
+To open a selected app from another device on the network, append `--public`
+to the same command:
+
+```bash
+npm run dev -- --app hello-speech --public
+```
+
+Replace `hello-speech` with any app in the workspace. The server binds to
+`0.0.0.0` and prints network URLs to open on the other device. Plain `npm run dev`
+keeps its localhost default. An explicit `--host` overrides the public bind
+address, and `--port` selects the port. Network reachability depends on the
+machine's firewall and network; the command does not configure internet hosting
+or HTTPS. Browser features requiring a [secure context](https://developer.mozilla.org/en-US/docs/Web/Security/Defenses/Secure_Contexts)
+need HTTPS when opened through a LAN address.
+
 Open the URL printed by the server. The generated page owns its import map and
 Arcane theme; its application module is `apps/hello-speech/modules/App.js`.
 `arcane/AI` is a managed **browser import**, not an npm-exported Node inference
@@ -363,7 +378,7 @@ uses the same controller for automatic memory extraction.
 Create a new repository-shaped Arcane application with the exact stable SDK:
 
 ```bash
-npx arcane-os@0.6.1 new my-app --path ./my-app --target portable --git
+npx arcane-os@0.7.0 new my-app --path ./my-app --target portable --git
 cd my-app
 npm install
 npm run dev
@@ -373,7 +388,7 @@ To enroll an existing repository, install the exact SDK and initialize only
 missing Arcane files:
 
 ```bash
-npm install --save-dev --save-exact arcane-os@0.6.1
+npm install --save-dev --save-exact arcane-os@0.7.0
 npm exec -- arcane init my-app --target portable
 ```
 
@@ -389,7 +404,7 @@ npm exec -- arcane-os targets
 No global SDK install or standalone Arcane CLI is required. The application
 repository's exact npm dependency and lockfile own the CLI and toolchain version.
 
-Use `npx arcane-os@0.6.1` for the initial bootstrap because it names this npm
+Use `npx arcane-os@0.7.0` for the initial bootstrap because it names this npm
 package explicitly; bare `npx arcane` outside an installed project could resolve
 a different package. Both installed commands invoke the same headless toolchain.
 Project-local npm scripts use the SDK pinned by that app's `package-lock.json`,
@@ -409,7 +424,7 @@ node ./bin/arcane.mjs new local-app --path ../local-app --target portable --git
 
 # From the generated app repository
 cd ../local-app
-npm install --save-dev --save-exact ../arcane-os-sdk/arcane-os-0.6.1.tgz
+npm install --save-dev --save-exact ../arcane-os-sdk/arcane-os-0.7.0.tgz
 npm ci
 ```
 
@@ -418,7 +433,7 @@ same location. The lockfile retains the selected package dependency while
 Arcane uses the installed package name and version. Local directory `file:` dependencies are not
 accepted because npm may install them as links; use a packed `.tgz`. A GitHub
 runner also needs that tarball at the locked path. After publication, replace
-the local declaration with the exact `arcane-os@0.6.1` registry package and
+the local declaration with the exact `arcane-os@0.7.0` registry package and
 commit the regenerated lock.
 
 Generated repositories use `npm ci --ignore-scripts` in CI. Run dependency
@@ -475,7 +490,7 @@ arcane new <id> [--path <directory>] [--display-name <name>] [--target <target>]
 arcane init [id] [--workspace <directory>] [--display-name <name>] [--target <target>]
 arcane doctor [--workspace <directory>] [--arcane-root <directory>]
 arcane import-map [--workspace <directory>] [--app <id>]
-arcane dev [--app <id>] [--host 127.0.0.1] [--port 8000]
+arcane dev [--app <id>] [--public] [--host <address>] [--port 8000]
 arcane test [--app <id>] [--scope app]
 arcane test --scope shared --test-file <repo-relative.test.mjs>
 arcane check [--app <id>] [--scope app] [--skip-tests]
@@ -557,7 +572,7 @@ package installation, or assertions.
 
 ## Current target support
 
-Version `0.6.1` exposes one browser target and five explicitly paired
+Version `0.7.0` exposes one browser target and five explicitly paired
 native development targets: a non-runnable portable directory, a
 Windows x64 unsigned-local-test EXE bundle, Linux x64 and Linux ARM64
 unsigned-local-test DEBs, and an Android development-signed APK. The
@@ -683,7 +698,7 @@ read-only managed-service assessment where the host supports one. Native apps
 use an app-scoped Arcane Core and `Arcane.localAI.status()` in this SDK version.
 Ollama is optional for packaging and non-AI applications.
 
-The loopback development server serves complete selected files and does not
+The development server serves complete selected files and does not
 apply byte-count or truncation gates. Optional hardening is inactive unless the
 caller explicitly selects `secure:true`. Required credential protection,
 malformed-request rejection, and unavoidable browser or operating-system safety

package/browser-runtime/ai/browser-device-settings.mjs

@@ -63,6 +63,7 @@ export function getBrowserDeviceSettings(navigatorObject = globalThis.navigator)
     ) {
       browserId = "chrome";
       name = "Google Chrome";
+      highPerformanceGpu = { name, url: "chrome://flags/#force-high-performance-gpu" };
     } else {
       browserId = "chromium";
       name = "your Chromium browser";
@@ -81,3 +82,23 @@ export function getBrowserDeviceSettings(navigatorObject = globalThis.navigator)
     webgpuAvailable: Boolean(navigatorObject?.gpu),
   };
 }
+
+// Adapter selection is availability evidence, not the browser flag's state or
+// proof that a model is executing on this GPU. The power preference is a hint.
+export async function detectBrowserGpu(navigatorObject = globalThis.navigator) {
+  if (!is.function(navigatorObject?.gpu?.requestAdapter)) {
+    return { available: false, reason: "api-unavailable" };
+  }
+  const adapter = await navigatorObject.gpu.requestAdapter({ powerPreference: "high-performance" });
+  if (!adapter) return { available: false, reason: "adapter-unavailable" };
+  const info = adapter.info;
+  return {
+    available: true,
+    name: info?.description
+      || [info?.vendor, info?.architecture, info?.device].filter(Boolean).join(" ")
+      || "WebGPU adapter",
+    isFallbackAdapter: is.boolean(info?.isFallbackAdapter)
+      ? info.isFallbackAdapter
+      : is.boolean(adapter.isFallbackAdapter) ? adapter.isFallbackAdapter : null,
+  };
+}

package/docs/architecture.md

@@ -236,7 +236,7 @@ Packaging writes the complete schema-1 release inventory to
 `ARCANE_APP_RELEASE.json`. Ordinary development, packaging, serving, and run
 paths do not count, limit, hash, truncate, tail, clip, or identify content by
 bytes, and they do not require provenance or verification receipts. The
-loopback server reads and returns each complete selected source, runtime, or
+development server reads and returns each complete selected source, runtime, or
 packaged file. Required credential protection, malformed-input rejection,
 applicable law, unavoidable protocol rules, and operating-system or browser
 safety remain in effect.
@@ -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 `0.6.1` runtime requires Arcane `0.8.12` or newer. Compatibility
+The SDK `0.7.0` 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/README.md

@@ -143,7 +143,7 @@ The synchronized browser payload exposes:
   ESM modules, classic vendor globals, one worker protocol, and one Node-oriented
   mail transport;
 - 14 shared entity modules under `runtime/arcane/entities/`;
-- 39 reusable HTML-import components under `runtime/arcane/components/`;
+- 40 reusable HTML-import components under `runtime/arcane/components/`;
 - seven shared CSS artifacts, images, optional physical-workspace security
   files where present, and the vendored `strong-type` dependency.
 

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

@@ -130,11 +130,12 @@ window.addEventListener('ai-tts-failure', function reportSpeechFailure(event) {
 
 Call `speechEvents.abort()` when disposing that interface to remove the listener.
 
-## Browser NPU setup
+## Browser NPU and GPU setup
 
 Applications can place the shared `browser-ai-setup.html` component in their
 profile or settings page. It reports whether this page exposes WebNN and
-WebGPU, without loading a model or creating an accelerator context:
+WebGPU, then requests one WebGPU adapter to display its availability and reported
+name. It does not load a model or create a GPU device or WebNN context:
 
 ```html
 <html-import
@@ -143,15 +144,31 @@ WebGPU, without loading a model or creating an accelerator context:
 </html-import>
 ```
 
-**Set up NPU** uses the same browser-settings approach as the existing
-high-performance GPU notice: attempt to open the browser's flags page, then
-show instructions including the full address to paste if navigation was
-blocked. 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 instead of an assumed target.
+The **NPU setup** section provides real browser-specific flags links. 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.
 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
+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
+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
+[component contract](../runtime-components.md#browser-ai-setuphtml) for result
+fields, failure handling, and disposal behavior.
+
 This control does not save an execution preference, change browser settings,
 restart the browser, or report that the NPU is active. A WebNN API presence
 result is not proof of NPU hardware, a compatible model, or physical execution.

package/docs/reference/cli.md

@@ -49,6 +49,7 @@ meaning and cardinality rules:
 | `--app` | app id | Workspace/app operations except shared scope and `verify-bundle`; also the exact `mail serve` caller id. |
 | `--arcane-root` | directory | `doctor`, native `build`/`run`, `native-doctor`, `native-prepare` |
 | `--host` / `--port` | host / integer 0–65535 | Browser `dev`/`run` default to `127.0.0.1:8000`; `mail serve` defaults to `127.0.0.1:8025` and admits numeric loopback only. |
+| `--public` | flag | `dev`; binds to `0.0.0.0` unless `--host` explicitly selects another address. |
 | `--target` | target id | `new`, `init`, native diagnostics, `build`, `run` |
 | `--format` / `--signing` | target-supported values | Native diagnostics, `build`, `run` |
 | `--output-root` | directory | Native `build` and `run` |
@@ -307,9 +308,9 @@ npm exec -- arcane upgrade --workspace . --app hello-world
 
 ### Overview
 
-Starts one loopback development server for one selected app and maps the exact
-workspace/runtime routes. It is a development convenience, not a production
-security boundary.
+Starts one development server for one selected app and maps the exact
+workspace/runtime routes. It defaults to localhost; `--public` enables access
+from other devices on the network.
 
 For an external workspace, the server exposes the selected projected
 `arcane/` root, including `arcane/sdk` and `arcane/dependencies`, alongside the
@@ -318,20 +319,35 @@ The explicit live-source SDK mapping remains unchanged and does not replace the
 installed projection.
 
 ```text
-arcane dev [--app <id>] [--host 127.0.0.1] [--port 8000]
+arcane dev [--app <id>] [--public] [--host <address>] [--port 8000]
 ```
 
 ### Lifecycle
 
 The command reports acceptance before bind/start work, emits the final URL,
 owns the server until cancellation, and restores failure to the process exit.
-The default host is loopback. Exposing another interface is an explicit
-development choice and does not add authentication.
+The default host is `127.0.0.1`. `--public` selects `0.0.0.0` (all IPv4
+interfaces); an explicit `--host` takes precedence. The command prints the
+local URL and available network URLs. Use a printed network URL on the other
+device, since `localhost` refers to that device and `0.0.0.0` is a bind address.
+Network URLs come from one interface snapshot at startup and do not establish
+remote reachability through the machine's firewall or network.
+
+This option changes the listener address. It does not configure a firewall,
+router forwarding, an internet tunnel, authentication, or HTTPS. Browser
+features that require a [secure context](https://developer.mozilla.org/en-US/docs/Web/Security/Defenses/Secure_Contexts)
+still require HTTPS when accessed through a LAN address.
 
 ### Example
 
 ```bash
 npm exec -- arcane dev --app hello-world --port 8000
+
+# Forward the public option through an existing dev script, for any app id.
+npm run dev -- --app hello-world --public
+
+# Equivalent direct CLI invocation, with an optional port.
+npm exec -- arcane dev --app hello-world --public --port 8000
 ```
 
 ## `arcane test`

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

@@ -5,9 +5,9 @@
     "repository": "https://github.com/TheWizardNexus/arcane-os-sdk.git",
     "branch": "main",
     "path": "runtime/arcane/components",
-    "sdkVersion": "0.5.18"
+    "sdkVersion": "0.6.2"
   },
-  "componentCount": 39,
+  "componentCount": 40,
   "loader": "/arcane/modules/HTMLImport.js",
   "artifacts": [
     {
@@ -76,6 +76,30 @@
       "transport": "HTMLImport + DOM; injected Arcane/provider modules where listed",
       "normalization": "DOM-normalized; caller/provider results remain external"
     },
+    {
+      "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.",
+      "methods": [
+        "refresh()",
+        "checkGpu()",
+        "open()",
+        "destroy()",
+        "ready"
+      ],
+      "events": [
+        "browser-ai-setup-ready"
+      ],
+      "slots": [],
+      "dependencies": [
+        "browser-device-settings.mjs",
+        "arcane-os/logging",
+        "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."
+    },
     {
       "file": "runtime/arcane/components/calculator.html",
       "name": "calculator.html",

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 availability and NPU setup instructions with browser-specific flags actions. | `refresh()`<br>`open()`<br>`destroy()`<br>`ready` | `browser-ai-setup-ready` | API presence only; settings navigation and hardware execution remain browser-owned |
+| [`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 |
 | [`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 |
@@ -158,15 +158,18 @@ Slots: `title`, `subtitle`, `identity`, `messages/message`, `composer`, `actions
 
 ### Overview
 
-Displays WebNN and WebGPU API availability for the current page and explains how
-to enable WebNN for NPU use. Its setup action uses the same browser-flags opening
-attempt and alert fallback as the existing high-performance GPU notice. Chrome
-and Edge receive their own flags addresses; other or unidentified browsers show
-both explicit choices.
+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,
+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.
 
 ### Public surface
 
-Methods/properties: `refresh()`, `open(browserId?)`, `destroy()`, `ready`.
+Methods/properties: `refresh()`, `checkGpu()`, `open(browserId?)`, `destroy()`, `ready`.
 
 Events: `browser-ai-setup-ready`.
 
@@ -176,6 +179,24 @@ the current settings record: `browserId`, `name`, `webnnFlagsURL`,
 after destruction. Availability means that the browser exposes the API to this
 page; it does not confirm a working adapter, loaded model, or hardware execution.
 
+`checkGpu()` returns a promise for one adapter-availability record. Concurrent
+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
+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
+`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
@@ -185,17 +206,19 @@ succeeded. Browsers may block internal-page navigation, so the full addresses
 also remain visible and copyable in the component.
 
 `destroy()` aborts owned listeners, disposes the event source, marks `ready`
-false, and suppresses UI updates from a pending clipboard operation. It returns
-`true` the first time and `false` thereafter. A BFCache-persisted `pagehide`
-preserves the component; nonpersisted `pagehide` destroys it.
+false, and suppresses UI updates from pending clipboard and adapter operations.
+It does not cancel the browser-owned adapter request. It returns `true` the
+first time and `false` thereafter. A BFCache-persisted `pagehide` preserves the
+component; nonpersisted `pagehide` destroys it.
 
 ### Availability and normalization
 
-**Browser and supported native WebViews.** API presence and setup instructions
-are normalized; browser flags, clipboard support, and device execution remain
-platform-owned. Mounting or refreshing creates no model, GPU adapter, or WebNN
-context. The component saves no preferences, changes no browser settings, and
-does not restart the browser. It cannot report model state owned by another page.
+**Browser and supported native WebViews.** API presence, adapter availability,
+and setup instructions are normalized; browser flag state, clipboard support,
+and model execution remain platform-owned. Mounting requests one WebGPU adapter
+when that API is available. No GPU device, model, or WebNN context is created.
+The component saves no preferences, changes no browser settings, and does not
+restart the browser. It cannot report model state owned by another page.
 
 ### Example
 

package/docs/reference/runtime-modules.md

@@ -3566,5 +3566,5 @@ console.log(Object.keys(module));
 ## Entity and component continuations
 
 - [Runtime entity modules](runtime-entities.md) explains all 14 modules, and [shared entity contracts](core/arcane-entities.md) owns all 29 exports.
-- [Runtime components](runtime-components.md) owns all 39 HTML-import fragments, methods, slots, and events.
+- [Runtime components](runtime-components.md) owns all 40 HTML-import fragments, methods, slots, and events.
 - [Arcane Ollama](arcane-ollama.md) expands the raw-versus-normalized behavior of `Ollama.js`.

package/docs/reference/sdk-api.md

@@ -3451,7 +3451,7 @@ async function useselectApp(...arguments_) {
 
 ### Overview
 
-Starts one owned browser development server with exact runtime/app route mappings and an unguessable session capability.
+Starts one owned browser development server with exact runtime/app route mappings and a caller-selected bind address.
 
 ### Signature, modes, and result
 
@@ -3464,19 +3464,23 @@ Import it from `arcane-os`. Source mode accepts
 signal, onEvent}` and serves one validated workspace application plus its
 complete SDK or integrated runtime. Packaged mode uses
 `{mode:'packaged', releaseRoot, host, port, signal, onEvent}` and serves the
-complete selected release files. `host` must be
-numeric loopback `127.0.0.1` or `::1`; port `0` asks the operating system for an
-available port.
+complete selected release files. `host` defaults to `127.0.0.1` and accepts an
+explicit network address or hostname. Use `0.0.0.0` for all IPv4 interfaces or
+`::` for the platform's IPv6 wildcard listener; port `0` asks the operating
+system for an available port.
 
 The promise settles after the listener is ready and resolves to
-`{server, mode, workspaceRoot, appId, host, port, origin, cleanUrl, url, close,
-closed, lifecycle}`. `server` is the raw Node HTTP server. `url` contains the
-unguessable bootstrap capability that establishes an HttpOnly session
-cookie; do not log, persist, or disclose it. `cleanUrl` contains no capability
-but does not bootstrap a new browser session by itself. In packaged mode,
-`workspaceRoot` and `appId` are `null`.
-
-Starting the server opens a loopback listener and emits awaited,
+`{server, mode, workspaceRoot, appId, host, port, origin, cleanUrl, url,
+networkUrls, close, closed, lifecycle}`. `server` is the raw Node HTTP server.
+`url` and `cleanUrl` are the same application URL. Wildcard listeners use
+`localhost` in that local URL; `host` retains the actual bound address.
+`networkUrls` lists application URLs for applicable non-loopback interface
+addresses discovered once at startup. These URLs are connection candidates,
+not evidence of reachability from another device. The server adds no session
+capability or authentication. In packaged mode, `workspaceRoot` and `appId`
+are `null`.
+
+Starting the server opens the selected listener and emits awaited,
 backpressured `server.starting` and `server.started` events. Request failures
 emit `server.request.failed`; shutdown emits `server.stopped` after owned
 requests and event delivery drain. Call `await result.close()` in a `finally`

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "arcane-os",
-  "version": "0.6.1",
+  "version": "0.7.0",
   "description": "Arcane OS JavaScript SDK, project-local CLI, browser runtime, and repository-portable application packager.",
   "type": "module",
   "main": "./src/index.mjs",

package/runtime/arcane/components/browser-ai-setup.html

@@ -8,9 +8,10 @@
     .flag-choice{margin-block:var(--arcane-space-4)}
     .flag-choice h3{margin-block:var(--arcane-space-2)}
     .actions{display:flex;flex-wrap:wrap;gap:var(--arcane-space-2)}
-    p,dt,dd,li,button{overflow-wrap:anywhere}
+    p,dt,dd,li,button,a{overflow-wrap:anywhere}
     button{white-space:normal}
-    code{display:block;white-space:pre-wrap;overflow-wrap:anywhere;user-select:text;margin-block:var(--arcane-space-2)}
+    .flag-address{display:block;white-space:pre-wrap;overflow-wrap:anywhere;user-select:text;margin-block:var(--arcane-space-2)}
+    #gpuAvailability{white-space:normal;overflow-wrap:anywhere;line-height:1.4;max-width:100%;box-sizing:border-box}
     li{margin-block:var(--arcane-space-2)}
 </style>
 
@@ -31,6 +32,21 @@
         <ol id="setupSteps"></ol>
         <p>Browser API availability does not confirm that an NPU or a particular model is supported. You choose whether to change the browser setting.</p>
         <p><a href="https://onnxruntime.ai/docs/tutorials/web/ep-webnn.html">ONNX Runtime WebNN setup documentation</a></p>
+        <h3>GPU performance</h3>
+        <p><span id="gpuAvailability" class="arcane-badge" role="status" aria-live="polite"></span></p>
+        <p id="gpuAdapterName" class="arcane-help"></p>
+        <button id="detectGpu" class="arcane-button arcane-button--secondary" type="button">Detect GPU</button>
+        <p id="gpuInstructions"></p>
+        <div id="gpuFlagAddress">
+            <a id="gpuFlagsLink" class="flag-address"></a>
+            <button id="copyGpuAddress" class="arcane-button arcane-button--secondary" type="button">Copy GPU flags address</button>
+            <ol>
+                <li>Open the address above. If the browser blocks the link, copy the address and paste it into the address bar.</li>
+                <li>Find “Force High Performance GPU” and choose Enabled.</li>
+                <li>Save your work, then completely close and reopen the browser before loading the model again.</li>
+            </ol>
+        </div>
+        <p class="arcane-help">GPU detection reports adapter availability. This page cannot read the flag's current setting or confirm that a model is using that adapter.</p>
         <p id="copyStatus" class="arcane-help" role="status" aria-live="polite"></p>
     </div>
     <footer class="arcane-card__footer">
@@ -46,6 +62,7 @@
 
     const {
         getBrowserDeviceSettings,
+        detectBrowserGpu,
         openBrowserDeviceSettings,
         WEBNN_BROWSER_SETTINGS
     }=await import('../sdk/ai/browser-device-settings.mjs');
@@ -58,6 +75,12 @@
     const lifecycleController=new AbortController();
     const status=root.querySelector('#setupStatus');
     const copyStatus=root.querySelector('#copyStatus');
+    const gpuAvailability=root.querySelector('#gpuAvailability');
+    const gpuAdapterName=root.querySelector('#gpuAdapterName');
+    const detectGpuButton=root.querySelector('#detectGpu');
+    const gpuFlagsLink=root.querySelector('#gpuFlagsLink');
+    const copyGpuButton=root.querySelector('#copyGpuAddress');
+    const copyButtons=[copyGpuButton];
     const choices=[];
     const setupSteps=[
         'If the flags page does not open, copy the matching address into your browser\'s address bar and open it.',
@@ -66,9 +89,11 @@
         'Return to this page and select Refresh browser availability.'
     ];
     let copyTask=null;
+    let gpuTask=null;
     let destroyed=false;
 
     host.refresh=refresh;
+    host.checkGpu=checkGpu;
     host.open=open;
     host.destroy=destroy;
 
@@ -85,18 +110,65 @@
             ?'Exposed to this page.'
             :'Unavailable to this page.';
         root.querySelector('#browserInstructions').textContent=settings.webnnFlagsURL
-            ?`Set up NPU attempts to open the WebNN flags page in ${settings.name} and shows the full instructions.`
+            ?`Open the ${settings.name} WebNN flags link below to set up NPU access.`
             :'A matching WebNN flags address is not known for this browser. Choose Google Chrome or Microsoft Edge below if that is the browser you are using.';
         for(const choice of choices){
             choice.row.hidden=Boolean(settings.webnnFlagsURL)
                 &&choice.target.url!==settings.webnnFlagsURL;
-            choice.setupButton.textContent=settings.webnnFlagsURL
-                ?'Set up NPU'
-                :`Set up NPU in ${choice.target.name}`;
+        }
+        const gpuTarget=settings.highPerformanceGpu;
+        root.querySelector('#gpuFlagAddress').hidden=!gpuTarget;
+        root.querySelector('#gpuInstructions').textContent=gpuTarget
+            ?`If this computer has multiple GPUs, use the ${gpuTarget.name} flags link to request the high-performance GPU when available.`
+            :'The Force High Performance GPU flag is documented here for desktop Windows Chromium browsers. No matching flags link is known for this browser and platform.';
+        if(gpuTarget){
+            gpuFlagsLink.href=gpuTarget.url;
+            gpuFlagsLink.textContent=gpuTarget.url;
+            gpuFlagsLink.setAttribute('aria-label',`${gpuTarget.name} Force High Performance GPU flags address: ${gpuTarget.url}`);
+        }else{
+            gpuFlagsLink.removeAttribute('href');
+            gpuFlagsLink.textContent='';
         }
         return settings;
     }
 
+    function checkGpu(){
+        if(destroyed)return Promise.resolve(false);
+        if(gpuTask)return gpuTask;
+        gpuAvailability.textContent='Detecting GPU...';
+        gpuAvailability.dataset.status='pending';
+        gpuAdapterName.textContent='';
+        detectGpuButton.disabled=true;
+        gpuTask=detect().finally(function finishGpuDetection(){
+            gpuTask=null;
+            if(!destroyed)detectGpuButton.disabled=false;
+        });
+        return gpuTask;
+
+        async function detect(){
+            try{
+                const result=await detectBrowserGpu();
+                if(destroyed)return false;
+                gpuAvailability.textContent=result.available
+                    ?result.isFallbackAdapter===true
+                        ?'Fallback WebGPU adapter detected.'
+                        :'GPU detected — WebGPU adapter available.'
+                    :'No WebGPU adapter available to this page.';
+                gpuAvailability.dataset.status=result.available&&result.isFallbackAdapter!==true
+                    ?'success':'warning';
+                gpuAdapterName.textContent=result.available?result.name:'';
+                return result;
+            }catch(error){
+                if(!destroyed){
+                    gpuAvailability.textContent='GPU detection failed. Select Detect GPU to try again.';
+                    gpuAvailability.dataset.status='error';
+                    arcaneLogging.error('Unable to detect the browser GPU adapter.',error);
+                }
+                return false;
+            }
+        }
+    }
+
     function open(browserId){
         if(destroyed)return false;
         const current=refresh();
@@ -105,7 +177,7 @@
             return browser.browserId===selectedId;
         });
         if(!target){
-            status.textContent='Choose the Google Chrome or Microsoft Edge setup button that matches your browser.';
+            status.textContent='Choose the Google Chrome or Microsoft Edge flags link that matches your browser.';
             return false;
         }
         const instructions=setupSteps.map(function numberSetupStep(step,index){
@@ -122,11 +194,11 @@
         if(destroyed)return Promise.resolve(false);
         if(copyTask)return copyTask;
         copyStatus.textContent='Copying address...';
-        for(const choice of choices)choice.copyButton.disabled=true;
+        for(const button of copyButtons)button.disabled=true;
         copyTask=writeAddress().finally(function finishAddressCopy(){
             copyTask=null;
             if(!destroyed){
-                for(const choice of choices)choice.copyButton.disabled=false;
+                for(const button of copyButtons)button.disabled=false;
             }
         });
         return copyTask;
@@ -140,7 +212,7 @@
             }catch(error){
                 if(!destroyed){
                     copyStatus.textContent='The address could not be copied automatically. Select the complete address above, copy it manually, and paste it into your browser\'s address bar.';
-                    arcaneLogging.error('Unable to copy the browser WebNN flags address.',error);
+                    arcaneLogging.error('Unable to copy the browser flags address.',error);
                 }
                 return false;
             }
@@ -151,32 +223,34 @@
     for(const target of WEBNN_BROWSER_SETTINGS){
         const row=document.createElement('section');
         const heading=document.createElement('h3');
-        const address=document.createElement('code');
+        const address=document.createElement('a');
         const actions=document.createElement('div');
-        const setupButton=document.createElement('button');
         const copyButton=document.createElement('button');
         row.className='flag-choice';
         heading.textContent=target.name;
+        address.className='flag-address';
+        address.href=target.url;
         address.textContent=target.url;
-        address.tabIndex=0;
-        address.setAttribute('aria-label',`${target.name} flags address`);
+        address.setAttribute('aria-label',`${target.name} WebNN flags address: ${target.url}`);
         actions.className='actions';
-        setupButton.type='button';
-        setupButton.className='arcane-button';
-        setupButton.addEventListener('click',function openSelectedBrowserSettings(){
-            return open(target.browserId);
-        },{signal:lifecycleController.signal});
         copyButton.type='button';
         copyButton.className='arcane-button arcane-button--secondary';
         copyButton.textContent=`Copy ${target.name} address`;
         copyButton.addEventListener('click',function copySelectedBrowserAddress(){
             return copyAddress(target.url);
         },{signal:lifecycleController.signal});
-        actions.append(setupButton,copyButton);
+        actions.append(copyButton);
         row.append(heading,address,actions);
         addressList.append(row);
-        choices.push({row,target,setupButton,copyButton});
+        choices.push({row,target});
+        copyButtons.push(copyButton);
     }
+
+    copyGpuButton.addEventListener('click',function copyGpuFlagsAddress(){
+        const target=getBrowserDeviceSettings().highPerformanceGpu;
+        if(target)return copyAddress(target.url);
+    },{signal:lifecycleController.signal});
+    detectGpuButton.addEventListener('click',checkGpu,{signal:lifecycleController.signal});
     const stepList=root.querySelector('#setupSteps');
     for(const step of setupSteps){
         const item=document.createElement('li');
@@ -187,6 +261,7 @@
     root.querySelector('#refreshAvailability').addEventListener('click',function refreshBrowserAvailability(){
         refresh();
         status.textContent='Browser API availability refreshed.';
+        return checkGpu();
     },{signal:lifecycleController.signal});
     host.ownerDocument.defaultView.addEventListener('pagehide',function disposeBeforePageUnload(event){
         if(!event.persisted)destroy();
@@ -204,6 +279,7 @@
     }
 
     refresh();
+    checkGpu();
     host.ready=true;
     const publication=events.dispatch('browser-ai-setup-ready',{}, {
         publicDetail:{ready:true}

package/src/cli/main.mjs

@@ -35,6 +35,7 @@ const VALUE_OPTIONS=new Set([
 ]);
 const FLAG_OPTIONS=new Set([
     'git',
+    'public',
     'skip-tests',
     'dry-run',
     'require-local-ai',
@@ -60,7 +61,7 @@ Usage:
   ${CLI_NAME} upgrade [--workspace <directory>] [--app <id>]
   ${CLI_NAME} doctor [--workspace <directory>] [--arcane-root <directory>]
   ${CLI_NAME} import-map [--workspace <directory>] [--app <id>]
-  ${CLI_NAME} dev [--app <id>] [--host 127.0.0.1] [--port 8000] [--sdk-runtime-source <sdk-root>]
+  ${CLI_NAME} dev [--app <id>] [--public] [--host <address>] [--port 8000] [--sdk-runtime-source <sdk-root>]
   ${CLI_NAME} test [--app <id>] [--scope app]
   ${CLI_NAME} test --scope shared --test-file <repo-relative.test.mjs>
   ${CLI_NAME} check [--app <id>] [--scope app] [--skip-tests]
@@ -83,6 +84,8 @@ Usage:
   ${CLI_NAME} mail serve --profile <profile> --from <address> --app <id> --origin <origin> [--allow-to <addresses>] [--app-key-stdin] [--host 127.0.0.1] [--port 8025] [--request-timeout <ms>]
 
 Development:
+  --public                      Bind all IPv4 interfaces (0.0.0.0) and print network URLs.
+  --host <address>               Override the bind address; takes precedence over --public.
   --sdk-runtime-source <sdk-root>  Dev-only live SDK checkout; omitted preserves the workspace runtime mode.
 
 Global:
@@ -445,6 +448,9 @@ function operationOptions(command,parsed,cwd){
     if(values['sdk-runtime-source']!==undefined&&command!=='dev'){
         usage('--sdk-runtime-source is supported only by dev.');
     }
+    if(flags.has('public')&&command!=='dev'){
+        usage('--public is supported only by dev.');
+    }
     if(flags.has('overwrite')&&command!=='bundle'){
         usage('--overwrite is supported only by bundle.');
     }
@@ -506,7 +512,7 @@ function operationOptions(command,parsed,cwd){
         noExtraPositionals(command,positionals);
         return {
             ...common,
-            host:values.host??'127.0.0.1',
+            host:values.host??(flags.has('public')?'0.0.0.0':'127.0.0.1'),
             port:readPort(values.port,8000),
             ...(values['sdk-runtime-source']===undefined?{}:{
                 sdkRuntimeSourceRoot:path.resolve(cwd,values['sdk-runtime-source'])
@@ -831,6 +837,7 @@ function serverSummary(result){
         host:result.host,
         port:result.port,
         url:result.url,
+        ...(result.networkUrls===undefined?{}:{networkUrls:result.networkUrls}),
         ...(result.callerAuthentication
             ?{callerAuthentication:result.callerAuthentication}
             :{}),
@@ -841,7 +848,11 @@ function serverSummary(result){
 }
 
 async function waitForServer(result,signal,reporter){
-    reporter.emit('server.ready',serverSummary(result),`Development server ready at ${result.url}`);
+    const readyMessage=[
+        `Development server ready at ${result.url}`,
+        ...(result.networkUrls??[]).map(function networkAddress(url){return `Network: ${url}`;})
+    ].join('\n');
+    reporter.emit('server.ready',serverSummary(result),readyMessage);
     if(result.lifecycle&&is.function(result.lifecycle.then)){
         const abort=()=>{
             void Promise.resolve().then(()=>result.close?.()).catch(()=>{});

package/src/dev-server.mjs

@@ -2,6 +2,7 @@ import Is from 'strong-type';
 import {constants as FS_CONSTANTS} from 'node:fs';
 import {lstat,open,readFile,realpath} from 'node:fs/promises';
 import http from 'node:http';
+import os from 'node:os';
 import path from 'node:path';
 import {resolveWorkspace} from './workspace.mjs';
 import {createEventQueue} from './event-queue.mjs';
@@ -489,6 +490,29 @@ function listen(server,{host,port,signal}){
     });
 }
 
+function browserHostname(host){
+    return host.includes(':')?`[${host}]`:host;
+}
+
+function networkUrlsForAddress(address,startPath){
+    if(address.address==='127.0.0.1'||address.address==='::1')return [];
+    const allIPv4=address.address==='0.0.0.0';
+    const allInterfaces=address.address==='::';
+    const urls=new Set();
+    for(const entries of Object.values(os.networkInterfaces())){
+        for(const entry of entries??[]){
+            if(entry.internal)continue;
+            // Scoped IPv6 addresses need the receiving device's interface,
+            // so they cannot provide a portable browser link for another device.
+            if(entry.family==='IPv6'&&entry.scopeid)continue;
+            if(allIPv4&&entry.family!=='IPv4')continue;
+            if(!allIPv4&&!allInterfaces&&entry.address!==address.address)continue;
+            urls.add(`http://${browserHostname(entry.address)}:${address.port}${startPath}`);
+        }
+    }
+    return [...urls];
+}
+
 async function startOwnedDevServer({
     workspaceRoot=process.cwd(),
     appId,
@@ -504,8 +528,8 @@ async function startOwnedDevServer({
     if(sdkRuntimeSourceRoot!==undefined&&mode!=='source'){
         fail('sdkRuntimeSourceRoot is supported only in source development mode.','ARCANE_USAGE');
     }
-    if(host!=='127.0.0.1'&&host!=='::1'){
-        fail('Development server host must be a numeric loopback address (127.0.0.1 or ::1).','ARCANE_POLICY_DENIED');
+    if(!is.string(host)||!host.trim()){
+        fail('Development server host must be a nonempty string.','ARCANE_USAGE');
     }
     if(!is.integer(port)||port<0||port>65535)fail('port must be an integer from 0 through 65535.','ARCANE_USAGE');
     const requestedRuntimeMode=mode==='source'&&sdkRuntimeSourceRoot!==undefined
@@ -631,7 +655,9 @@ async function startOwnedDevServer({
         server.close();
         fail('Development server did not expose a TCP address.');
     }
-    const visibleHost=address.family==='IPv6'?`[${address.address}]`:address.address;
+    const visibleHost=address.address==='0.0.0.0'||address.address==='::'
+        ?'localhost'
+        :browserHostname(address.address);
     const endpoint=new URL(`http://${visibleHost}:${address.port}`);
     const origin=endpoint.origin;
     const cleanUrl=`${origin}${routeSet.startPath}`;
@@ -728,17 +754,20 @@ async function startOwnedDevServer({
         origin,
         cleanUrl,
         url,
+        networkUrls:[],
         close,
         closed:lifecycle,
         lifecycle
     };
     try{
+        result.networkUrls=networkUrlsForAddress(address,routeSet.startPath);
         await events.send({
             type:'server.started',
             mode,
             host:result.host,
             port:result.port,
             url,
+            networkUrls:result.networkUrls,
             appId:result.appId,
             ...(routeSet.runtime?{
                 runtimeMode:routeSet.runtime.mode,