@codingame/monaco-vscode-chat-service-override 36.2.7 → 37.0.0

package/vscode/src/vs/workbench/contrib/agentsVoice/browser/agentsVoice.contribution.js

@@ -1,72 +1,91 @@
 
 import { __decorate, __param } from '@codingame/monaco-vscode-api/external/tslib/tslib.es6';
-import '../../chat/browser/voiceClient/micCaptureService.js';
+import '@codingame/monaco-vscode-api/vscode/vs/workbench/contrib/chat/browser/voiceClient/micCaptureService';
 import '../../chat/browser/voiceClient/ttsPlaybackService.js';
 import '../../chat/browser/voiceClient/voiceClientService.js';
+import { isVoiceEntitled } from '@codingame/monaco-vscode-katex-common/vscode/vs/workbench/contrib/chat/browser/voiceClient/voiceSessionController';
 import { IVoiceSessionController } from '@codingame/monaco-vscode-api/vscode/vs/workbench/contrib/chat/browser/voiceClient/voiceSessionController.service';
-import '../../chat/browser/voiceClient/voiceToolDispatchService.js';
+import { VOICE_AGENT_PROGRESS_SETTING, normalizeAgentsVoiceId } from '@codingame/monaco-vscode-katex-common/vscode/vs/workbench/contrib/chat/common/voiceClient/voiceClientService';
+import '@codingame/monaco-vscode-katex-common/vscode/vs/workbench/contrib/chat/browser/voiceClient/voiceToolDispatchService';
 import '../../chat/common/voicePlaybackService.js';
 import '../common/voiceTranscriptStore.js';
 import './transcriptsView/voiceTranscripts.contribution.js';
 import { Disposable } from '@codingame/monaco-vscode-api/vscode/vs/base/common/lifecycle';
 import '@codingame/monaco-vscode-api/vscode/vs/base/common/observableInternal/index';
 import { KeyMod, KeyCode } from '@codingame/monaco-vscode-api/vscode/vs/base/common/keyCodes';
+import { PolicyCategory } from '@codingame/monaco-vscode-api/vscode/vs/base/common/policy';
+import { URI } from '@codingame/monaco-vscode-api/vscode/vs/base/common/uri';
 import { localize2, localize } from '@codingame/monaco-vscode-api/vscode/vs/nls';
 import { registerAction2, Action2, MenuId } from '@codingame/monaco-vscode-api/vscode/vs/platform/actions/common/actions';
 import { Extensions, ConfigurationScope } from '@codingame/monaco-vscode-api/vscode/vs/platform/configuration/common/configurationRegistry';
 import { RawContextKey, ContextKeyExpr } from '@codingame/monaco-vscode-api/vscode/vs/platform/contextkey/common/contextkey';
 import { IContextKeyService } from '@codingame/monaco-vscode-api/vscode/vs/platform/contextkey/common/contextkey.service';
+import { SegmentedVoiceInputModePillInactive } from '@codingame/monaco-vscode-katex-common/vscode/vs/workbench/contrib/chat/browser/voiceInputMode/voiceInputModeContextKeys';
+import { IKeybindingService } from '@codingame/monaco-vscode-api/vscode/vs/platform/keybinding/common/keybinding.service';
 import { KeybindingWeight } from '@codingame/monaco-vscode-api/vscode/vs/platform/keybinding/common/keybindingsRegistry';
 import { Registry } from '@codingame/monaco-vscode-api/vscode/vs/platform/registry/common/platform';
 import { registerWorkbenchContribution2, WorkbenchPhase } from '@codingame/monaco-vscode-api/vscode/vs/workbench/common/contributions';
+import { Extensions as Extensions$1 } from '@codingame/monaco-vscode-api/vscode/vs/workbench/common/configuration';
 import { StorageScope, StorageTarget } from '@codingame/monaco-vscode-api/vscode/vs/platform/storage/common/storage';
 import { IStorageService } from '@codingame/monaco-vscode-api/vscode/vs/platform/storage/common/storage.service';
-import { AgentsVoiceStorageKeys } from '../common/agentsVoice.js';
+import { AGENTS_VOICE_ENTITLED, AGENTS_VOICE_CONNECTED, AGENTS_VOICE_CONNECTING, AGENTS_VOICE_LISTENING, AGENTS_VOICE_RECONNECTING, AGENTS_VOICE_ENABLED, AgentsVoiceSettingId, AgentsVoiceStorageKeys, getAgentsVoicePolicyValue } from '@codingame/monaco-vscode-api/vscode/vs/workbench/contrib/agentsVoice/common/agentsVoice';
 import { IConfigurationService } from '@codingame/monaco-vscode-api/vscode/vs/platform/configuration/common/configuration.service';
-import { IQuickInputService } from '@codingame/monaco-vscode-api/vscode/vs/platform/quickinput/common/quickInput.service';
+import { IChatEntitlementService } from '@codingame/monaco-vscode-api/vscode/vs/workbench/services/chat/common/chatEntitlementService.service';
 import { ITelemetryService } from '@codingame/monaco-vscode-api/vscode/vs/platform/telemetry/common/telemetry.service';
-import { mainWindow } from '@codingame/monaco-vscode-api/vscode/vs/base/browser/window';
+import { INotificationService } from '@codingame/monaco-vscode-api/vscode/vs/platform/notification/common/notification.service';
 import { Codicon } from '@codingame/monaco-vscode-api/vscode/vs/base/common/codicons';
 import { ChatContextKeys } from '@codingame/monaco-vscode-api/vscode/vs/workbench/contrib/chat/common/actions/chatContextKeys';
+import { EditorContextKeys } from '@codingame/monaco-vscode-api/vscode/vs/editor/common/editorContextKeys';
+import { getActiveWindow } from '@codingame/monaco-vscode-api/vscode/vs/base/browser/dom';
 import { ChatAgentLocation } from '@codingame/monaco-vscode-api/vscode/vs/workbench/contrib/chat/common/constants';
-import { IChatWidgetService } from '@codingame/monaco-vscode-api/vscode/vs/workbench/contrib/chat/browser/chat.service';
+import { ICommandService } from '@codingame/monaco-vscode-api/vscode/vs/platform/commands/common/commands.service';
+import { CONFIGURE_VOICE_INSTRUCTIONS_ACTION_ID } from '@codingame/monaco-vscode-katex-common/vscode/vs/workbench/contrib/chat/browser/actions/configureVoiceInstructionsAction';
+import { IVoiceModeOnboardingService } from '@codingame/monaco-vscode-api/vscode/vs/workbench/contrib/agentsVoice/browser/voiceModeOnboarding.service';
+import { SHOW_VOICE_MODE_ONBOARDING_COMMAND } from '@codingame/monaco-vscode-katex-common/vscode/vs/workbench/contrib/chat/browser/speechToText/micButtonMenuActions';
+import { IsSessionsWindowContext } from '@codingame/monaco-vscode-api/vscode/vs/workbench/common/contextkeys';
 import { autorun } from '@codingame/monaco-vscode-api/vscode/vs/base/common/observableInternal/reactions/autorun';
 
 var AgentsVoiceTelemetryContribution_1;
 const AGENTS_VOICE_WIDGET_FOCUSED = ( new RawContextKey("agentsVoiceWidgetFocused", false));
-const AGENTS_VOICE_CONNECTED = ( new RawContextKey("agentsVoiceConnected", false));
-const AGENTS_VOICE_CONNECTING = ( new RawContextKey("agentsVoiceConnecting", false));
-const AGENTS_VOICE_LISTENING = ( new RawContextKey("agentsVoiceListening", false));
-const AGENTS_VOICE_ACTIVE = ( new RawContextKey("agentsVoiceActive", false));
-const AGENTS_VOICE_INITIATED_HERE = ( new RawContextKey("agentsVoiceInitiatedHere", false));
+const AGENTS_VOICE_INITIATED_HERE = ( ContextKeyExpr.equals("agentsVoiceInitiatedHere", true));
+const VOICE_ACTIVE_ON_SURFACE = ( ContextKeyExpr.or(( IsSessionsWindowContext.negate()), AGENTS_VOICE_INITIATED_HERE));
+let AgentsVoiceEntitlementKeyContribution = class AgentsVoiceEntitlementKeyContribution extends Disposable {
+    static {
+        this.ID = "workbench.contrib.agentsVoiceEntitlementKey";
+    }
+    constructor(chatEntitlementService, contextKeyService) {
+        super();
+        const entitledKey = AGENTS_VOICE_ENTITLED.bindTo(contextKeyService);
+        const update = () => entitledKey.set(isVoiceEntitled(chatEntitlementService));
+        update();
+        this._register(chatEntitlementService.onDidChangeEntitlement(update));
+    }
+};
+AgentsVoiceEntitlementKeyContribution = ( __decorate([( __param(0, IChatEntitlementService)), ( __param(1, IContextKeyService))], AgentsVoiceEntitlementKeyContribution));
+registerWorkbenchContribution2(
+    AgentsVoiceEntitlementKeyContribution.ID,
+    AgentsVoiceEntitlementKeyContribution,
+    WorkbenchPhase.AfterRestored
+);
 let AgentsVoiceConnectedKeyContribution = class AgentsVoiceConnectedKeyContribution extends Disposable {
     static {
         this.ID = "workbench.contrib.agentsVoiceConnectedKey";
     }
-    constructor(voiceSessionController, contextKeyService, chatWidgetService) {
+    constructor(voiceSessionController, contextKeyService) {
         super();
         const connectedKey = AGENTS_VOICE_CONNECTED.bindTo(contextKeyService);
         const connectingKey = AGENTS_VOICE_CONNECTING.bindTo(contextKeyService);
         const listeningKey = AGENTS_VOICE_LISTENING.bindTo(contextKeyService);
-        const activeKey = AGENTS_VOICE_ACTIVE.bindTo(contextKeyService);
-        let wasConnected = false;
+        const reconnectingKey = AGENTS_VOICE_RECONNECTING.bindTo(contextKeyService);
         this._register(autorun(reader => {
-            const connected = voiceSessionController.isConnected.read(reader);
-            connectedKey.set(connected);
+            connectedKey.set(voiceSessionController.isConnected.read(reader));
             connectingKey.set(voiceSessionController.isConnecting.read(reader));
-            const state = voiceSessionController.voiceState.read(reader);
-            listeningKey.set(state === "listening");
-            activeKey.set(state === "listening" || state === "speaking");
-            if (wasConnected && !connected) {
-                for (const widget of chatWidgetService.getAllWidgets()) {
-                    AGENTS_VOICE_INITIATED_HERE.bindTo(widget.scopedContextKeyService).set(false);
-                }
-            }
-            wasConnected = connected;
+            reconnectingKey.set(voiceSessionController.isReconnecting.read(reader));
+            listeningKey.set(voiceSessionController.voiceState.read(reader) === "listening");
         }));
     }
 };
-AgentsVoiceConnectedKeyContribution = ( __decorate([( __param(0, IVoiceSessionController)), ( __param(1, IContextKeyService)), ( __param(2, IChatWidgetService))], AgentsVoiceConnectedKeyContribution));
+AgentsVoiceConnectedKeyContribution = ( __decorate([( __param(0, IVoiceSessionController)), ( __param(1, IContextKeyService))], AgentsVoiceConnectedKeyContribution));
 registerWorkbenchContribution2(
     AgentsVoiceConnectedKeyContribution.ID,
     AgentsVoiceConnectedKeyContribution,
@@ -119,16 +138,35 @@ registerWorkbenchContribution2(
     AgentsVoiceTelemetryContribution,
     WorkbenchPhase.AfterRestored
 );
+let AgentsVoiceOnboardingContribution = class AgentsVoiceOnboardingContribution extends Disposable {
+    static {
+        this.ID = "workbench.contrib.agentsVoiceOnboarding";
+    }
+    constructor(voiceSessionController, voiceModeOnboardingService) {
+        super();
+        this._register(autorun(reader => {
+            if (voiceSessionController.isConnecting.read(reader) || voiceSessionController.isConnected.read(reader)) {
+                voiceModeOnboardingService.showIfNeeded();
+            }
+        }));
+    }
+};
+AgentsVoiceOnboardingContribution = ( __decorate([( __param(0, IVoiceSessionController)), ( __param(1, IVoiceModeOnboardingService))], AgentsVoiceOnboardingContribution));
+registerWorkbenchContribution2(
+    AgentsVoiceOnboardingContribution.ID,
+    AgentsVoiceOnboardingContribution,
+    WorkbenchPhase.Eventually
+);
 registerAction2(class extends Action2 {
     constructor() {
         super({
             id: "agentsVoice.connecting",
-            title: ( localize2(4767, "Connecting...")),
-            icon: Codicon.loading,
-            precondition: ( ContextKeyExpr.and(( ContextKeyExpr.equals("config.agents.voice.enabled", true)), ( AGENTS_VOICE_CONNECTING.isEqualTo(true)))),
+            title: ( localize2(5067, "Connecting...")),
+            icon: Codicon.loadingCompact,
+            precondition: ( ContextKeyExpr.and(AGENTS_VOICE_ENABLED, ( ContextKeyExpr.or(( AGENTS_VOICE_CONNECTING.isEqualTo(true)), ( AGENTS_VOICE_RECONNECTING.isEqualTo(true)))))),
             menu: {
                 id: MenuId.ChatExecute,
-                when: ( ContextKeyExpr.and(( ContextKeyExpr.equals("config.agents.voice.enabled", true)), ( ChatContextKeys.location.isEqualTo(ChatAgentLocation.Chat)), ( AGENTS_VOICE_CONNECTING.isEqualTo(true)), ( AGENTS_VOICE_INITIATED_HERE.isEqualTo(true)))),
+                when: ( ContextKeyExpr.and(SegmentedVoiceInputModePillInactive, AGENTS_VOICE_ENABLED, ( ContextKeyExpr.notEquals(`config.${AgentsVoiceSettingId.ShowButton}`, false)), ( ChatContextKeys.location.isEqualTo(ChatAgentLocation.Chat)), ( ContextKeyExpr.or(( AGENTS_VOICE_CONNECTING.isEqualTo(true)), ( AGENTS_VOICE_RECONNECTING.isEqualTo(true)))), VOICE_ACTIVE_ON_SURFACE)),
                 group: "navigation",
                 order: -10
             }
@@ -140,89 +178,164 @@ registerAction2(class extends Action2 {
     constructor() {
         super({
             id: "agentsVoice.startVoiceInChat",
-            title: ( localize2(4768, "Voice Mode")),
-            icon: Codicon.voiceMode,
-            precondition: ( ContextKeyExpr.equals("config.agents.voice.enabled", true)),
+            title: ( localize2(5068, "Voice Mode")),
+            icon: Codicon.voiceModeCompact,
+            precondition: AGENTS_VOICE_ENABLED,
             menu: {
                 id: MenuId.ChatExecute,
-                when: ( ContextKeyExpr.and(( ContextKeyExpr.equals("config.agents.voice.enabled", true)), ( ChatContextKeys.location.isEqualTo(ChatAgentLocation.Chat)), ( ChatContextKeys.currentlyEditing.negate()), ( AGENTS_VOICE_ACTIVE.negate()), ( AGENTS_VOICE_CONNECTING.negate()))),
+                when: ( ContextKeyExpr.and(SegmentedVoiceInputModePillInactive, AGENTS_VOICE_ENABLED, ( ContextKeyExpr.notEquals(`config.${AgentsVoiceSettingId.ShowButton}`, false)), ( ChatContextKeys.location.isEqualTo(ChatAgentLocation.Chat)), ( ChatContextKeys.currentlyEditing.negate()), ( AGENTS_VOICE_LISTENING.negate()), ( AGENTS_VOICE_CONNECTING.negate()), ( AGENTS_VOICE_RECONNECTING.negate()), (
+                ChatContextKeys.speechToTextRecording.negate()), ( ChatContextKeys.speechToTextPreparing.negate()))),
                 group: "navigation",
                 order: -10
             },
             keybinding: {
                 weight: KeybindingWeight.WorkbenchContrib,
                 primary: KeyMod.CtrlCmd | KeyMod.Shift | KeyCode.Space,
-                when: ( ContextKeyExpr.and(( ContextKeyExpr.equals("config.agents.voice.enabled", true)), ChatContextKeys.inChatInput))
+                when: ( ContextKeyExpr.and(
+                    SegmentedVoiceInputModePillInactive,
+                    AGENTS_VOICE_ENABLED,
+                    ChatContextKeys.inChatInput
+                ))
             }
         });
     }
     async run(accessor) {
         const voiceController = accessor.get(IVoiceSessionController);
-        if (!voiceController.isConnected.get()) {
-            const chatWidgetService = accessor.get(IChatWidgetService);
-            const widget = chatWidgetService.lastFocusedWidget;
-            if (widget) {
-                AGENTS_VOICE_INITIATED_HERE.bindTo(widget.scopedContextKeyService).set(true);
-            }
-            await voiceController.connect(mainWindow);
-        } else {
-            voiceController.pttDown();
+        const keybindingService = accessor.get(IKeybindingService);
+        const handsFree = accessor.get(IConfigurationService).getValue("agents.voice.handsFree") === true;
+        const activeWindow = getActiveWindow();
+        voiceController.setActiveWindow(activeWindow);
+        const holdMode = keybindingService.enableKeybindingHoldMode("agentsVoice.startVoiceInChat");
+        const currentSession = await accessor.get(ICommandService).executeCommand("_chat.voice.getCurrentSession");
+        if (currentSession) {
+            try {
+                const resource = ( URI.parse(currentSession));
+                if (resource.scheme === "sessions-voice") {
+                    voiceController.setDraftTarget();
+                } else {
+                    voiceController.setTargetSession(resource);
+                    voiceController.activateSession(resource);
+                }
+            } catch {}
+        }
+        const wasConnected = voiceController.isConnected.get();
+        if (!wasConnected) {
+            await voiceController.connect(activeWindow);
+        }
+        if (!holdMode && !handsFree && !wasConnected) {
+            return;
+        }
+        voiceController.pttDown();
+        if (!holdMode) {
             voiceController.pttUp();
+            return;
         }
+        await holdMode;
+        voiceController.pttUp();
     }
 });
 registerAction2(class extends Action2 {
     constructor() {
         super({
             id: "agentsVoice.pttStopInChat",
-            title: ( localize2(4769, "Voice Mode: Stop Recording")),
-            icon: Codicon.voiceMode,
-            precondition: ( ContextKeyExpr.and(( ContextKeyExpr.equals("config.agents.voice.enabled", true)), ( AGENTS_VOICE_ACTIVE.isEqualTo(true)))),
+            title: ( localize2(5069, "Voice Mode: Stop Recording")),
+            icon: Codicon.voiceModeCompact,
+            precondition: ( ContextKeyExpr.and(AGENTS_VOICE_ENABLED, ( AGENTS_VOICE_LISTENING.isEqualTo(true)))),
             menu: {
                 id: MenuId.ChatExecute,
-                when: ( ContextKeyExpr.and(( ContextKeyExpr.equals("config.agents.voice.enabled", true)), ( ChatContextKeys.location.isEqualTo(ChatAgentLocation.Chat)), ( ChatContextKeys.currentlyEditing.negate()), ( AGENTS_VOICE_ACTIVE.isEqualTo(true)), ( AGENTS_VOICE_INITIATED_HERE.isEqualTo(true)))),
+                when: ( ContextKeyExpr.and(SegmentedVoiceInputModePillInactive, AGENTS_VOICE_ENABLED, ( ContextKeyExpr.notEquals(`config.${AgentsVoiceSettingId.ShowButton}`, false)), ( ChatContextKeys.location.isEqualTo(ChatAgentLocation.Chat)), ( ChatContextKeys.currentlyEditing.negate()), ( AGENTS_VOICE_LISTENING.isEqualTo(true)), VOICE_ACTIVE_ON_SURFACE)),
                 group: "navigation",
                 order: -10
-            },
-            keybinding: {
-                weight: KeybindingWeight.WorkbenchContrib,
-                primary: KeyMod.CtrlCmd | KeyMod.Shift | KeyCode.Space,
-                when: ( ContextKeyExpr.and(( ContextKeyExpr.equals("config.agents.voice.enabled", true)), ChatContextKeys.inChatInput, ( AGENTS_VOICE_ACTIVE.isEqualTo(true))))
             }
         });
     }
     async run(accessor) {
         const voiceController = accessor.get(IVoiceSessionController);
-        const configService = accessor.get(IConfigurationService);
-        const autoSendDelay = configService.getValue("agents.voice.autoSendDelay") ?? 500;
-        if (autoSendDelay >= 0) {
-            voiceController.disconnect();
-        } else {
-            voiceController.pttDown();
-            voiceController.pttUp();
-        }
+        voiceController.stopListening();
     }
 });
 registerAction2(class extends Action2 {
     constructor() {
         super({
             id: "agentsVoice.disconnect",
-            title: ( localize2(4770, "Disconnect Voice Mode")),
-            icon: Codicon.debugDisconnect,
+            title: ( localize2(5070, "Disconnect Voice Mode")),
+            icon: Codicon.debugDisconnectCompact,
             f1: true,
-            precondition: ( ContextKeyExpr.and(( ContextKeyExpr.equals("config.agents.voice.enabled", true)), ( AGENTS_VOICE_CONNECTED.isEqualTo(true))))
+            precondition: ( ContextKeyExpr.and(AGENTS_VOICE_ENABLED, ( AGENTS_VOICE_CONNECTED.isEqualTo(true)))),
+            menu: {
+                id: MenuId.ChatExecute,
+                when: ( ContextKeyExpr.and(AGENTS_VOICE_ENABLED, ( ContextKeyExpr.notEquals(`config.${AgentsVoiceSettingId.ShowButton}`, false)), ( ChatContextKeys.location.isEqualTo(ChatAgentLocation.Chat)), ( ChatContextKeys.currentlyEditing.negate()), ( AGENTS_VOICE_CONNECTED.isEqualTo(true)), VOICE_ACTIVE_ON_SURFACE,
+                SegmentedVoiceInputModePillInactive)),
+                group: "navigation",
+                order: -9
+            },
+            keybinding: {
+                weight: KeybindingWeight.EditorContrib - 5,
+                primary: KeyCode.Escape,
+                when: ( ContextKeyExpr.and(AGENTS_VOICE_ENABLED, ChatContextKeys.inChatInput, ( AGENTS_VOICE_CONNECTED.isEqualTo(true)), VOICE_ACTIVE_ON_SURFACE, (
+                ChatContextKeys.hasActiveRequest.negate()), ( EditorContextKeys.hoverVisible.toNegated()), ( EditorContextKeys.hasNonEmptySelection.toNegated()), ( EditorContextKeys.hasMultipleSelections.toNegated())))
+            }
         });
     }
     async run(accessor) {
         const voiceController = accessor.get(IVoiceSessionController);
-        voiceController.disconnect();
+        voiceController.disconnect("explicit");
+    }
+});
+registerAction2(class extends Action2 {
+    constructor() {
+        super({
+            id: "agentsVoice.cancelActiveRequest",
+            title: ( localize2(5071, "Voice Mode: Cancel Request")),
+            f1: false,
+            keybinding: {
+                weight: KeybindingWeight.EditorContrib - 5,
+                primary: KeyCode.Escape,
+                when: ( ContextKeyExpr.and(AGENTS_VOICE_ENABLED, ChatContextKeys.inChatInput, ( AGENTS_VOICE_CONNECTED.isEqualTo(true)),
+                ChatContextKeys.hasActiveRequest, ( EditorContextKeys.hoverVisible.toNegated()), ( EditorContextKeys.hasNonEmptySelection.toNegated()), ( EditorContextKeys.hasMultipleSelections.toNegated())))
+            }
+        });
+    }
+    async run(accessor) {
+        await accessor.get(ICommandService).executeCommand("workbench.action.chat.cancel");
+    }
+});
+registerAction2(class extends Action2 {
+    constructor() {
+        super({
+            id: "agentsVoice.openSettings",
+            title: ( localize2(5072, "Voice Mode Settings")),
+            f1: true,
+            precondition: AGENTS_VOICE_ENABLED
+        });
+    }
+    async run(accessor) {
+        const commandService = accessor.get(ICommandService);
+        await commandService.executeCommand("workbench.action.openSettings", {
+            query: "agents.voice"
+        });
+    }
+});
+registerAction2(class extends Action2 {
+    constructor() {
+        super({
+            id: SHOW_VOICE_MODE_ONBOARDING_COMMAND,
+            title: ( localize2(5073, "Voice Mode: Show Introduction")),
+            f1: true,
+            precondition: AGENTS_VOICE_ENABLED
+        });
+    }
+    run(accessor) {
+        if (!accessor.get(IVoiceModeOnboardingService).show()) {
+            accessor.get(INotificationService).info(( localize(5074, "Open a chat to see the Voice Mode introduction.")));
+        }
     }
 });
 registerAction2(class extends Action2 {
     constructor() {
         super({
             id: "agentsVoice.simulateConnection",
-            title: ( localize2(4771, "Voice: Simulate Connection (Dev)")),
+            title: ( localize2(5075, "Voice: Simulate Connection (Dev)")),
             f1: true
         });
     }
@@ -235,22 +348,23 @@ registerAction2(class extends Action2 {
     constructor() {
         super({
             id: "agentsVoice.resetOnboarding",
-            title: ( localize2(4772, "Voice: Reset Onboarding")),
+            title: ( localize2(5076, "Voice: Reset Onboarding")),
             f1: true
         });
     }
     async run(accessor) {
         const storageService = accessor.get(IStorageService);
         storageService.remove(AgentsVoiceStorageKeys.OnboardingCompleted, StorageScope.PROFILE);
+        storageService.remove(AgentsVoiceStorageKeys.IntroBannerShown, StorageScope.APPLICATION);
     }
 });
 registerAction2(class extends Action2 {
     constructor() {
         super({
             id: "agentsVoice.pushToTalk",
-            title: ( localize2(4773, "Voice Mode: Push to Talk")),
+            title: ( localize2(5077, "Voice Mode: Push to Talk")),
             f1: true,
-            precondition: ( ContextKeyExpr.equals("config.agents.voice.enabled", true)),
+            precondition: AGENTS_VOICE_ENABLED,
             keybinding: {
                 weight: KeybindingWeight.WorkbenchContrib,
                 primary: KeyMod.CtrlCmd | KeyMod.Shift | KeyCode.Space,
@@ -260,140 +374,213 @@ registerAction2(class extends Action2 {
     }
     async run(accessor) {
         const voiceController = accessor.get(IVoiceSessionController);
+        const keybindingService = accessor.get(IKeybindingService);
+        const holdMode = keybindingService.enableKeybindingHoldMode("agentsVoice.pushToTalk");
         if (!voiceController.isConnected.get() && !voiceController.isConnecting.get()) {
-            await voiceController.connect(mainWindow);
+            await voiceController.connect(getActiveWindow());
         }
         if (!voiceController.isConnected.get()) {
             return;
         }
         voiceController.pttDown();
-    }
-});
-registerAction2(class extends Action2 {
-    constructor() {
-        super({
-            id: "agentsVoice.selectMicrophone",
-            title: ( localize2(4774, "Voice: Select Microphone")),
-            f1: true,
-            precondition: ( ContextKeyExpr.equals("config.agents.voice.enabled", true))
-        });
-    }
-    async run(accessor) {
-        const quickInputService = accessor.get(IQuickInputService);
-        const storageService = accessor.get(IStorageService);
-        const devices = await navigator.mediaDevices.enumerateDevices();
-        const audioInputs = devices.filter(d => d.kind === "audioinput" && d.deviceId !== "default");
-        if (audioInputs.length === 0) {
-            quickInputService.pick([{
-                label: ( localize(4775, "No microphones found"))
-            }]);
+        if (!holdMode) {
+            voiceController.pttUp();
             return;
         }
-        const currentDeviceId = storageService.get(AgentsVoiceStorageKeys.MicrophoneDevice, StorageScope.APPLICATION, "");
-        const items = [];
-        items.push({
-            label: ( localize(4776, "System Default")),
-            description: currentDeviceId === "" ? ( localize(4777, "(current)")) : undefined,
-            deviceId: ""
-        });
-        for (const d of audioInputs) {
-            const label = d.label || ( localize(4778, "Unknown Device ({0})", d.deviceId.slice(0, 8)));
-            items.push({
-                label,
-                description: d.deviceId === currentDeviceId ? ( localize(4777, "(current)")) : undefined,
-                deviceId: d.deviceId
-            });
-        }
-        const picked = await quickInputService.pick(items, {
-            placeHolder: ( localize(4779, "Select a microphone for Voice Mode"))
-        });
-        if (picked) {
-            const selection = picked;
-            if (selection.deviceId) {
-                storageService.store(
-                    AgentsVoiceStorageKeys.MicrophoneDevice,
-                    selection.deviceId,
-                    StorageScope.APPLICATION,
-                    StorageTarget.MACHINE
-                );
-            } else {
-                storageService.remove(AgentsVoiceStorageKeys.MicrophoneDevice, StorageScope.APPLICATION);
-            }
-        }
+        await holdMode;
+        voiceController.pttUp();
     }
 });
 const configurationRegistry = ( Registry.as(Extensions.Configuration));
 configurationRegistry.registerConfiguration({
     id: "agentsVoice",
-    title: ( localize(4780, "Voice Mode")),
+    title: ( localize(5078, "Voice Mode")),
     type: "object",
     properties: {
         "agents.voice.enabled": {
             type: "boolean",
             description: ( localize(
-                4781,
+                5079,
                 "Enable the Voice Mode panel in the chat view for voice-driven coding conversations."
             )),
             default: false,
+            experiment: {
+                mode: "auto"
+            },
+            tags: ["experimental"],
             scope: ConfigurationScope.APPLICATION,
             restricted: true,
-            included: false
+            policy: {
+                name: "AgentsVoice",
+                category: PolicyCategory.InteractiveSession,
+                minimumVersion: "1.137",
+                value: getAgentsVoicePolicyValue,
+                localization: {
+                    description: {
+                        key: "agents.voice.enabled",
+                        value: ( localize(
+                            5079,
+                            "Enable the Voice Mode panel in the chat view for voice-driven coding conversations."
+                        ))
+                    }
+                }
+            }
+        },
+        [AgentsVoiceSettingId.ShowButton]: {
+            type: "boolean",
+            markdownDescription: ( localize(
+                5080,
+                "Controls whether the Voice Mode button is shown in the chat input. When hidden, Voice Mode can still be started with its keyboard shortcut."
+            )),
+            default: true,
+            tags: ["experimental"],
+            scope: ConfigurationScope.APPLICATION
         },
         "agents.voice.backendUrl": {
             type: "string",
             description: ( localize(
-                4782,
+                5081,
                 "Voice backend WebSocket URL. Leave empty to use the default hosted backend. Set to e.g. `ws://localhost:8000/api/v1/realtime/voice` to point at a backend running on your machine."
             )),
             default: "",
             scope: ConfigurationScope.APPLICATION,
             included: false
         },
-        "agents.voice.textToSpeech": {
+        "agents.voice.speakResponses": {
             type: "boolean",
-            description: ( localize(
-                4783,
-                "When enabled, the assistant reads responses aloud. When disabled, responses appear as text transcripts only."
+            markdownDescription: ( localize(
+                5082,
+                "When enabled, the assistant reads responses aloud. When disabled, responses are not spoken; enable `#agents.voice.showTranscript#` to read them as a text transcript instead."
             )),
             default: true,
-            scope: ConfigurationScope.APPLICATION,
-            included: false
+            scope: ConfigurationScope.APPLICATION
+        },
+        [VOICE_AGENT_PROGRESS_SETTING]: {
+            type: "boolean",
+            markdownDescription: ( localize(
+                5083,
+                "Allow Agent mode to speak brief semantic progress updates while it investigates, plans, edits, validates, or recovers from a problem."
+            )),
+            default: true,
+            tags: ["experimental"],
+            scope: ConfigurationScope.APPLICATION
+        },
+        "agents.voice.voice": {
+            type: "string",
+            enum: ["harper_neutral", "birch_neutral", "junho_neutral", "oak_neutral"],
+            enumItemLabels: ["Harper", "Birch", "Junho", "Oak"],
+            enumDescriptions: [( localize(5084, "Harper.")), ( localize(5085, "Birch.")), ( localize(5086, "Junho.")), ( localize(5087, "Oak."))],
+            markdownDescription: ( localize(
+                5088,
+                "The voice used when the assistant reads responses aloud. Changing this while voice mode is connected takes effect immediately. Use [Voice Mode instructions](command:{0}) to customize Voice Mode behavior and terminology.",
+                CONFIGURE_VOICE_INSTRUCTIONS_ACTION_ID
+            )),
+            default: "birch_neutral",
+            scope: ConfigurationScope.APPLICATION
+        },
+        "agents.voice.language": {
+            type: "string",
+            enum: ["auto", "en", "de", "es", "fr", "it", "pt", "ja", "ko", "zh"],
+            enumItemLabels: [( localize(5089, "Automatic")), ( localize(5090, "English")), ( localize(5091, "German")), ( localize(5092, "Spanish")), ( localize(5093, "French")), ( localize(5094, "Italian")), ( localize(5095, "Portuguese")), ( localize(5096, "Japanese")), ( localize(5097, "Korean")), ( localize(5098, "Chinese"))],
+            markdownDescription: ( localize(
+                5099,
+                "The language used for speech recognition, dictation, and spoken responses. The selectable languages support native voice output. Automatic uses the configured display language for speech recognition and dictation when supported; otherwise, it follows the system or browser locale. English voice output is used when the detected language does not support native voice output. Changing this while voice mode is connected takes effect immediately."
+            )),
+            default: "auto",
+            scope: ConfigurationScope.APPLICATION
         },
         "agents.voice.showTranscript": {
             type: "boolean",
-            description: ( localize(
-                4784,
-                "Show the voice transcript overlay in the chat input area while voice mode is active."
+            markdownDescription: ( localize(
+                5100,
+                "Show the voice transcript overlay in the chat input area while voice mode is active. Enable this to read responses as text when `#agents.voice.speakResponses#` is disabled."
             )),
             default: false,
-            scope: ConfigurationScope.APPLICATION,
-            included: false,
-            tags: ["advanced"]
+            scope: ConfigurationScope.APPLICATION
+        },
+        "agents.voice.liveTranscript": {
+            type: "boolean",
+            markdownDescription: ( localize(
+                5101,
+                "Show your speech as a live, word-by-word transcript while you are speaking. When disabled, your transcript appears only once you finish speaking. Requires `#agents.voice.showTranscript#` to be enabled to be visible."
+            )),
+            default: false,
+            scope: ConfigurationScope.APPLICATION
+        },
+        "agents.voice.handsFree": {
+            type: "boolean",
+            markdownDescription: ( localize(
+                5102,
+                "When enabled, voice mode automatically re-enters listening after the assistant finishes speaking, so you can hold a hands-free back-and-forth conversation. When disabled, you start and end each turn manually, and ending the turn sends it. Turns are not ended automatically on trailing silence or a stop phrase unless {0} or {1} is explicitly configured.",
+                "`#agents.voice.turn.silenceMs#`",
+                "`#agents.voice.turn.stopPhrases#`"
+            )),
+            default: true,
+            scope: ConfigurationScope.APPLICATION
         },
-        "agents.voice.autoSendDelay": {
+        "agents.voice.turn.silenceMs": {
             type: "number",
-            description: ( localize(
-                4785,
-                "In toggle voice mode (short tap), automatically finish recording after this many milliseconds of silence. Set to -1 to disable."
+            markdownDescription: ( localize(
+                5103,
+                "Trailing silence in milliseconds before the backend ends the turn automatically. Set to `-1` to disable ending the turn on silence, in which case the turn ends only via a stop phrase ({0}) or manually. When enabled, the backend clamps this to its supported range (currently 200-5000 ms) and is the source of truth. When hands-free mode ({1}) is disabled, the turn is not ended on silence by default unless this setting is explicitly configured, so you keep manual control over when a turn is sent.",
+                "`#agents.voice.turn.stopPhrases#`",
+                "`#agents.voice.handsFree#`"
             )),
-            default: 500,
-            minimum: -1,
-            scope: ConfigurationScope.APPLICATION,
-            included: false,
-            tags: ["advanced"]
+            default: 800,
+            anyOf: [{
+                const: -1,
+                description: ( localize(5104, "Do not end the turn on trailing silence."))
+            }, {
+                type: "number",
+                minimum: 200,
+                maximum: 5000
+            }],
+            scope: ConfigurationScope.APPLICATION
         },
-        "agents.voice.sendKeyword": {
-            type: "string",
-            description: ( localize(
-                4786,
-                "A keyword phrase (e.g. \"send it\") that, when spoken at the end of an utterance in toggle mode, triggers sending the request immediately. The keyword is stripped from the sent message. Leave empty to disable."
+        "agents.voice.turn.stopPhrases": {
+            type: "array",
+            items: {
+                type: "string"
+            },
+            markdownDescription: ( localize(
+                5105,
+                "Phrases that end the turn when spoken at the end of an utterance. Leave empty to disable ending the turn on a stop phrase, in which case the turn ends only on trailing silence ({0}) or manually. The backend strips the matched phrase from the transcript before it reaches the agent. When hands-free mode ({1}) is disabled, stop phrases do not end the turn by default unless this setting is explicitly configured, so you keep manual control over when a turn is sent.",
+                "`#agents.voice.turn.silenceMs#`",
+                "`#agents.voice.handsFree#`"
             )),
-            default: "",
-            scope: ConfigurationScope.APPLICATION,
-            included: false,
-            tags: ["advanced"]
+            default: ["send it"],
+            scope: ConfigurationScope.APPLICATION
         }
     }
 });
+( Registry.as(Extensions$1.ConfigurationMigration)).registerConfigurationMigrations([{
+    key: "agents.voice.voice",
+    includeApplication: true,
+    migrateFn: value => ({
+        value: normalizeAgentsVoiceId(value)
+    })
+}, {
+    key: "agents.voice.turn.autoEndMode",
+    migrateFn: value => {
+        const result = [["agents.voice.turn.autoEndMode", {
+            value: undefined
+        }]];
+        if (value === "off" || value === "vad" || value === "phrase" || value === "both") {
+            const silenceEnabled = value === "vad" || value === "both";
+            const phraseEnabled = value === "phrase" || value === "both";
+            if (!silenceEnabled) {
+                result.push(["agents.voice.turn.silenceMs", {
+                    value: -1
+                }]);
+            }
+            if (!phraseEnabled) {
+                result.push(["agents.voice.turn.stopPhrases", {
+                    value: []
+                }]);
+            }
+        }
+        return result;
+    }
+}]);
 
 export { AGENTS_VOICE_WIDGET_FOCUSED };