feat(desktop): add Voice Control dictation

apps/desktop/package.json

@@ -265,6 +265,7 @@
 		"electron": "40.8.5",
 		"electron-builder": "26.8.1",
 		"electron-vite": "4.0.1",
+		"happy-dom": "20.9.0",
 		"material-icon-theme": "5.32.0",
 		"rimraf": "6.1.3",
 		"rollup-plugin-inject-process-env": "1.3.1",

apps/desktop/src/lib/trpc/routers/index.ts

@@ -27,6 +27,7 @@ import { createSettingsRouter } from "./settings";
 import { createSystemRouter } from "./system";
 import { createTerminalRouter } from "./terminal";
 import { createUiStateRouter } from "./ui-state";
+import { createVoiceInputRouter } from "./voice-input";
 import { createWindowRouter } from "./window";
 import { createWorkspacesRouter } from "./workspaces";
 
@@ -61,6 +62,7 @@ export const createAppRouter = (getWindow: () => BrowserWindow | null) => {
 		hostServiceCoordinator: createHostServiceCoordinatorRouter(),
 		keyboardLayout: createKeyboardLayoutRouter(),
 		migration: createMigrationRouter(),
+		voiceInput: createVoiceInputRouter(),
 	});
 };
 

apps/desktop/src/lib/trpc/routers/permissions.test.ts

@@ -0,0 +1,59 @@
+import { beforeEach, describe, expect, it, mock } from "bun:test";
+
+const askForMediaAccessMock = mock(async () => false);
+const getMediaAccessStatusMock = mock(() => "not-determined");
+const isTrustedAccessibilityClientMock = mock(() => false);
+const openExternalMock = mock(async () => undefined);
+
+mock.module("electron", () => ({
+	app: {
+		relaunch: mock(() => {}),
+	},
+	shell: {
+		openExternal: openExternalMock,
+	},
+	systemPreferences: {
+		askForMediaAccess: askForMediaAccessMock,
+		getMediaAccessStatus: getMediaAccessStatusMock,
+		isTrustedAccessibilityClient: isTrustedAccessibilityClientMock,
+	},
+}));
+
+const { createPermissionsRouter } = await import("./permissions");
+
+function createCaller() {
+	return createPermissionsRouter().createCaller({});
+}
+
+describe("permissions router", () => {
+	beforeEach(() => {
+		askForMediaAccessMock.mockClear();
+		getMediaAccessStatusMock.mockClear();
+		isTrustedAccessibilityClientMock.mockClear();
+		openExternalMock.mockClear();
+	});
+
+	it("doesNotRequestMicrophoneOnStatusRead", async () => {
+		const caller = createCaller();
+
+		const status = await caller.getStatus();
+
+		expect(status.microphone).toBe(false);
+		expect(status.microphoneStatus).toBe("promptable");
+		expect(getMediaAccessStatusMock).toHaveBeenCalledWith("microphone");
+		expect(askForMediaAccessMock).not.toHaveBeenCalled();
+		expect(openExternalMock).not.toHaveBeenCalled();
+	});
+
+	it("requestsMicrophoneOnlyThroughExplicitMutation", async () => {
+		const caller = createCaller();
+
+		await caller.requestMicrophone();
+
+		if (process.platform === "darwin") {
+			expect(askForMediaAccessMock).toHaveBeenCalledWith("microphone");
+		} else {
+			expect(openExternalMock).toHaveBeenCalledTimes(1);
+		}
+	});
+});

apps/desktop/src/lib/trpc/routers/permissions/native-permissions.test.ts

@@ -1,6 +1,9 @@
 import { describe, expect, it, mock } from "bun:test";
 
 mock.module("electron", () => ({
+	app: {
+		relaunch: mock(() => {}),
+	},
 	shell: {
 		openExternal: mock(async () => {}),
 	},

apps/desktop/src/lib/trpc/routers/permissions/native-permissions.ts

@@ -22,6 +22,11 @@ type SystemPreferencesApi = Pick<
 	typeof electronSystemPreferences,
 	"askForMediaAccess" | "getMediaAccessStatus" | "isTrustedAccessibilityClient"
 >;
+export type MicrophonePermissionStatus =
+	| "granted"
+	| "denied"
+	| "promptable"
+	| "unknown";
 
 function getElectronShell(): ShellApi {
 	return (require("electron") as Partial<typeof import("electron")>)
@@ -49,20 +54,41 @@ export function checkMicrophone({
 }: {
 	systemPreferencesApi?: Pick<SystemPreferencesApi, "getMediaAccessStatus">;
 } = {}): boolean {
+	return getMicrophonePermissionStatus({ systemPreferencesApi }) === "granted";
+}
+
+export function getMicrophonePermissionStatus({
+	systemPreferencesApi = getElectronSystemPreferences(),
+}: {
+	systemPreferencesApi?: Pick<SystemPreferencesApi, "getMediaAccessStatus">;
+} = {}): MicrophonePermissionStatus {
 	try {
-		return (
-			systemPreferencesApi?.getMediaAccessStatus("microphone") === "granted"
-		);
+		const status = systemPreferencesApi?.getMediaAccessStatus("microphone");
+
+		if (status === "granted") {
+			return "granted";
+		}
+		if (status === "denied" || status === "restricted") {
+			return "denied";
+		}
+		if (status === "not-determined") {
+			return "promptable";
+		}
+
+		return "unknown";
 	} catch {
-		return false;
+		return "unknown";
 	}
 }
 
 export function getPermissionStatus() {
+	const microphoneStatus = getMicrophonePermissionStatus();
+
 	return {
 		fullDiskAccess: checkFullDiskAccess(),
 		accessibility: checkAccessibility(),
-		microphone: checkMicrophone(),
+		microphone: microphoneStatus === "granted",
+		microphoneStatus,
 	};
 }
 

apps/desktop/src/lib/trpc/routers/settings/index.ts

@@ -50,6 +50,7 @@ import {
 	DEFAULT_SHOW_RESOURCE_MONITOR,
 	DEFAULT_TERMINAL_LINK_BEHAVIOR,
 	DEFAULT_USE_COMPACT_TERMINAL_ADD_BUTTON,
+	DEFAULT_VOICE_INPUT_ENABLED,
 } from "shared/constants";
 import { normalizePresetProjectIds } from "shared/preset-project-targeting";
 import {
@@ -260,8 +261,25 @@ export function getPresetsForTrigger(
 	);
 }
 
+function getSettingsPayload() {
+	const row = getSettings();
+	return {
+		confirmOnQuit: row.confirmOnQuit ?? DEFAULT_CONFIRM_ON_QUIT,
+		fileOpenMode: row.fileOpenMode ?? DEFAULT_FILE_OPEN_MODE,
+		openLinksInApp: row.openLinksInApp ?? DEFAULT_OPEN_LINKS_IN_APP,
+		showResourceMonitor:
+			row.showResourceMonitor ?? DEFAULT_SHOW_RESOURCE_MONITOR,
+		terminalLinkBehavior:
+			row.terminalLinkBehavior ?? DEFAULT_TERMINAL_LINK_BEHAVIOR,
+		voiceInputEnabled: row.voiceInputEnabled ?? DEFAULT_VOICE_INPUT_ENABLED,
+	};
+}
+
 export const createSettingsRouter = () => {
 	return router({
+		getSettings: publicProcedure
+			.input(z.void())
+			.query(() => getSettingsPayload()),
 		getTerminalPresets: publicProcedure.query(() => {
 			const row = getSettings();
 			if (!row.terminalPresetsInitialized) {
@@ -925,6 +943,25 @@ export const createSettingsRouter = () => {
 			return row.showResourceMonitor ?? DEFAULT_SHOW_RESOURCE_MONITOR;
 		}),
 
+		getVoiceInputEnabled: publicProcedure.input(z.void()).query(() => {
+			return getSettingsPayload().voiceInputEnabled;
+		}),
+
+		setVoiceInputEnabled: publicProcedure
+			.input(z.object({ enabled: z.boolean() }))
+			.mutation(({ input }) => {
+				localDb
+					.insert(settings)
+					.values({ id: 1, voiceInputEnabled: input.enabled })
+					.onConflictDoUpdate({
+						target: settings.id,
+						set: { voiceInputEnabled: input.enabled },
+					})
+					.run();
+
+				return { success: true };
+			}),
+
 		setShowResourceMonitor: publicProcedure
 			.input(z.object({ enabled: z.boolean() }))
 			.mutation(({ input }) => {

apps/desktop/src/lib/trpc/routers/settings/voice-input.test.ts

@@ -0,0 +1,178 @@
+import { Database } from "bun:sqlite";
+import { beforeEach, describe, expect, it, mock } from "bun:test";
+import { readFileSync } from "node:fs";
+import { resolve } from "node:path";
+import { drizzle } from "drizzle-orm/bun-sqlite";
+import * as schema from "../../../../../../../packages/local-db/src";
+
+const sqlite = new Database(":memory:");
+
+function createPreVoiceInputSettingsTable() {
+	sqlite.exec(`
+	DROP TABLE IF EXISTS settings;
+	CREATE TABLE settings (
+		id integer PRIMARY KEY DEFAULT 1,
+		last_active_workspace_id text,
+		terminal_presets text,
+		terminal_presets_initialized integer,
+		agent_preset_overrides text,
+		agent_custom_definitions text,
+		agent_preset_permissions_migrated_at integer,
+		selected_ringtone_id text,
+		active_organization_id text,
+		confirm_on_quit integer,
+		terminal_link_behavior text,
+		persist_terminal integer DEFAULT true,
+		auto_apply_default_preset integer,
+		branch_prefix_mode text,
+		branch_prefix_custom text,
+		notification_sounds_muted integer,
+		notification_volume integer,
+		delete_local_branch integer,
+		file_open_mode text,
+		show_presets_bar integer,
+		use_compact_terminal_add_button integer,
+		terminal_font_family text,
+		terminal_font_size integer,
+		editor_font_family text,
+		editor_font_size integer,
+		show_resource_monitor integer,
+		worktree_base_dir text,
+		open_links_in_app integer,
+		default_editor text,
+		expose_host_service_via_relay integer
+	);
+`);
+}
+
+function applyVoiceInputMigration() {
+	const migrationSql = readFileSync(
+		resolve(
+			process.cwd(),
+			"packages/local-db/drizzle/0042_add_voice_input_enabled.sql",
+		),
+		"utf8",
+	);
+
+	for (const statement of migrationSql.split("--> statement-breakpoint")) {
+		const trimmedStatement = statement.trim();
+		if (trimmedStatement.length > 0) {
+			sqlite.exec(trimmedStatement);
+		}
+	}
+}
+
+const testLocalDb = drizzle(sqlite, { schema });
+
+mock.module("@superset/local-db", () => schema);
+
+const getHostServiceCoordinatorMock = mock(() => ({
+	getActiveOrganizationIds: () => [],
+	restartAll: async () => {},
+}));
+const loadTokenMock = mock(async () => ({ token: null }));
+
+mock.module("electron", () => ({
+	app: {
+		relaunch: mock(() => {}),
+	},
+	shell: {
+		openExternal: mock(async () => undefined),
+	},
+	systemPreferences: {
+		askForMediaAccess: mock(async () => false),
+		getMediaAccessStatus: mock(() => "not-determined"),
+		isTrustedAccessibilityClient: mock(() => false),
+	},
+}));
+
+mock.module("main/env.main", () => ({
+	env: {
+		NODE_ENV: "test",
+		NEXT_PUBLIC_API_URL: "https://api.superset.test",
+	},
+}));
+
+mock.module("main/index", () => ({
+	exitImmediately: mock(() => {}),
+}));
+
+mock.module("main/lib/custom-ringtones", () => ({
+	hasCustomRingtone: mock(() => false),
+}));
+
+mock.module("main/lib/host-service-coordinator", () => ({
+	getHostServiceCoordinator: getHostServiceCoordinatorMock,
+}));
+
+mock.module("main/lib/local-db", () => ({
+	localDb: testLocalDb,
+}));
+
+mock.module("../auth/utils/auth-functions", () => ({
+	loadToken: loadTokenMock,
+}));
+
+mock.module("../workspaces/utils/git", () => ({
+	NotGitRepoError: class NotGitRepoError extends Error {},
+	getGitAuthorName: mock(async () => null),
+	getGitHubUsername: mock(async () => null),
+}));
+
+const { createSettingsRouter } = await import("./index");
+
+function createCaller() {
+	return createSettingsRouter().createCaller({});
+}
+
+describe("voice input settings", () => {
+	beforeEach(() => {
+		createPreVoiceInputSettingsTable();
+		applyVoiceInputMigration();
+		getHostServiceCoordinatorMock.mockClear();
+		loadTokenMock.mockClear();
+	});
+
+	it("returnsDefaultDisabledVoiceInputSetting", async () => {
+		const caller = createCaller();
+
+		expect(await caller.getVoiceInputEnabled()).toBe(false);
+	});
+
+	it("persistsVoiceInputEnabledSetting", async () => {
+		const caller = createCaller();
+
+		await caller.setVoiceInputEnabled({ enabled: true });
+		expect(await caller.getVoiceInputEnabled()).toBe(true);
+
+		await caller.setVoiceInputEnabled({ enabled: false });
+		expect(await caller.getVoiceInputEnabled()).toBe(false);
+	});
+
+	it("exposesVoiceInputInGetSettings", async () => {
+		const caller = createCaller();
+
+		await caller.setVoiceInputEnabled({ enabled: true });
+
+		expect(await caller.getSettings()).toMatchObject({
+			confirmOnQuit: true,
+			fileOpenMode: "split-pane",
+			openLinksInApp: false,
+			showResourceMonitor: true,
+			terminalLinkBehavior: "file-viewer",
+			voiceInputEnabled: true,
+		});
+	});
+
+	it("keepsVoiceSettingsLocalOnly", async () => {
+		const caller = createCaller();
+
+		await caller.setVoiceInputEnabled({ enabled: true });
+		expect(await caller.getVoiceInputEnabled()).toBe(true);
+
+		const payload = await caller.getSettings();
+		expect(payload).not.toHaveProperty("activeOrganizationId");
+		expect(getHostServiceCoordinatorMock).not.toHaveBeenCalled();
+		expect(loadTokenMock).not.toHaveBeenCalled();
+	});
+});