Address review feedback on reading responses aloud

- Bind the stored key to the endpoint it was given for, so that pointing the
  setting at another host stops reading aloud until a key for that host is
  entered, rather than sending this one to it. The setup action writes the
  endpoint before asking for the replacement key, which left a window where
  the two did not belong together.

- Register the endpoint and voice settings statically. They were registered
  only once speech was already available, so on a platform without a speech
  synthesizer the setup action could not write the endpoint, and the engine
  could never become available at all.

- Strip keycap sequences that omit the variation selector, which the Unicode
  grammar allows: `1\u20E3` was read out as "one".

- Mention reading aloud in the chat accessibility help, including how to stop.

Author: federicobrancasi

src/vs/workbench/contrib/accessibility/browser/accessibilityConfiguration.ts

@@ -922,6 +922,22 @@ export function registerAccessibilityConfiguration() {
 				tags: ['accessibility'],
 				scope: ConfigurationScope.APPLICATION,
 			},
+			[AccessibilityVoiceSettingId.MaiSpeechEndpoint]: {
+				'markdownDescription': localize('voice.maiSpeechEndpoint', "The endpoint of the speech service used to read text aloud, for example `https://eastus2.tts.speech.microsoft.com`. Run `Speech: Set Up Read Aloud` to configure it together with its key, which is stored securely rather than in your settings. Note that the text being read is sent to this service."),
+				'type': 'string',
+				'default': '',
+				// Application scope so that a workspace cannot point reading aloud
+				// at another server and have the key sent there.
+				'scope': ConfigurationScope.APPLICATION,
+				'tags': ['accessibility', 'usesOnlineServices']
+			},
+			[AccessibilityVoiceSettingId.MaiVoice]: {
+				'markdownDescription': localize('voice.maiVoice', "The voice used to read text aloud, for example `en-US-Harper:MAI-Voice-2`. Leave empty to pick a voice for {0} automatically.", `\`#${AccessibilityVoiceSettingId.SpeechLanguage}#\``),
+				'type': 'string',
+				'default': '',
+				'scope': ConfigurationScope.APPLICATION,
+				'tags': ['accessibility']
+			},
 			[AccessibilityWorkbenchSettingId.HideAccessibleView]: {
 				description: localize('accessibility.hideAccessibleView', "Controls whether the Accessible View is hidden."),
 				type: 'boolean',
@@ -998,22 +1014,6 @@ export class DynamicSpeechAccessibilityConfiguration extends Disposable implemen
 					'enumDescriptions': languagesSorted.map(key => languages[key].name),
 					'enumItemLabels': languagesSorted.map(key => languages[key].name)
 				},
-				[AccessibilityVoiceSettingId.MaiSpeechEndpoint]: {
-					'markdownDescription': localize('voice.maiSpeechEndpoint', "The endpoint of the speech service used to read text aloud, for example `https://eastus2.tts.speech.microsoft.com`. Run `Speech: Set Up Read Aloud` to configure it together with its key, which is stored securely rather than in your settings. Note that the text being read is sent to this service."),
-					'type': 'string',
-					'default': '',
-					// Application scope so that a workspace cannot point reading aloud
-					// at another server and have the key sent there.
-					'scope': ConfigurationScope.APPLICATION,
-					'tags': ['accessibility', 'usesOnlineServices']
-				},
-				[AccessibilityVoiceSettingId.MaiVoice]: {
-					'markdownDescription': localize('voice.maiVoice', "The voice used to read text aloud, for example `en-US-Harper:MAI-Voice-2`. Leave empty to pick a voice for {0} automatically.", `\`#${AccessibilityVoiceSettingId.SpeechLanguage}#\``),
-					'type': 'string',
-					'default': '',
-					'scope': ConfigurationScope.APPLICATION,
-					'tags': ['accessibility']
-				},
 				[AccessibilityVoiceSettingId.AutoSynthesize]: {
 					'type': 'string',
 					'enum': ['on', 'off'],

src/vs/workbench/contrib/chat/browser/actions/chatAccessibilityHelp.ts

@@ -176,6 +176,7 @@ export function getAccessibilityHelpText(type: 'panelChat' | 'inlineChat' | 'qui
 		content.push(localize('chat.find', 'To search the chat transcript, invoke Find in Chat{0}. Find Next{1} and Find Previous{2} move between results, scrolling each one into view.', '<keybinding:workbench.action.chat.find>', '<keybinding:workbench.action.chat.findNext>', '<keybinding:workbench.action.chat.findPrevious>'));
 	}
 	content.push(localize('chat.attachments.pastedText', "Long pasted text is stored as an attached text item and replaced in the input with a numbered inline reference."));
+	content.push(localize('chat.readAloud', "A response can be read aloud with Read Aloud{0}, which is also the first button in the response toolbar. Reading stops with Stop Read Aloud{1}, or by starting to read another response.", '<keybinding:workbench.action.chat.readChatResponseAloud>', '<keybinding:workbench.action.chat.stopReadChatItemAloud>'));
 	content.push(localize('chat.signals', "Accessibility Signals can be changed via settings with a prefix of signals.chat. By default, if a request takes more than 4 seconds, you will hear a sound indicating that progress is still occurring."));
 	return content.join('\n');
 }

src/vs/workbench/contrib/speech/browser/maiSpeechCredentials.ts

@@ -107,19 +107,26 @@ export class MaiSpeechCredentialsService extends Disposable implements IMaiSpeec
 
 	private async refreshHasKey(): Promise<void> {
 		const wasConfigured = this.isConfigured;
-		this.hasKey = !!await this.readKey();
+		this.hasKey = !!await this.resolve();
 
 		if (this.isConfigured !== wasConfigured) {
 			this._onDidChangeConfigured.fire();
 		}
 	}
 
-	private async readKey(): Promise<string | undefined> {
+	/**
+	 * The stored key together with the endpoint it was given for, or `undefined`
+	 * when nothing is stored.
+	 */
+	private async readStoredKey(): Promise<IMaiSpeechCredentials | undefined> {
 		try {
-			return await this.secretStorageService.get(MAI_SPEECH_KEY_SECRET) || undefined;
+			const stored = await this.secretStorageService.get(MAI_SPEECH_KEY_SECRET);
+
+			return stored ? JSON.parse(stored) as IMaiSpeechCredentials : undefined;
 		} catch (error) {
-			// Secret storage is unavailable on some platforms; reading aloud then
-			// falls back to the speech synthesizer of the platform.
+			// Secret storage is unavailable on some platforms, and a value written
+			// by an older version is not in this shape; reading aloud then falls
+			// back to the speech synthesizer of the platform.
 			this.logService.warn(`[speech] could not read the MAI speech key: ${error}`);
 
 			return undefined;
@@ -128,14 +135,18 @@ export class MaiSpeechCredentialsService extends Disposable implements IMaiSpeec
 
 	async resolve(): Promise<IMaiSpeechCredentials | undefined> {
 		const endpoint = this.endpoint;
-		const key = await this.readKey();
+		const stored = await this.readStoredKey();
 
-		return endpoint && key ? { endpoint, key } : undefined;
+		// A key is only ever sent to the endpoint it was given for. Pointing the
+		// setting at another host therefore stops reading aloud until a key for
+		// that host is entered, instead of handing this one to it.
+		return endpoint && stored?.key && stored.endpoint === endpoint ? { endpoint, key: stored.key } : undefined;
 	}
 
 	async setKey(key: string | undefined): Promise<void> {
-		if (key?.trim()) {
-			await this.secretStorageService.set(MAI_SPEECH_KEY_SECRET, key.trim());
+		const endpoint = this.endpoint;
+		if (key?.trim() && endpoint) {
+			await this.secretStorageService.set(MAI_SPEECH_KEY_SECRET, JSON.stringify({ endpoint, key: key.trim() } satisfies IMaiSpeechCredentials));
 		} else {
 			await this.secretStorageService.delete(MAI_SPEECH_KEY_SECRET);
 		}

src/vs/workbench/contrib/speech/common/speechText.ts

@@ -18,7 +18,7 @@
  * `\uFE0F`.
  */
 const TEXTUAL_SYMBOL = /[\u00A9\u00AE\u2122\u203C\u2049]/;
-const EMOJI = new RegExp(`(?!${TEXTUAL_SYMBOL.source}(?!\\uFE0F))(?:\\p{Extended_Pictographic}\\uFE0F?(?:[\\u{1F3FB}-\\u{1F3FF}]|\\u20E3)?(?:\\u200D\\p{Extended_Pictographic}\\uFE0F?(?:[\\u{1F3FB}-\\u{1F3FF}])?)*|\\p{RI}\\p{RI}|[0-9#*]\\uFE0F\\u20E3)`, 'gu');
+const EMOJI = new RegExp(`(?!${TEXTUAL_SYMBOL.source}(?!\\uFE0F))(?:\\p{Extended_Pictographic}\\uFE0F?(?:[\\u{1F3FB}-\\u{1F3FF}]|\\u20E3)?(?:\\u200D\\p{Extended_Pictographic}\\uFE0F?(?:[\\u{1F3FB}-\\u{1F3FF}])?)*|\\p{RI}\\p{RI}|[0-9#*]\\uFE0F?\\u20E3)`, 'gu');
 
 /**
  * Removes emoji from `text` so they are not read aloud. Emoji are decoration in

src/vs/workbench/contrib/speech/test/common/speechText.test.ts

@@ -37,12 +37,15 @@ suite('stripEmoji', () => {
 			skinTone: stripEmoji('Nice 👍🏽 work'),
 			flag: stripEmoji('Ships to 🇺🇸 only'),
 			keycap: stripEmoji('Step 1️⃣ first'),
+			// The variation selector is optional in the Unicode keycap grammar.
+			keycapWithoutVariationSelector: stripEmoji('Step 1\u20E3 first'),
 			variationSelector: stripEmoji('Warning ⚠️ ahead')
 		}, {
 			zwjSequence: 'Ask the team',
 			skinTone: 'Nice work',
 			flag: 'Ships to only',
 			keycap: 'Step first',
+			keycapWithoutVariationSelector: 'Step first',
 			variationSelector: 'Warning ahead'
 		});
 	});