fix: Language and speech opens SUBMENU with closed-captions-button menuitemcheckbox, not a panel with switches

Co-authored-by: Cursor <cursoragent@cursor.com>

src/bot/captionsProcedure.ts

@@ -106,44 +106,134 @@ export class CaptionsProcedure {
    * Enable captions or transcription from the "More" menu.
    *
    * Strategies in priority order:
-   * 1. Direct captions button (anonymous / light-meetings UI)
+   * 1. Direct captions button (anonymous / light-meetings UI — closed-captions-button in main menu)
-   * 2. "Record and transcribe" → "Start transcription" (authenticated Teams 2025+)
+   * 2. "Language and speech" → SUBMENU → "Show live captions" (authenticated Teams 2025+)
-   *    → triggers spoken-language-selection-dialog handled by _handleLanguageDialog()
+   *    The submenu item is a menuitemcheckbox with id="closed-captions-button"
-   *    → then "Show transcript" to open scraping panel
+   *    aria-checked="false" + "Show live captions" → click to enable
-   * 3. "Captions & transcripts" submenu (older authenticated Teams)
+   *    aria-checked="true" + "Hide live captions" → already on, don't click
-   * 4. "Language and speech" panel toggle (fallback)
+   * 3. "Record and transcribe" → "Start transcription" (fallback with transcript panel)
-   * 5. Generic text / DOM scan fallback
+   * 4. Generic text / DOM scan fallback
    */
   private async _clickEnableCaptions(): Promise<void> {
     await this._logVisibleMenuItems();
 
     // ── Strategy 1: Direct captions button (anonymous / light-meetings UI) ──
-    const directSelectors = [
+    // In anonymous mode, closed-captions-button appears directly in the More menu
-      'div[id="closed-captions-button"]',
+    const directBtn = await this._page.$('#closed-captions-button');
-      '[data-tid="closed-captions-button"]',
+    if (directBtn) {
-      '[data-tid="captions-toggle"]',
+      const ariaChecked = await directBtn.getAttribute('aria-checked');
+      const text = await directBtn.evaluate(el => el.textContent?.trim() || '');
+      this._logger.info(`Direct captions button found: aria-checked="${ariaChecked}", text="${text}"`);
+
+      if (ariaChecked === 'true' || text.toLowerCase().includes('hide')) {
+        this._logger.info('Live captions already ON (found "Hide live captions")');
+        await this._page.keyboard.press('Escape');
+        return;
+      }
+
+      await directBtn.click();
+      this._logger.info('Clicked "Show live captions" (direct button)');
+      await this._page.waitForTimeout(1000);
+      return;
+    }
+
+    // ── Strategy 2: "Language and speech" → submenu → "Show live captions" ──
+    // Authenticated Teams 2025+: "Language and speech" has aria-haspopup="menu"
+    // Opening it reveals a submenu with closed-captions-button as menuitemcheckbox
+    const langSpeechSelectors = [
+      '#LanguageSpeechMenuControl-id',
+      '[data-tid="LanguageSpeechMenuControl-id"]',
+      'div[role="menuitem"]:has-text("Language and speech")',
+      'div[role="menuitem"]:has-text("Sprache und Spracheingabe")',
     ];
 
-    for (const selector of directSelectors) {
+    for (const selector of langSpeechSelectors) {
       try {
-        const button = await this._page.$(selector);
+        const item = await this._page.$(selector);
-        if (button) {
+        if (item) {
-          await button.click();
+          await item.click();
-          this._logger.info(`Clicked direct captions button: ${selector}`);
+          this._logger.info(`Clicked "Language and speech": ${selector}`);
-          await this._page.waitForTimeout(1000);
+          await this._page.waitForTimeout(1500);
-          return;
+
+          // Now look for closed-captions-button in the submenu
+          const captionsBtn = await this._page.$('#closed-captions-button');
+          if (captionsBtn) {
+            const ariaChecked = await captionsBtn.getAttribute('aria-checked');
+            const btnText = await captionsBtn.evaluate(el => el.textContent?.trim() || '');
+            this._logger.info(`Submenu captions button: aria-checked="${ariaChecked}", text="${btnText}"`);
+
+            if (ariaChecked === 'true' || btnText.toLowerCase().includes('hide')) {
+              this._logger.info('Live captions already ON (found "Hide live captions")');
+              await this._page.keyboard.press('Escape');
+              return;
+            }
+
+            await captionsBtn.click();
+            this._logger.info('Clicked "Show live captions" in Language and speech submenu');
+            await this._page.waitForTimeout(1000);
+            return;
+          }
+
+          // Fallback: try menuitemcheckbox with captions-related text
+          const fallbackSelectors = [
+            '[role="menuitemcheckbox"]:has-text("captions")',
+            '[role="menuitemcheckbox"]:has-text("Untertitel")',
+            '[role="menuitemcheckbox"]:has-text("caption")',
+          ];
+
+          for (const fbSel of fallbackSelectors) {
+            try {
+              const fbBtn = await this._page.$(fbSel);
+              if (fbBtn) {
+                const ariaChecked = await fbBtn.getAttribute('aria-checked');
+                const fbText = await fbBtn.evaluate(el => el.textContent?.trim() || '');
+                this._logger.info(`Fallback captions button: aria-checked="${ariaChecked}", text="${fbText}"`);
+
+                if (ariaChecked === 'true' || fbText.toLowerCase().includes('hide')) {
+                  this._logger.info('Live captions already ON');
+                  await this._page.keyboard.press('Escape');
+                  return;
+                }
+
+                await fbBtn.click();
+                this._logger.info(`Clicked captions button: ${fbSel}`);
+                await this._page.waitForTimeout(1000);
+                return;
+              }
+            } catch {
+              // Continue
+            }
+          }
+
+          this._logger.warn('Language and speech submenu opened but no captions button found');
+          await this._page.keyboard.press('Escape');
+          break;
         }
       } catch {
         // Continue
       }
     }
 
-    // ── Strategy 2: "Record and transcribe" → "Start transcription" + "Show transcript" ──
+    // ── Strategy 3 (fallback): "Record and transcribe" → "Start transcription" + panel ──
+    this._logger.info('Live captions not found, trying transcription fallback...');
+
+    // Clean menu state first
+    await this._page.keyboard.press('Escape');
+    await this._page.waitForTimeout(500);
+    await this._page.keyboard.press('Escape');
+    await this._page.waitForTimeout(500);
+
+    try {
+      await this._openMoreMenu();
+    } catch {
+      this._logger.warn('Could not re-open More menu for transcription fallback');
+    }
+
     const recordMenuSelectors = [
       '[data-tid="RecordingMenuControl-id"]',
+      '#RecordingMenuControl-id',
       'div[role="menuitem"]:has-text("Record and transcribe")',
       'div[role="menuitem"]:has-text("Aufzeichnen und transkribieren")',
-      'div[role="menuitem"]:has-text("Aufnehmen und transkribieren")',
     ];
 
     for (const selector of recordMenuSelectors) {
@@ -179,13 +269,10 @@ export class CaptionsProcedure {
           }
 
           if (!alreadyRunning) {
-            // Click "Start transcription" (only explicit "Start" selectors)
             const startSelectors = [
               '[data-tid="call-transcript-button"]:has-text("Start")',
               '[role="menuitem"]:has-text("Start transcription")',
               '[role="menuitem"]:has-text("Transkription starten")',
-              'button:has-text("Start transcription")',
-              'button:has-text("Transkription starten")',
             ];
 
             let started = false;
@@ -230,7 +317,6 @@ export class CaptionsProcedure {
             }
           }
 
-          // Close any remaining menu overlay
           await this._page.keyboard.press('Escape');
           await this._page.waitForTimeout(500);
           return;
@@ -240,124 +326,7 @@ export class CaptionsProcedure {
       }
     }
 
-    // ── Strategy 3: "Captions & transcripts" submenu (older Teams) ──
+    // ── Strategy 4: DOM scan for anything containing "caption" / "transcri" ──
-    const submenuSelectors = [
-      '[data-tid="captions-and-transcripts-button"]',
-      '[role="menuitem"]:has-text("Captions & transcripts")',
-      '[role="menuitem"]:has-text("Captions and transcripts")',
-      '[role="menuitem"]:has-text("Untertitel und Transkripte")',
-      '[role="menuitem"]:has-text("Untertitel")',
-    ];
-
-    for (const selector of submenuSelectors) {
-      try {
-        const item = await this._page.$(selector);
-        if (item) {
-          await item.click();
-          this._logger.info(`Clicked captions submenu: ${selector}`);
-          await this._page.waitForTimeout(1500);
-
-          const enableSelectors = [
-            'button:has-text("Turn on live captions")',
-            'button:has-text("Live captions")',
-            'button:has-text("Live-Untertitel aktivieren")',
-            '[role="menuitem"]:has-text("Turn on live captions")',
-            '[role="menuitem"]:has-text("Live captions")',
-            '[role="menuitemcheckbox"]:has-text("captions")',
-            '[data-tid="toggle-captions"]',
-          ];
-
-          for (const enableSel of enableSelectors) {
-            try {
-              const enableBtn = await this._page.$(enableSel);
-              if (enableBtn) {
-                await enableBtn.click();
-                this._logger.info(`Clicked enable captions: ${enableSel}`);
-                await this._page.waitForTimeout(1000);
-                return;
-              }
-            } catch {
-              // Continue
-            }
-          }
-
-          this._logger.info('Opened captions submenu but could not find enable button');
-          break;
-        }
-      } catch {
-        // Continue
-      }
-    }
-
-    // ── Strategy 4 (fallback): "Language and speech" panel toggle ──
-    this._logger.info('Trying "Language and speech" as fallback...');
-
-    // Ensure clean menu state: close any open panels/menus first
-    await this._page.keyboard.press('Escape');
-    await this._page.waitForTimeout(500);
-    await this._page.keyboard.press('Escape');
-    await this._page.waitForTimeout(500);
-
-    try {
-      await this._openMoreMenu();
-    } catch {
-      this._logger.warn('Could not re-open More menu for Language and speech fallback');
-    }
-
-    const langSpeechSelectors = [
-      '[data-tid="LanguageSpeechMenuControl-id"]',
-      'div[role="menuitem"]:has-text("Language and speech")',
-      'div[role="menuitem"]:has-text("Sprache und Spracheingabe")',
-    ];
-
-    for (const selector of langSpeechSelectors) {
-      try {
-        const item = await this._page.$(selector);
-        if (item) {
-          await item.click();
-          this._logger.info(`Clicked "Language and speech": ${selector}`);
-          await this._page.waitForTimeout(2000);
-
-          const toggleResult = await this._page.evaluate(() => {
-            const switches = document.querySelectorAll(
-              'input[role="switch"], [role="switch"], input[type="checkbox"]'
-            );
-            for (const sw of Array.from(switches)) {
-              const label = (sw.getAttribute('aria-label') || '').toLowerCase();
-              const tid = (sw.getAttribute('data-tid') || '').toLowerCase();
-              const parentEl = sw.closest('div, label, span') as HTMLElement;
-              const nearText = (parentEl?.textContent || '').toLowerCase();
-              const isCaptions =
-                label.includes('caption') || label.includes('untertitel') ||
-                tid.includes('caption') || tid.includes('subtitle') ||
-                nearText.includes('live caption') || nearText.includes('liveuntertitel');
-              if (isCaptions) {
-                if (!(sw as HTMLInputElement).checked) {
-                  (sw as HTMLElement).click();
-                  return { found: true, clicked: true, info: label || tid || nearText.substring(0, 60) };
-                }
-                return { found: true, clicked: false, info: `already on: ${label || tid}` };
-              }
-            }
-            return { found: false, clicked: false, info: '' };
-          });
-
-          this._logger.info(`Captions toggle result: ${JSON.stringify(toggleResult)}`);
-          if (toggleResult.found && toggleResult.clicked) {
-            await this._page.waitForTimeout(1500);
-          }
-          await this._page.keyboard.press('Escape');
-          if (toggleResult.found) return;
-
-          this._logger.warn('Language panel opened but no captions toggle found');
-          break;
-        }
-      } catch {
-        // Continue
-      }
-    }
-
-    // ── Strategy 5: DOM scan for anything containing "caption" / "transcri" ──
     const found = await this._page.evaluate(() => {
       const keywords = ['caption', 'captions', 'untertitel', 'live caption', 'transcri', 'transkri'];
       const candidates = document.querySelectorAll(
@@ -379,13 +348,6 @@ export class CaptionsProcedure {
     if (found.clicked) {
       this._logger.info(`Clicked via DOM scan: "${found.clicked}"`);
       await this._page.waitForTimeout(1500);
-
-      const turnOnBtn = await this._page.$('button:has-text("Turn on"), [role="menuitem"]:has-text("Turn on")');
-      if (turnOnBtn) {
-        await turnOnBtn.click();
-        this._logger.info('Clicked "Turn on" in submenu');
-        await this._page.waitForTimeout(1000);
-      }
       return;
     }