feat: ensure mic ON on pre-join + voice greeting via Gateway TTS

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

src/bot/orchestrator.ts

@@ -137,6 +137,9 @@ export class BotOrchestrator {
     // STEP 1: Navigate to meeting URL and click "Continue on this browser"
     await this._joinProcedure!.startMeetingLauncherFlow(this._meetingUrl);
 
+    // Ensure microphone is ON (required for voice playback)
+    await this._ensureMicOn();
+
     // STEP 2: Enter bot name and click "Join now"
     await this._joinProcedure!.joinMeetingLobbyFlow();
 
@@ -282,6 +285,9 @@ export class BotOrchestrator {
     // Camera stays OFF — no video injection, focus on communication stability
     this._logger.info('Camera left OFF (video disabled for stability)');
 
+    // Ensure microphone is ON (required for voice playback)
+    await this._ensureMicOn();
+
     await this._page!.waitForTimeout(2000);
 
     const joinNowSelectors = [
@@ -470,6 +476,67 @@ export class BotOrchestrator {
     }
   }
 
+  /**
+   * Ensure the microphone is turned on in the pre-join screen.
+   * Required for voice playback (TTS audio is injected into the mic stream).
+   *
+   * Teams pre-join uses a fui-Switch input:
+   *   <input data-tid="toggle-audio" role="switch" type="checkbox" checked>
+   *   - checked present = mic ON
+   *   - checked absent  = mic OFF
+   */
+  private async _ensureMicOn(): Promise<void> {
+    try {
+      let micToggle = await this._page!.$('input[data-tid="toggle-audio"]');
+
+      if (!micToggle) {
+        const fallbacks = [
+          '[data-tid="toggle-audio"]',
+          'input[role="switch"][title*="microphone" i]',
+          'input[role="switch"][title*="Mikrofon" i]',
+          'input[role="switch"][title*="mic" i]',
+          'input[role="switch"][title*="audio" i]',
+        ];
+        for (const sel of fallbacks) {
+          micToggle = await this._page!.$(sel);
+          if (micToggle) {
+            this._logger.info(`Mic toggle found via fallback: ${sel}`);
+            break;
+          }
+        }
+      }
+
+      if (!micToggle) {
+        this._logger.warn('Mic toggle not found on pre-join screen');
+        return;
+      }
+
+      const state = await micToggle.evaluate((el: HTMLInputElement) => ({
+        checked: el.checked,
+        dataCid: el.getAttribute('data-cid') || '',
+        title: el.getAttribute('title') || '',
+      }));
+
+      this._logger.info(`Mic state: checked=${state.checked}, data-cid="${state.dataCid}", title="${state.title}"`);
+
+      if (!state.checked) {
+        await micToggle.click();
+        this._logger.info('Mic toggled ON');
+        await this._page!.waitForTimeout(1000);
+
+        const afterState = await micToggle.evaluate((el: HTMLInputElement) => ({
+          checked: el.checked,
+          dataCid: el.getAttribute('data-cid') || '',
+        }));
+        this._logger.info(`Mic after toggle: checked=${afterState.checked}, data-cid="${afterState.dataCid}"`);
+      } else {
+        this._logger.info('Mic already ON');
+      }
+    } catch (err) {
+      this._logger.warn(`Could not toggle mic: ${err}`);
+    }
+  }
+
   /**
    * Start a keepalive timer that periodically moves the mouse and sends
    * a WebSocket ping. Prevents Teams from detecting the bot as idle
@@ -1055,8 +1122,9 @@ export class BotOrchestrator {
   }
 
   /**
-   * Send a greeting message in the meeting chat after joining.
+   * Send a greeting message in the meeting chat AND via voice after joining.
    * Uses the bot's display name and the configured language.
+   * Voice greeting confirms that the audio pipeline (TTS -> mic) is working.
    */
   private async _sendJoinGreeting(): Promise<void> {
     try {
@@ -1074,8 +1142,18 @@ export class BotOrchestrator {
         greeting = `Hello, this is ${firstName}. I'm ready.`;
       }
 
-      this._logger.info(`Sending join greeting: ${greeting}`);
+      this._logger.info(`Sending join greeting (chat + voice): ${greeting}`);
+
+      // Chat greeting
       await this.sendChatMessageToMeeting(greeting);
+
+      // Voice greeting — ask Gateway to generate TTS and send back playAudio
+      this._sendToGateway({
+        type: 'voiceGreeting',
+        sessionId: this._sessionId,
+        text: greeting,
+        language: this._options.language || 'de-DE',
+      });
     } catch (error) {
       this._logger.warn('Could not send join greeting:', error);
     }