From 77aa4f3d07dd84cbbe4030234a7e395d579a2f31 Mon Sep 17 00:00:00 2001
From: ValueOn AG <p.motsch@valueon.ch>
Date: Thu, 26 Feb 2026 09:22:07 +0100
Subject: [PATCH] feat: enforce aggressive hybrid mode for names-only captions

Made-with: Cursor
---
 src/bot/orchestrator.ts | 45 +++++++++--------------------------------
 1 file changed, 10 insertions(+), 35 deletions(-)

diff --git a/src/bot/orchestrator.ts b/src/bot/orchestrator.ts
index f621457..fcbe9ea 100644
--- a/src/bot/orchestrator.ts
+++ b/src/bot/orchestrator.ts
@@ -926,14 +926,9 @@ export class BotOrchestrator {
       this._page,
       this._logger,
       (entry) => {
-        const transferMode = this._getEffectiveTransferMode();
-        if (transferMode === 'audio') {
-          // In audio mode, captions are only used as speaker hints.
-          this._sendTranscript(entry.speaker, entry.text, entry.isFinal, 'speakerHint');
-        } else {
-          this._sendTranscript(entry.speaker, entry.text, entry.isFinal, 'caption');
-          this._callbacks.onTranscript(entry);
-        }
+        // Aggressive hybrid mode: captions are always speaker hints only.
+        // Caption text is never persisted as transcript due to language quality issues.
+        this._sendTranscript(entry.speaker, entry.text, entry.isFinal, 'speakerHint');
       },
       this._options.language
     );
@@ -958,11 +953,8 @@ export class BotOrchestrator {
     // This ensures Teams gets our controlled audio stream when it calls getUserMedia
     await this._audioProcedure.injectAudioOverride();
 
-    // Inject audio capture (WebRTC interception) if transfer mode requires it
-    const transferMode = this._getEffectiveTransferMode();
-    if (transferMode === 'audio') {
-      await this._audioCaptureProcedure!.injectCaptureOverride();
-    }
+    // Aggressive hybrid mode: always capture meeting audio as transcript source.
+    await this._audioCaptureProcedure!.injectCaptureOverride();
 
     // Handle page errors
     this._page.on('pageerror', (error) => {
@@ -1108,32 +1100,15 @@ export class BotOrchestrator {
     }
   }
 
-  /**
-   * Subscribe to captions only as speaker hints (without enabling captions flow).
-   * This supports hybrid mode: audio text + caption-based speaker names.
-   */
-  private async _enableSpeakerHintsFromCaptions(): Promise<void> {
-    try {
-      await this._captionsProcedure!.subscribeToCaptions();
-      this._logger.info('Speaker hints from captions subscribed (audio mode)');
-    } catch (error) {
-      this._logger.warn('Could not subscribe to captions for speaker hints:', error);
-    }
-  }
-
   /**
    * Enable transcript capture (captions or audio) based on transfer mode.
    */
   private async _enableTranscriptCapture(): Promise<void> {
-    const transferMode = this._getEffectiveTransferMode();
-    this._logger.info(`Transfer mode: ${transferMode} (configured: ${this._options.transferMode || 'auto'})`);
-
-    if (transferMode === 'caption') {
-      await this._enableCaptions();
-    } else {
-      await this._enableAudioCapture();
-      await this._enableSpeakerHintsFromCaptions();
-    }
+    this._logger.info(
+      `Aggressive hybrid mode active: audio STT + background captions for speaker hints (configured transferMode: ${this._options.transferMode || 'auto'})`,
+    );
+    await this._enableAudioCapture();
+    await this._enableCaptions();
   }
 
   /**