agentgui 1.0.222 → 1.0.223

package/lib/speech.js

@@ -27,6 +27,12 @@ const POCKET_TTS_VOICES = [
 ];
 
 const PREDEFINED_IDS = new Set(POCKET_TTS_VOICES.filter(v => v.id !== 'default').map(v => v.id));
+const FALLBACK_VOICE = 'cosette';
+
+function isVoiceCloningError(err) {
+  const msg = (err.message || err.stderr || String(err)).toLowerCase();
+  return msg.includes('voice cloning') || msg.includes('could not download the weights');
+}
 const POCKET_PORT = 8787;
 
 const needsPatch = !serverTTS.getVoices(EXTRA_VOICE_DIRS).some(v => v.id === 'alba' && !v.isCustom);
@@ -79,11 +85,19 @@ function getSTT() {
   return serverSTT.getSTT();
 }
 
-function synthesize(text, voiceId) {
-  if (needsPatch && voiceId && PREDEFINED_IDS.has(voiceId)) {
-    return synthesizeDirect(text, voiceId);
+async function synthesize(text, voiceId) {
+  try {
+    if (needsPatch && voiceId && PREDEFINED_IDS.has(voiceId)) {
+      return await synthesizeDirect(text, voiceId);
+    }
+    return await serverTTS.synthesize(text, voiceId, EXTRA_VOICE_DIRS);
+  } catch (err) {
+    if (isVoiceCloningError(err) && voiceId && !PREDEFINED_IDS.has(voiceId)) {
+      console.log(`[TTS] Voice cloning failed for "${voiceId}", falling back to ${FALLBACK_VOICE}`);
+      return synthesize(text, FALLBACK_VOICE);
+    }
+    throw err;
   }
-  return serverTTS.synthesize(text, voiceId, EXTRA_VOICE_DIRS);
 }
 
 function synthesizeStream(text, voiceId) {
@@ -95,7 +109,27 @@ function synthesizeStream(text, voiceId) {
       }
     })();
   }
-  return serverTTS.synthesizeStream(text, voiceId, EXTRA_VOICE_DIRS);
+  return synthesizeStreamWithFallback(text, voiceId);
+}
+
+function synthesizeStreamWithFallback(text, voiceId) {
+  let fellBack = false;
+  const upstream = serverTTS.synthesizeStream(text, voiceId, EXTRA_VOICE_DIRS);
+  return (async function* () {
+    try {
+      for await (const chunk of upstream) {
+        yield chunk;
+      }
+    } catch (err) {
+      if (!fellBack && isVoiceCloningError(err) && voiceId && !PREDEFINED_IDS.has(voiceId)) {
+        fellBack = true;
+        console.log(`[TTS] Voice cloning failed for "${voiceId}", falling back to ${FALLBACK_VOICE}`);
+        yield* synthesizeStream(text, FALLBACK_VOICE);
+      } else {
+        throw err;
+      }
+    }
+  })();
 }
 
 function getVoices() {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentgui",
-  "version": "1.0.222",
+  "version": "1.0.223",
   "description": "Multi-agent ACP client with real-time communication",
   "type": "module",
   "main": "server.js",