agentgui 1.0.292 → 1.0.294

package/lib/speech.js

@@ -12,12 +12,12 @@ const ROOT = path.dirname(__dirname);
 let serverTTS = null;
 let serverSTT = null;
 let audioDecode = null;
-let sttttsmodels = null;
+let ttsUtils = null;
 
 try { serverTTS = require('webtalk/server-tts'); } catch(e) { console.warn('[TTS] webtalk/server-tts unavailable:', e.message); }
 try { serverSTT = require('webtalk/server-stt'); } catch(e) { console.warn('[STT] webtalk/server-stt unavailable:', e.message); }
 try { audioDecode = require('audio-decode'); } catch(e) { console.warn('[TTS] audio-decode unavailable:', e.message); }
-try { sttttsmodels = require('sttttsmodels'); } catch(e) { console.warn('[TTS] sttttsmodels unavailable:', e.message); }
+try { ttsUtils = require('webtalk/tts-utils'); } catch(e) {}
 
 // Detect webtalk API type: old (server-tts.js with getVoices/synthesizeViaPocket)
 // vs new ONNX (server-tts-onnx.js with encodeVoiceAudio)
@@ -51,10 +51,6 @@ const SAMPLE_RATE = 24000;
 const voiceEmbeddingCache = new Map();
 
 function getModelDir() {
-  if (sttttsmodels && sttttsmodels.ttsDir && fs.existsSync(sttttsmodels.ttsDir)) {
-    return sttttsmodels.ttsDir;
-  }
-  // Fallback to persistent cache dir
   return path.join(os.homedir(), '.gmgui', 'models', 'tts');
 }
 
@@ -98,17 +94,20 @@ async function getVoiceEmbedding(voiceId) {
   if (!voicePath) return null;
   if (!audioDecode || !serverTTS || !isOnnxApi) return null;
 
+  const modelDir = getModelDir();
+  if (serverTTS.loadModels) await serverTTS.loadModels(modelDir);
+
   const raw = fs.readFileSync(voicePath);
   const decoded = await audioDecode.default(raw);
-  // Get mono float32 PCM, resample to 24kHz if needed
   let pcm = decoded.getChannelData(0);
   if (decoded.sampleRate !== SAMPLE_RATE) {
-    // Simple linear resampling
-    const ratio = decoded.sampleRate / SAMPLE_RATE;
-    const outLen = Math.floor(pcm.length / ratio);
-    const resampled = new Float32Array(outLen);
-    for (let i = 0; i < outLen; i++) resampled[i] = pcm[Math.floor(i * ratio)];
-    pcm = resampled;
+    pcm = ttsUtils ? ttsUtils.resample(pcm, decoded.sampleRate, SAMPLE_RATE)
+      : (() => {
+        const ratio = decoded.sampleRate / SAMPLE_RATE;
+        const out = new Float32Array(Math.round(pcm.length / ratio));
+        for (let i = 0; i < out.length; i++) out[i] = pcm[Math.floor(i * ratio)];
+        return out;
+      })();
   }
 
   const embedding = await serverTTS.encodeVoiceAudio(pcm);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentgui",
-  "version": "1.0.292",
+  "version": "1.0.294",
   "description": "Multi-agent ACP client with real-time communication",
   "type": "module",
   "main": "server.js",
@@ -30,8 +30,7 @@
     "fsbrowse": "^0.2.18",
     "google-auth-library": "^10.5.0",
     "onnxruntime-node": "^1.24.1",
-    "sttttsmodels": "github:AnEntrypoint/sttttsmodels",
-    "webtalk": "^1.0.12",
+    "webtalk": "^1.0.13",
     "ws": "^8.14.2"
   },
   "overrides": {