agentgui 1.0.170 → 1.0.171

package/lib/speech.js

@@ -8,9 +8,19 @@ const require = createRequire(import.meta.url);
 const __dirname = path.dirname(fileURLToPath(import.meta.url));
 const ROOT = path.dirname(__dirname);
 const DATA_DIR = path.join(ROOT, 'data');
-const VOICES_DIR = path.join(ROOT, 'voices');
-const HOME_VOICES_DIR = path.join(os.homedir(), 'voices');
 const AUDIO_EXTENSIONS = ['.mp3', '.wav', '.ogg', '.flac', '.m4a'];
+
+function getVoiceDirs() {
+  const dirs = [];
+  const seen = new Set();
+  const add = (d) => { const r = path.resolve(d); if (!seen.has(r)) { seen.add(r); dirs.push(r); } };
+  const startupCwd = process.env.STARTUP_CWD || process.cwd();
+  add(path.join(startupCwd, 'voices'));
+  add(path.join(ROOT, 'voices'));
+  add(path.join(os.homedir(), 'voices'));
+  return dirs;
+}
+
 const MIN_WAV_SIZE = 1000;
 
 const BASE_VOICES = [
@@ -24,6 +34,24 @@ const BASE_VOICES = [
   { id: 'ksp', name: 'KSP', gender: 'male', accent: 'Indian' },
 ];
 
+async function convertToWav(filePath) {
+  const wavPath = filePath.replace(/\.[^.]+$/, '.wav');
+  if (fs.existsSync(wavPath)) return wavPath;
+  try {
+    console.log('[VOICES] Converting to WAV:', filePath);
+    const audio = await decodeAudioFile(filePath);
+    const wav = encodeWav(audio, SAMPLE_RATE_STT);
+    fs.writeFileSync(wavPath, wav);
+    console.log('[VOICES] Converted:', path.basename(wavPath));
+    return wavPath;
+  } catch (err) {
+    console.error('[VOICES] Conversion failed for', filePath + ':', err.message);
+    return null;
+  }
+}
+
+const pendingConversions = new Map();
+
 function scanVoiceDir(dir) {
   const voices = [];
   try {
@@ -32,6 +60,18 @@ function scanVoiceDir(dir) {
       const ext = path.extname(file).toLowerCase();
       if (!AUDIO_EXTENSIONS.includes(ext)) continue;
       const baseName = path.basename(file, ext);
+      if (ext !== '.wav') {
+        const wavFile = baseName + '.wav';
+        if (fs.existsSync(path.join(dir, wavFile))) continue;
+        const fullPath = path.join(dir, file);
+        if (!pendingConversions.has(fullPath)) {
+          pendingConversions.set(fullPath, convertToWav(fullPath).then(result => {
+            pendingConversions.delete(fullPath);
+            return result;
+          }));
+        }
+        continue;
+      }
       const id = 'custom_' + baseName.replace(/[^a-zA-Z0-9_-]/g, '_');
       const name = baseName.replace(/_/g, ' ');
       voices.push({ id, name, gender: 'custom', accent: 'custom', isCustom: true, sourceDir: dir });
@@ -45,7 +85,7 @@ function scanVoiceDir(dir) {
 function loadCustomVoices() {
   const seen = new Set();
   const voices = [];
-  for (const dir of [VOICES_DIR, HOME_VOICES_DIR]) {
+  for (const dir of getVoiceDirs()) {
     for (const v of scanVoiceDir(dir)) {
       if (seen.has(v.id)) continue;
       seen.add(v.id);
@@ -170,7 +210,7 @@ async function getSpeakerEmbeddingPipeline() {
 
 function findCustomVoiceFile(voiceId) {
   const baseName = voiceId.replace(/^custom_/, '');
-  for (const dir of [VOICES_DIR, HOME_VOICES_DIR]) {
+  for (const dir of getVoiceDirs()) {
     for (const ext of AUDIO_EXTENSIONS) {
       const candidate = path.join(dir, baseName + ext);
       if (fs.existsSync(candidate)) return candidate;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentgui",
-  "version": "1.0.170",
+  "version": "1.0.171",
   "description": "Multi-agent ACP client with real-time communication",
   "type": "module",
   "main": "server.js",