agentvibes 5.10.1 → 5.11.1

package/src/services/provider-voice-catalog.js

@@ -0,0 +1,126 @@
+/**
+ * Provider voice catalogs — single source of truth for the static voice lists
+ * of the cloud/neural providers, plus a helper that returns the voice pool for
+ * whichever provider is currently active.
+ *
+ * Piper voices are NOT listed here (they are discovered on disk via
+ * scanInstalledVoices); this module covers the providers whose voices are a
+ * fixed catalog: ElevenLabs and Kokoro. Single-voice providers (e.g. soprano)
+ * resolve to a single synthetic entry.
+ *
+ * @module services/provider-voice-catalog
+ */
+
+import { isSingleVoiceProvider } from './agent-voice-store.js';
+
+/** ElevenLabs library voices (voice_id + display metadata). */
+export const ELEVENLABS_VOICES = [
+  { id: 'EXAVITQu4vr4xnSDxMaL', name: 'Sarah',   gender: 'Female', lang: 'en-US', desc: 'Mature, reassuring'      },
+  { id: 'CwhRBWXzGAHq8TQ4Fs17', name: 'Roger',   gender: 'Male',   lang: 'en-US', desc: 'Laid-back, casual'       },
+  { id: 'FGY2WhTYpPnrIDTdsKH5', name: 'Laura',   gender: 'Female', lang: 'en-US', desc: 'Enthusiast, quirky'      },
+  { id: 'IKne3meq5aSn9XLyUdCD', name: 'Charlie', gender: 'Male',   lang: 'en-AU', desc: 'Deep, confident'         },
+  { id: 'JBFqnCBsd6RMkjVDRZzb', name: 'George',  gender: 'Male',   lang: 'en-GB', desc: 'Warm storyteller'        },
+  { id: 'N2lVS1w4EtoT3dr4eOWO', name: 'Callum',  gender: 'Male',   lang: 'en-US', desc: 'Husky trickster'         },
+  { id: 'SAz9YHcvj6GT2YYXdXww', name: 'River',   gender: '',       lang: 'en-US', desc: 'Relaxed, neutral'        },
+  { id: 'SOYHLrjzK2X1ezoPC6cr', name: 'Harry',   gender: 'Male',   lang: 'en-US', desc: 'Fierce warrior'          },
+  { id: 'TX3LPaxmHKxFdv7VOQHJ', name: 'Liam',    gender: 'Male',   lang: 'en-US', desc: 'Energetic creator'       },
+  { id: 'Xb7hH8MSUJpSbSDYk0k2', name: 'Alice',   gender: 'Female', lang: 'en-GB', desc: 'Clear educator'          },
+  { id: 'XrExE9yKIg1WjnnlVkGX', name: 'Matilda', gender: 'Female', lang: 'en-US', desc: 'Knowledgable, pro'       },
+  { id: 'bIHbv24MWmeRgasZH58o', name: 'Will',    gender: 'Male',   lang: 'en-US', desc: 'Relaxed optimist'        },
+  { id: 'cgSgspJ2msm6clMCkdW9', name: 'Jessica', gender: 'Female', lang: 'en-US', desc: 'Playful, bright'         },
+  { id: 'cjVigY5qzO86Huf0OWal', name: 'Eric',    gender: 'Male',   lang: 'en-US', desc: 'Smooth, trustworthy'     },
+  { id: 'hpp4J3VqNfWAUOO0d1Us', name: 'Bella',   gender: 'Female', lang: 'en-US', desc: 'Professional, warm'      },
+  { id: 'iP95p4xoKVk53GoZ742B', name: 'Chris',   gender: 'Male',   lang: 'en-US', desc: 'Charming, down-to-earth' },
+  { id: 'nPczCjzI2devNBz1zQrb', name: 'Brian',   gender: 'Male',   lang: 'en-US', desc: 'Deep, comforting'        },
+  { id: 'onwK4e9ZLuTAKqWW03F9', name: 'Daniel',  gender: 'Male',   lang: 'en-GB', desc: 'Steady broadcaster'      },
+  { id: 'pFZP5JQG7iQjIQuC4Bku', name: 'Lily',    gender: 'Female', lang: 'en-GB', desc: 'Velvety actress'         },
+  { id: 'pNInz6obpgDQGcFmaJgB', name: 'Adam',    gender: 'Male',   lang: 'en-US', desc: 'Dominant, firm'          },
+  { id: 'pqHfZKP75CvOlQylNhV4', name: 'Bill',    gender: 'Male',   lang: 'en-US', desc: 'Wise, mature'            },
+  // ── Added from the ElevenLabs voice library (raw library voice_ids) ───────
+  { id: 'HBDoL4wkcalemIO0nUAu', name: 'Emily',   gender: 'Female', lang: 'en-GB', desc: 'RPG, immersive'    },
+  { id: 'Mtmp3KhFIjYpWYRycDe3', name: 'John',    gender: 'Male',   lang: 'en-US', desc: 'Raspy surfer'      },
+  { id: '8kgj5469z1URcH4MB2G4', name: 'Sakuya',  gender: 'Female', lang: 'en-US', desc: 'Cheerful anime'    },
+  { id: '2ajXGJNYBR0iNHpS4VZb', name: 'Rob',     gender: 'Male',   lang: 'en-GB', desc: 'Tough, gritty'     },
+  { id: 'mMdTuD2nnFiCH88UdXSb', name: 'Jane',    gender: 'Female', lang: 'en-US', desc: 'Cinematic villain' },
+];
+
+/** Kokoro voice ids. Gender is encoded in the id: the 2nd char is f|m. */
+export const KOKORO_VOICE_IDS = [
+  // American English
+  'af_heart','af_alloy','af_aoede','af_bella','af_jessica','af_kore','af_nicole','af_nova','af_river','af_sarah','af_sky',
+  'am_adam','am_echo','am_eric','am_fenrir','am_liam','am_michael','am_onyx','am_puck',
+  // British English
+  'bf_alice','bf_emma','bf_isabella','bf_lily',
+  'bm_daniel','bm_fable','bm_george','bm_lewis',
+  // Japanese
+  'jf_alpha','jf_gongitsune','jf_nezumi','jf_tebukuro','jm_kumo',
+  // Mandarin Chinese
+  'zf_xiaobei','zf_xiaoni','zf_xiaoxiao','zf_xiaoyi','zm_yunxi','zm_yunxia','zm_yunyang',
+  // Spanish
+  'ef_dora','em_alex','em_santa',
+  // French
+  'ff_siwis',
+  // Hindi
+  'hf_alpha','hm_omega',
+  // Italian
+  'if_sara','im_nicola',
+  // Brazilian Portuguese
+  'pf_dora','pm_alex','pm_santa',
+  // Korean
+  'kf_alpha','km_hyunsu',
+];
+
+/**
+ * Infer a Kokoro voice's gender from its id. Kokoro ids follow `<lang><sex>_name`
+ * where the 2nd character is `f` (female) or `m` (male).
+ * @param {string} id
+ * @returns {'Female'|'Male'|''}
+ */
+export function kokoroGender(id) {
+  const c = typeof id === 'string' && id.length > 1 ? id[1].toLowerCase() : '';
+  if (c === 'f') return 'Female';
+  if (c === 'm') return 'Male';
+  return '';
+}
+
+/**
+ * Return the voice pool for the currently-active provider as `{ id, gender }`
+ * entries, suitable for gender-aware auto-assignment.
+ *
+ * Piper voices are discovered on disk, so the caller injects `scanInstalledVoices`
+ * and `getVoiceMeta` (both live in the voices-tab module) to avoid a circular
+ * import here.
+ *
+ * @param {string} provider - active provider id (piper|kokoro|elevenlabs|soprano|…)
+ * @param {object} helpers
+ * @param {() => string[]} helpers.scanInstalledVoices - lists installed Piper voice ids
+ * @param {(id: string) => {gender?: string}} helpers.getVoiceMeta - Piper voice metadata
+ * @returns {Array<{id: string, gender: string}>}
+ */
+export function voicesForProvider(provider, { scanInstalledVoices, getVoiceMeta } = {}) {
+  const p = (provider || 'piper').toLowerCase();
+
+  if (p === 'kokoro') {
+    return KOKORO_VOICE_IDS.map(id => ({ id, gender: kokoroGender(id) }));
+  }
+
+  if (p === 'elevenlabs') {
+    return ELEVENLABS_VOICES.map(v => ({ id: v.id, gender: v.gender || '' }));
+  }
+
+  // Single-voice providers (e.g. soprano): one synthetic entry named after the
+  // provider — every agent ends up sharing it, which is correct for these.
+  if (isSingleVoiceProvider(p)) {
+    return [{ id: p, gender: '' }];
+  }
+
+  // Piper (and any disk-discovered provider): read installed models.
+  if (typeof scanInstalledVoices === 'function') {
+    const metaOf = typeof getVoiceMeta === 'function' ? getVoiceMeta : () => ({});
+    return scanInstalledVoices().map(id => ({ id, gender: metaOf(id).gender || '' }));
+  }
+
+  return [];
+}
+
+export default { ELEVENLABS_VOICES, KOKORO_VOICE_IDS, kokoroGender, voicesForProvider };

package/src/services/tts-engine-service.js

@@ -8,6 +8,10 @@
 import { execFileSync } from 'node:child_process';
 import path from 'node:path';
 import fs from 'node:fs';
+import os from 'node:os';
+import { fileURLToPath } from 'node:url';
+
+const _hooksDir = path.join(path.dirname(fileURLToPath(import.meta.url)), '..', '..', '.claude', 'hooks');
 
 const TTS_ENGINES = [
   {
@@ -16,10 +20,29 @@ const TTS_ENGINES = [
       ? 'winget install --id Rhasspy.Piper --accept-package-agreements --accept-source-agreements'
       : process.platform === 'darwin'
         ? 'brew install piper'
-        : 'pip install piper-tts',
+        : `${path.join(_hooksDir, 'piper-installer.sh')} --non-interactive`,
+    // Structured form — robust against spaces in paths (do NOT split installCmd on ' ').
+    installSpec: process.platform === 'win32'
+      ? { cmd: 'winget', args: ['install', '--id', 'Rhasspy.Piper', '--accept-package-agreements', '--accept-source-agreements'] }
+      : process.platform === 'darwin'
+        ? { cmd: 'brew', args: ['install', 'piper'] }
+        : { cmd: path.join(_hooksDir, 'piper-installer.sh'), args: ['--non-interactive'] },
+  },
+  {
+    id: 'kokoro', name: 'Kokoro TTS', desc: 'Local neural TTS — 60+ voices, 8 languages, no API key', native: false,
+    installCmd: `${path.join(_hooksDir, 'kokoro-installer.sh')} --non-interactive`,
+    // Structured form — robust against spaces in paths (do NOT split installCmd on ' ').
+    installSpec: { cmd: path.join(_hooksDir, 'kokoro-installer.sh'), args: ['--non-interactive'] },
+    requiresApiKey: false,
   },
   {
-    id: 'soprano', name: 'Soprano TTS', desc: 'Web-based TTS service with premium voices', native: false,
+    id: 'elevenlabs', name: 'ElevenLabs', desc: 'Premium cloud voices — free tier available, 32+ languages', native: false,
+    installCmd: `echo "Set ELEVENLABS_API_KEY in your shell profile. Free key at elevenlabs.io"`,
+    requiresApiKey: true,
+    apiKeyEnvVar: 'ELEVENLABS_API_KEY',
+  },
+  {
+    id: 'soprano', name: 'Soprano TTS', desc: 'Local neural TTS via Gradio WebUI', native: false,
     installCmd: 'pip install soprano-tts',
   },
   { id: 'sapi', name: 'Windows SAPI', desc: 'Built-in Windows speech — no install needed', native: true, platform: 'win32' },
@@ -35,6 +58,32 @@ export function checkEngineInstalled(engineId) {
   if (!engine) return false;
   if (engine.native) return true; // Native engines are always available on their platform
 
+  // ElevenLabs: check API key in env or key file
+  if (engineId === 'elevenlabs') {
+    if (process.env.ELEVENLABS_API_KEY?.trim()) return true;
+    try {
+      const keyFile = path.join(os.homedir(), '.agentvibes', 'elevenlabs-key.txt');
+      return fs.existsSync(keyFile) && fs.readFileSync(keyFile, 'utf8').trim().length > 0;
+    } catch { return false; }
+  }
+
+  // Kokoro: use find_spec (no import) to avoid the slow torch load that causes ETIMEDOUT.
+  // Try each platform-appropriate python command (Windows lacks `python3`).
+  if (engineId === 'kokoro') {
+    const pythonCommands = process.platform === 'win32'
+      ? ['py', 'python', 'python3']
+      : ['python3', 'python'];
+    for (const pythonCmd of pythonCommands) {
+      try {
+        execFileSync(pythonCmd, ['-c', "import importlib.util,sys; sys.exit(0 if importlib.util.find_spec('kokoro') else 1)"], { stdio: 'ignore', timeout: 3000 }); // NOSONAR
+        return true;
+      } catch {
+        // try next python command
+      }
+    }
+    return false;
+  }
+
   // Check binary availability — soprano has two possible binaries
   const binaryMap = { piper: ['piper'], soprano: ['soprano-tts', 'soprano-webui'] };
   const binaries = binaryMap[engineId];

package/src/utils/audio-duration-validator.js

@@ -49,10 +49,15 @@ export const DURATION_LIMITS = {
  */
 export async function getAudioDuration(filePath) {
   return new Promise((resolve) => {
+    // Resolve exactly once: an 'error' event, a 'close' event, and a timeout can
+    // all race, and a double-resolve would mask the first (correct) result.
+    let settled = false;
+    const done = (result) => { if (!settled) { settled = true; resolve(result); } };
+
     try {
       // Verify file exists
       if (!fs.existsSync(filePath)) {
-        resolve({
+        done({
           success: false,
           duration: null,
           error: 'File does not exist'
@@ -62,13 +67,35 @@ export async function getAudioDuration(filePath) {
 
       // Check if ffprobe is available
       const checkFFprobe = spawn('which', ['ffprobe']); // NOSONAR
-      let ffprobeExists = false;
+
+      // `which` does not exist on Windows, so this spawn can fail outright.
+      // Without an 'error' listener Node re-throws as an unhandled error and
+      // crashes the caller; handle it and degrade gracefully instead.
+      const checkTimer = setTimeout(() => {
+        try { checkFFprobe.kill(); } catch { /* already exited */ }
+        done({
+          success: false,
+          duration: null,
+          error: 'ffprobe availability check timed out'
+        });
+      }, 5000);
+      checkTimer.unref?.();
+
+      checkFFprobe.on('error', (err) => {
+        clearTimeout(checkTimer);
+        done({
+          success: false,
+          duration: null,
+          error: `ffprobe not found: ${err.message}. Please install ffmpeg.`
+        });
+      });
 
       checkFFprobe.on('close', (code) => {
-        ffprobeExists = (code === 0);
+        clearTimeout(checkTimer);
+        if (settled) return;
 
-        if (!ffprobeExists) {
-          resolve({
+        if (code !== 0) {
+          done({
             success: false,
             duration: null,
             error: 'ffprobe not found. Please install ffmpeg: sudo apt install ffmpeg (Linux) or brew install ffmpeg (macOS)'
@@ -85,6 +112,18 @@ export async function getAudioDuration(filePath) {
           filePath
         ]);
 
+        // A malformed or locked file can make ffprobe hang; cap it so the promise
+        // always settles instead of leaving the caller awaiting forever.
+        const ffTimer = setTimeout(() => {
+          try { ffprobe.kill(); } catch { /* already exited */ }
+          done({
+            success: false,
+            duration: null,
+            error: 'ffprobe timed out reading the file'
+          });
+        }, 10000);
+        ffTimer.unref?.();
+
         let stdoutChunks = [];
         let stderrChunks = [];
 
@@ -97,11 +136,14 @@ export async function getAudioDuration(filePath) {
         });
 
         ffprobe.on('close', (code) => {
+          clearTimeout(ffTimer);
+          if (settled) return;
+
           const stdout = Buffer.concat(stdoutChunks).toString();
           const stderr = Buffer.concat(stderrChunks).toString();
 
           if (code !== 0) {
-            resolve({
+            done({
               success: false,
               duration: null,
               error: `ffprobe exited with code ${code}: ${stderr || 'Unknown error'}`
@@ -112,7 +154,7 @@ export async function getAudioDuration(filePath) {
           const duration = parseFloat(stdout.trim());
 
           if (isNaN(duration) || duration <= 0) {
-            resolve({
+            done({
               success: false,
               duration: null,
               error: 'Invalid duration value from ffprobe'
@@ -120,7 +162,7 @@ export async function getAudioDuration(filePath) {
             return;
           }
 
-          resolve({
+          done({
             success: true,
             duration,
             error: null
@@ -128,7 +170,8 @@ export async function getAudioDuration(filePath) {
         });
 
         ffprobe.on('error', (err) => {
-          resolve({
+          clearTimeout(ffTimer);
+          done({
             success: false,
             duration: null,
             error: `Failed to spawn ffprobe: ${err.message}. Please install ffmpeg.`
@@ -137,7 +180,7 @@ export async function getAudioDuration(filePath) {
       });
 
     } catch (err) {
-      resolve({
+      done({
         success: false,
         duration: null,
         error: `Unexpected error: ${err.message}`