agentgui 1.0.276 → 1.0.278

package/lib/speech.js

@@ -134,11 +134,20 @@ function preloadTTS() {
     console.log('[TTS] pocket-tts functions not available');
     return;
   }
+  if (typeof serverTTS.isInstalled === 'function' && !serverTTS.isInstalled()) {
+    console.log('[TTS] pocket-tts not installed yet - will install on first use');
+    return;
+  }
+  const portableDataDir = process.env.PORTABLE_DATA_DIR;
+  const binaryPaths = portableDataDir ? [
+    path.join(portableDataDir, 'pocket-venv', 'Scripts', 'pocket-tts.exe'),
+    path.join(portableDataDir, 'pocket-venv', 'bin', 'pocket-tts'),
+  ] : undefined;
   const defaultVoice = serverTTS.findVoiceFile('custom_cleetus', EXTRA_VOICE_DIRS) || '/config/voices/cleetus.wav';
   const voicePath = fs.existsSync(defaultVoice) ? defaultVoice : null;
-  serverTTS.start(voicePath, {}).then(ok => {
+  serverTTS.start(voicePath, binaryPaths ? { binaryPaths } : {}).then(ok => {
     if (ok) console.log('[TTS] pocket-tts sidecar started');
-    else console.log('[TTS] pocket-tts failed to start');
+    else console.log('[TTS] pocket-tts not available - will use edge-tts fallback');
   }).catch(err => {
     console.error('[TTS] pocket-tts start error:', err.message);
   });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentgui",
-  "version": "1.0.276",
+  "version": "1.0.278",
   "description": "Multi-agent ACP client with real-time communication",
   "type": "module",
   "main": "server.js",

package/server.js

@@ -8,6 +8,9 @@ import { fileURLToPath } from 'url';
 import { WebSocketServer } from 'ws';
 import { execSync, spawn } from 'child_process';
 import { createRequire } from 'module';
+import express from 'express';
+import Busboy from 'busboy';
+import fsbrowse from 'fsbrowse';
 import { OAuth2Client } from 'google-auth-library';
 import { queries, dataDir } from './database.js';
 import { runClaudeWithStreaming } from './lib/claude-runner.js';
@@ -69,12 +72,9 @@ async function ensureModelsDownloaded() {
   }
 
   try {
-    const { createRequire: cr } = await import('module');
-    const r = cr(import.meta.url);
-
-    const gmguiModels = path.join(dataDir, 'models');
-    const sttDir = path.join(gmguiModels, 'onnx-community', 'whisper-base');
-    const ttsDir = path.join(gmguiModels, 'tts');
+    const gmguiModels = path.join(dataDir, 'models');
+    const sttDir = path.join(gmguiModels, 'onnx-community', 'whisper-base');
+    const ttsDir = path.join(gmguiModels, 'tts');
 
     const sttOk = fs.existsSync(sttDir) && fs.readdirSync(sttDir).length > 0;
     const ttsOk = fs.existsSync(ttsDir) && fs.readdirSync(ttsDir).length > 0;
@@ -91,116 +91,89 @@ async function ensureModelsDownloaded() {
     const totalFiles = 16;
     let completedFiles = 0;
 
+    const require = createRequire(import.meta.url);
+
     if (!sttOk) {
-      console.log('[MODELS] Downloading STT model via IPFS...');
       broadcastModelProgress({ started: true, done: false, downloading: true, type: 'stt', source: 'ipfs', completedFiles, totalFiles });
+      let sttDownloaded = false;
+
+      const ipfsCid = queries.getIpfsCidByModel('whisper-base', 'stt');
+      if (ipfsCid) {
+        console.log('[MODELS] Downloading STT from Lighthouse IPFS:', ipfsCid.cid);
+        fs.mkdirSync(sttDir, { recursive: true });
+        const sttUrl = `https://gateway.lighthouse.storage/ipfs/${ipfsCid.cid}/stt/onnx-community/whisper-base/onnx/`;
+        const sttFile = path.join(sttDir, 'whisper-onnx.tar');
+        try {
+          await IPFSDownloader.downloadWithProgress(sttUrl, sttFile, (progress) => {
+            broadcastModelProgress({ started: true, done: false, downloading: true, type: 'stt', source: 'lighthouse-ipfs', gateway: 'gateway.lighthouse.storage', ...progress, completedFiles, totalFiles });
+          });
+          console.log('[MODELS] STT model downloaded from Lighthouse IPFS');
+          sttDownloaded = true;
+        } catch (err) {
+          console.error('[MODELS] IPFS STT download failed:', err.message, '- falling back to HuggingFace');
+        }
+      } else {
+        console.warn('[MODELS] No STT IPFS CID registered - using HuggingFace directly');
+      }
 
-      try {
-        const ipfsCid = queries.getIpfsCidByModel('whisper-base', 'stt');
-        if (!ipfsCid) {
-          console.warn('[MODELS] STT IPFS CID not registered in database');
-          console.warn('[MODELS] To enable STT: Pin whisper-base model to IPFS and register CID via: queries.recordIpfsCid(cid, "whisper-base", "stt", hash, gateway)');
-          broadcastModelProgress({
-            done: true,
-            error: 'STT model CID not registered - speech will be unavailable. Register via IPFS.',
-            type: 'stt',
-            completedFiles,
-            totalFiles
+      if (!sttDownloaded) {
+        console.log('[MODELS] Downloading STT model via HuggingFace...');
+        broadcastModelProgress({ started: true, done: false, downloading: true, type: 'stt', source: 'huggingface', completedFiles, totalFiles });
+        try {
+          const whisperModels = require('webtalk/whisper-models');
+          const modelsDir = path.join(dataDir, 'models');
+          fs.mkdirSync(modelsDir, { recursive: true });
+          await whisperModels.ensureModel('onnx-community/whisper-base', {
+            modelsDir,
+            whisperBaseUrl: 'https://huggingface.co/',
           });
-        } else {
-          console.log('[MODELS] Downloading STT from Lighthouse IPFS:', ipfsCid.cid);
-          fs.mkdirSync(sttDir, { recursive: true });
-
-          // Download from Lighthouse gateway: https://gateway.lighthouse.storage/ipfs/CID/stt/onnx-community/whisper-base/
-          const lighthouseGateway = 'https://gateway.lighthouse.storage/ipfs';
-          const sttUrl = `${lighthouseGateway}/${ipfsCid.cid}/stt/onnx-community/whisper-base/onnx/`;
-          const sttFile = path.join(sttDir, 'whisper-onnx.tar');
-
-          await IPFSDownloader.downloadWithProgress(
-            sttUrl,
-            sttFile,
-            (progress) => {
-              broadcastModelProgress({
-                started: true,
-                done: false,
-                downloading: true,
-                type: 'stt',
-                source: 'lighthouse-ipfs',
-                gateway: 'gateway.lighthouse.storage',
-                ...progress,
-                completedFiles,
-                totalFiles
-              });
-            }
-          );
-          console.log('[MODELS] STT model downloaded successfully from Lighthouse IPFS');
+          console.log('[MODELS] STT model downloaded from HuggingFace');
+        } catch (hfErr) {
+          console.error('[MODELS] HuggingFace STT download failed:', hfErr.message);
+          broadcastModelProgress({ done: true, error: `STT download failed: ${hfErr.message}`, type: 'stt', completedFiles, totalFiles });
         }
-      } catch (err) {
-        console.error('[MODELS] IPFS STT download failed:', err.message);
-        broadcastModelProgress({
-          done: true,
-          error: `IPFS STT download failed: ${err.message}`,
-          type: 'stt',
-          completedFiles,
-          totalFiles
-        });
       }
       completedFiles += 10;
     }
 
     if (!ttsOk) {
-      console.log('[MODELS] Downloading TTS models via IPFS...');
       broadcastModelProgress({ started: true, done: false, downloading: true, type: 'tts', source: 'ipfs', completedFiles, totalFiles });
+      let ttsDownloaded = false;
+
+      const ipfsCid = queries.getIpfsCidByModel('tts', 'voice');
+      if (ipfsCid) {
+        console.log('[MODELS] Downloading TTS from Lighthouse IPFS:', ipfsCid.cid);
+        fs.mkdirSync(ttsDir, { recursive: true });
+        const ttsUrl = `https://gateway.lighthouse.storage/ipfs/${ipfsCid.cid}/tts/`;
+        const ttsFile = path.join(ttsDir, 'tts-models.tar');
+        try {
+          await IPFSDownloader.downloadWithProgress(ttsUrl, ttsFile, (progress) => {
+            broadcastModelProgress({ started: true, done: false, downloading: true, type: 'tts', source: 'lighthouse-ipfs', gateway: 'gateway.lighthouse.storage', ...progress, completedFiles, totalFiles });
+          });
+          console.log('[MODELS] TTS models downloaded from Lighthouse IPFS');
+          ttsDownloaded = true;
+        } catch (err) {
+          console.error('[MODELS] IPFS TTS download failed:', err.message, '- falling back to HuggingFace');
+        }
+      } else {
+        console.warn('[MODELS] No TTS IPFS CID registered - using HuggingFace directly');
+      }
 
-      try {
-        const ipfsCid = queries.getIpfsCidByModel('tts', 'voice');
-        if (!ipfsCid) {
-          console.warn('[MODELS] TTS IPFS CID not registered in database');
-          console.warn('[MODELS] To enable TTS: Pin TTS models to IPFS and register CID via: queries.recordIpfsCid(cid, "tts", "voice", hash, gateway)');
-          broadcastModelProgress({
-            done: true,
-            error: 'TTS model CID not registered - speech synthesis will be unavailable. Register via IPFS.',
-            type: 'tts',
-            completedFiles,
-            totalFiles
+      if (!ttsDownloaded) {
+        console.log('[MODELS] Downloading TTS models via HuggingFace...');
+        broadcastModelProgress({ started: true, done: false, downloading: true, type: 'tts', source: 'huggingface', completedFiles, totalFiles });
+        try {
+          const ttsModels = require('webtalk/tts-models');
+          await ttsModels.ensureTTSModels({
+            ttsModelsDir: ttsDir,
+            ttsDir: path.join(dataDir, 'models', 'tts'),
+            ttsBaseUrl: 'https://huggingface.co/datasets/AnEntrypoint/sttttsmodels/resolve/main/tts/',
           });
-        } else {
-          console.log('[MODELS] Downloading TTS from Lighthouse IPFS:', ipfsCid.cid);
-          fs.mkdirSync(ttsDir, { recursive: true });
-
-          // Download from Lighthouse gateway: https://gateway.lighthouse.storage/ipfs/CID/tts/
-          const lighthouseGateway = 'https://gateway.lighthouse.storage/ipfs';
-          const ttsUrl = `${lighthouseGateway}/${ipfsCid.cid}/tts/`;
-          const ttsFile = path.join(ttsDir, 'tts-models.tar');
-
-          await IPFSDownloader.downloadWithProgress(
-            ttsUrl,
-            ttsFile,
-            (progress) => {
-              broadcastModelProgress({
-                started: true,
-                done: false,
-                downloading: true,
-                type: 'tts',
-                source: 'lighthouse-ipfs',
-                gateway: 'gateway.lighthouse.storage',
-                ...progress,
-                completedFiles,
-                totalFiles
-              });
-            }
-          );
-          console.log('[MODELS] TTS models downloaded successfully from Lighthouse IPFS');
+          console.log('[MODELS] TTS models downloaded from HuggingFace');
+        } catch (hfErr) {
+          console.error('[MODELS] HuggingFace TTS download failed:', hfErr.message);
+          broadcastModelProgress({ done: true, error: `TTS download failed: ${hfErr.message}`, type: 'tts', completedFiles, totalFiles });
         }
-      } catch (err) {
-        console.error('[MODELS] IPFS TTS download failed:', err.message);
-        broadcastModelProgress({
-          done: true,
-          error: `IPFS TTS download failed: ${err.message}`,
-          type: 'tts',
-          completedFiles,
-          totalFiles
-        });
       }
       completedFiles += 6;
     }
@@ -276,10 +249,6 @@ function pushTTSAudio(cacheKey, wav, conversationId, sessionId, voiceId) {
   });
 }
 
-const require = createRequire(import.meta.url);
-const express = require('express');
-const Busboy = require('busboy');
-const fsbrowse = require('fsbrowse');
 
 const SYSTEM_PROMPT = `Your output will be spoken aloud by a text-to-speech system. Write ONLY plain conversational sentences that sound natural when read aloud. Never use markdown, bold, italics, headers, bullet points, numbered lists, tables, or any formatting. Never use colons to introduce lists or options. Never use labels like "Option A" or "1." followed by a title. Instead of listing options, describe them conversationally in flowing sentences. For example, instead of "**Option 1**: Do X" say "One approach would be to do X." Keep sentences short and simple. Use transition words like "also", "another option", "or alternatively" to connect ideas. When mentioning file names, spell out the dot between the name and extension as the word "dot" so it is spoken clearly. For example, say "server dot js" instead of "server.js", "index dot html" instead of "index.html", and "package dot json" instead of "package.json". Write as if you are speaking to someone in a casual conversation.`;
 
@@ -296,16 +265,16 @@ const debugLog = (msg) => {
   console.error(`[${timestamp}] ${msg}`);
 };
 
-const __dirname = path.dirname(fileURLToPath(import.meta.url));
-const PORT = process.env.PORT || 3000;
-const BASE_URL = (process.env.BASE_URL || '/gm').replace(/\/+$/, '');
-const watch = process.argv.includes('--no-watch') ? false : (process.argv.includes('--watch') || process.env.HOT_RELOAD !== 'false');
-
-const STARTUP_CWD = process.env.STARTUP_CWD || process.cwd();
-const staticDir = process.env.PORTABLE_EXE_DIR 
-  ? path.join(process.env.PORTABLE_EXE_DIR, 'static')
-  : path.join(__dirname, 'static');
-if (!fs.existsSync(staticDir)) fs.mkdirSync(staticDir, { recursive: true });
+const __dirname = path.dirname(fileURLToPath(import.meta.url));
+const PORT = process.env.PORT || 3000;
+const BASE_URL = (process.env.BASE_URL || '/gm').replace(/\/+$/, '');
+const watch = process.argv.includes('--no-watch') ? false : (process.argv.includes('--watch') || process.env.HOT_RELOAD !== 'false');
+
+const STARTUP_CWD = process.env.STARTUP_CWD || process.cwd();
+const staticDir = process.env.PORTABLE_EXE_DIR 
+  ? path.join(process.env.PORTABLE_EXE_DIR, 'static')
+  : path.join(__dirname, 'static');
+if (!fs.existsSync(staticDir)) fs.mkdirSync(staticDir, { recursive: true });
 
 // Express sub-app for fsbrowse file browser and file upload
 const expressApp = express();
@@ -3697,21 +3666,21 @@ function onServerReady() {
 
   resumeInterruptedStreams().catch(err => console.error('[RESUME] Startup error:', err.message));
 
-  ensureModelsDownloaded().then(ok => {
+  ensureModelsDownloaded().then(async ok => {
     if (ok) console.log('[MODELS] Speech models ready');
     else console.log('[MODELS] Speech model download failed');
     try {
-      const { getVoices } = require('./lib/speech.js');
+      const { getVoices } = await getSpeech();
       const voices = getVoices();
       broadcastSync({ type: 'voice_list', voices });
     } catch (err) {
       debugLog('[VOICE] Failed to broadcast voices: ' + err.message);
       broadcastSync({ type: 'voice_list', voices: [] });
     }
-  }).catch(err => {
+  }).catch(async err => {
     console.error('[MODELS] Download error:', err.message);
     try {
-      const { getVoices } = require('./lib/speech.js');
+      const { getVoices } = await getSpeech();
       const voices = getVoices();
       broadcastSync({ type: 'voice_list', voices });
     } catch (err2) {