clay-server 4.3.1-beta.5 → 4.4.0-beta.1

package/lib/public/modules/stt.js

@@ -1,266 +1,127 @@
-// Speech-to-Text module using Web Speech API
-// Uses browser's built-in speech recognition (Chrome/Edge/Safari → Google servers)
-
+// Compact shared voice controls for project, DM and Home composers.
 import { iconHtml, refreshIcons } from './icons.js';
-import { autoResize } from './input.js';
+import { showToast } from './utils.js';
 import { store } from './store.js';
-import { VENDOR_NAMES } from './app-rendering.js';
-
-var ctx;
-
-// --- State ---
-var recording = false;
-var recognition = null;
-var selectedLang = null;
-var textBeforeSTT = '';
-var interimText = '';
-
-// DOM refs
-var sttBtn = null;
-var langPopover = null;
-
-// --- Language options ---
-// Web Speech API uses BCP-47 language tags
-var LANGUAGES = [
-  { code: 'en-US', name: 'English' },
-  { code: 'ko-KR', name: 'Korean' },
-  { code: 'ja-JP', name: 'Japanese' },
-  { code: 'zh-CN', name: 'Chinese' },
-  { code: 'es-ES', name: 'Spanish' },
-  { code: 'fr-FR', name: 'French' },
-  { code: 'de-DE', name: 'German' },
-];
+import { initSpeechLifecycle, speechConstructor, setSpeechStatus, startSTT, stopSTT } from './stt-controller.js';
+import { startRecording, finishRecording, retryRecording } from './stt-recording.js';
+import { voiceModel, voiceProvider, loadSpeechSettings } from './stt-settings.js';
+import { openSpeechPicker, initSpeechPicker } from './stt-picker.js';
+import { createSpeechMeter } from './stt-meter.js';
+export { stopSTT } from './stt-controller.js';
 
-// --- Check browser support ---
-function getSpeechRecognition() {
-  return window.SpeechRecognition || window.webkitSpeechRecognition || null;
+export function setSTTLang(code) {
+  if (typeof code !== 'string' || !code) return;
+  if (store.get('speechLang') !== code) stopSTT();
+  store.set({ speechLang: code });
 }
-
-// --- Init ---
-export function initSTT(_ctx) {
-  ctx = _ctx;
-
-  sttBtn = document.getElementById('stt-btn');
-  if (!sttBtn) return;
-
-  if (!getSpeechRecognition()) {
-    sttBtn.style.display = 'none';
-    console.warn('[STT] Web Speech API not supported in this browser');
-    return;
-  }
-
-  sttBtn.addEventListener('click', function(e) {
-    e.stopPropagation();
-
-    if (recording) {
-      stopRecording();
+export function getSTTLang() { return store.get('speechLang') || 'en-US'; }
+export function isSTTRecording() { return !!store.get('speechActive'); }
+export function isSTTInitializing() { var active = store.get('speechActive'); return !!active && !active.started; }
+
+function bindComposer(id, inputId, wrapperId) {
+  var button = document.getElementById(id);
+  var input = document.getElementById(inputId);
+  var wrapper = document.getElementById(wrapperId);
+  if (!button || !input || !wrapper || button.dataset.speechBound) return;
+  var composer = document.getElementById(id === 'stt-btn' ? 'input-row' : 'home-mate-chat-composer') || wrapper;
+  var toolbar = document.getElementById(id === 'stt-btn' ? 'input-bottom' : 'home-mate-chat-composer') || composer;
+  button.dataset.speechBound = 'true'; button.classList.add('stt-button');
+  var group = document.createElement('span'); group.className = 'stt-controls';
+  button.before(group); group.appendChild(button);
+  var picker = document.createElement('button'); picker.type = 'button'; picker.id = id + '-picker';
+  picker.className = 'stt-picker-toggle'; picker.innerHTML = iconHtml('chevron-down');
+  picker.setAttribute('aria-label', 'Choose voice model'); picker.setAttribute('aria-haspopup', 'dialog'); picker.setAttribute('aria-expanded', 'false');
+  picker.title = 'Choose voice model'; group.appendChild(picker);
+  picker.addEventListener('click', function () { openSpeechPicker(picker); });
+  var row = document.createElement('div'); row.className = 'stt-progress'; row.hidden = true;
+  var status = document.createElement('span'); status.id = id + '-status'; status.className = 'stt-status'; status.setAttribute('role', 'status');
+  var timer = document.createElement('span'); timer.className = 'stt-timer'; timer.setAttribute('aria-hidden', 'true');
+  var cancel = document.createElement('button'); cancel.type = 'button'; cancel.innerHTML = iconHtml('x');
+  cancel.className = 'stt-cancel'; cancel.title = 'Cancel voice input'; cancel.setAttribute('aria-label', 'Cancel voice input');
+  cancel.addEventListener('click', function () { stopSTT(); store.set({ speechRetry: null }); });
+  var meter = createSpeechMeter(id);
+  var retry = document.createElement('button'); retry.type = 'button'; retry.textContent = 'Retry';
+  retry.addEventListener('click', retryRecording);
+  var account = document.createElement('a'); account.textContent = 'Provider account'; account.target = '_blank'; account.rel = 'noopener noreferrer';
+  var dismiss = document.createElement('button'); dismiss.type = 'button'; dismiss.textContent = 'Dismiss';
+  dismiss.addEventListener('click', function () { setSpeechStatus(id, ''); });
+  row.append(cancel, meter.element, timer, status, retry, account, dismiss); composer.prepend(row);
+  button.setAttribute('aria-describedby', status.id);
+  button.addEventListener('click', function () {
+    var active = store.get('speechActive');
+    if (active && active.id === id) {
+      if (active.model && active.phase === 'recording') finishRecording(); else stopSTT();
       return;
     }
-
-    if (!selectedLang) {
-      showLangPopover();
-    } else {
-      startRecording();
-    }
+    var model = voiceModel(); var settings = store.get('speechSettings');
+    if (model.provider !== 'browser' && (!settings || settings.configured.indexOf(model.provider) < 0)) { openSpeechPicker(picker); return; }
+    if (model.provider === 'browser') startSTT(id, input); else startRecording(id, input, model);
   });
-
-  // Right-click to change language
-  sttBtn.addEventListener('contextmenu', function(e) {
-    e.preventDefault();
-    e.stopPropagation();
-    if (recording) stopRecording();
-    showLangPopover();
+  input.addEventListener('beforeinput', function () {
+    var active = store.get('speechActive');
+    if (active && active.id === id) stopSTT('Voice input stopped so you can edit your draft.');
+    store.set({ speechRetry: null });
   });
-}
-
-// --- Language popover ---
-function showLangPopover() {
-  if (langPopover) {
-    hideLangPopover();
-    return;
-  }
-
-  langPopover = document.createElement('div');
-  langPopover.className = 'stt-lang-popover';
-
-  var html = '<div class="stt-lang-title">Voice Input Language</div>';
-  for (var i = 0; i < LANGUAGES.length; i++) {
-    var l = LANGUAGES[i];
-    var activeClass = (selectedLang === l.code) ? ' stt-lang-active' : '';
-    html += '<button class="stt-lang-option' + activeClass + '" data-lang="' + l.code + '">' +
-      '<span class="stt-lang-name">' + l.name + '</span>' +
-      '</button>';
-  }
-  langPopover.innerHTML = html;
-
-  langPopover.querySelectorAll('.stt-lang-option').forEach(function(btn) {
-    btn.addEventListener('click', function() {
-      onLangSelected(btn.dataset.lang);
-    });
+  input.addEventListener('input', function () {
+    var active = store.get('speechActive');
+    if (active && active.id === id && !active.writing) stopSTT();
   });
-
-  var wrapper = document.getElementById('input-wrapper');
-  wrapper.appendChild(langPopover);
-
-  setTimeout(function() {
-    document.addEventListener('click', closeLangOnOutside);
-  }, 0);
-}
-
-function closeLangOnOutside(e) {
-  if (langPopover && !langPopover.contains(e.target) && e.target !== sttBtn && !sttBtn.contains(e.target)) {
-    hideLangPopover();
-  }
-}
-
-function hideLangPopover() {
-  if (langPopover) {
-    langPopover.remove();
-    langPopover = null;
-  }
-  document.removeEventListener('click', closeLangOnOutside);
-}
-
-function onLangSelected(code) {
-  selectedLang = code;
-  hideLangPopover();
-  startRecording();
-}
-
-// --- Recording ---
-function startRecording() {
-  if (recording) return;
-
-  var SpeechRecognition = getSpeechRecognition();
-  if (!SpeechRecognition) return;
-
-  recognition = new SpeechRecognition();
-  recognition.lang = selectedLang || 'en-US';
-  recognition.continuous = true;
-  recognition.interimResults = true;
-
-  textBeforeSTT = ctx.inputEl.value;
-  interimText = '';
-
-  recognition.onresult = function(e) {
-    var final = '';
-    var interim = '';
-
-    for (var i = 0; i < e.results.length; i++) {
-      var result = e.results[i];
-      if (result.isFinal) {
-        final += result[0].transcript;
-      } else {
-        interim += result[0].transcript;
-      }
-    }
-
-    var text = textBeforeSTT;
-    if (final) {
-      if (text && text.length > 0 && text[text.length - 1] !== ' ' && text[text.length - 1] !== '\n') {
-        text += ' ';
-      }
-      text += final;
-    }
-    if (interim) {
-      if (text && text.length > 0 && text[text.length - 1] !== ' ' && text[text.length - 1] !== '\n') {
-        text += ' ';
-      }
-      text += interim;
-    }
-
-    ctx.inputEl.value = text;
-    autoResize();
-    if (ctx.scrollToBottom) ctx.scrollToBottom();
-  };
-
-  recognition.onerror = function(e) {
-    console.error('[STT] Recognition error:', e.error);
-    if (e.error === 'not-allowed') {
-      if (ctx.addSystemMessage) {
-        ctx.addSystemMessage('Microphone access denied.\n\nTo fix: click the lock icon in the address bar → Site settings → Microphone → Allow, then reload.', true);
-      }
-      stopRecording();
-    } else if (e.error === 'no-speech') {
-      // Silence — just keep listening
-    } else if (e.error === 'network') {
-      if (ctx.addSystemMessage) {
-        ctx.addSystemMessage('Speech recognition unavailable.\n\nWeb Speech API sends audio to Google servers for recognition. Some Chromium forks (Arc, Brave) block this connection.\n\nSupported: Chrome, Edge, Safari 14.1+, Samsung Internet\nNot supported: Arc, Brave, Firefox', true);
-      }
-      stopRecording();
-    }
-  };
-
-  recognition.onend = function() {
-    // Auto-restart if still recording (browser may stop after silence)
-    if (recording) {
-      // Save confirmed text so far
-      textBeforeSTT = ctx.inputEl.value;
-      try {
-        recognition.start();
-      } catch (e) {
-        // Already started or other error
-        stopRecording();
+  function render() {
+    var active = store.get('speechActive'); var own = !!active && active.id === id;
+    var model = voiceModel(); var provider = voiceProvider(model.provider);
+    var pending = own && active.model && active.phase !== 'recording';
+    composer.classList.toggle('stt-recording', own);
+    row.classList.toggle('stt-progress-active', own);
+    row.classList.toggle('stt-progress-pending', !!pending);
+    var parent = own ? toolbar : composer;
+    if (row.parentElement !== parent || row !== parent.firstElementChild) parent.prepend(row);
+    meter.update(own ? active : null);
+    picker.hidden = own;
+    button.disabled = input.disabled || input.readOnly || !store.get('speechSettings');
+    picker.disabled = input.disabled || input.readOnly || own;
+    button.classList.toggle('stt-active', own);
+    button.setAttribute('aria-pressed', String(own));
+    var action = own ? (pending ? 'Cancel voice input' : (model.mode === 'recorded' ? 'Stop and transcribe' : 'Stop voice input')) : 'Start voice input';
+    button.setAttribute('aria-label', action + ' · ' + (provider ? provider.name + ' · ' : '') + model.name);
+    button.title = action + ' · ' + model.name;
+    var signature = [own, pending, provider && provider.id].join(':');
+    if (button.dataset.speechIcon !== signature) {
+      button.dataset.speechIcon = signature;
+      button.innerHTML = iconHtml(own ? (pending ? 'x' : 'square') : 'mic');
+      if (provider && !own) {
+        var badge = document.createElement('img'); badge.className = 'stt-vendor-badge'; badge.src = provider.icon; badge.alt = ''; button.appendChild(badge);
       }
+      refreshIcons();
     }
-  };
-
-  try {
-    recognition.start();
-    recording = true;
-    sttBtn.classList.add('stt-active');
-    sttBtn.innerHTML =
-      '<span class="stt-wave">' +
-        '<span class="stt-wave-bar"></span>' +
-        '<span class="stt-wave-bar"></span>' +
-        '<span class="stt-wave-bar"></span>' +
-        '<span class="stt-wave-bar"></span>' +
-        '<span class="stt-wave-bar"></span>' +
-      '</span>' +
-      '<span class="stt-stop-label">Stop</span>';
-    ctx.inputEl.setAttribute('placeholder', 'Listening...');
-  } catch (err) {
-    console.error('[STT] Failed to start:', err);
-    if (ctx.addSystemMessage) {
-      ctx.addSystemMessage('Failed to start voice input: ' + err.message, true);
-    }
+    var message = (store.get('speechStatuses') || {})[id] || '';
+    var failure = (store.get('speechErrors') || {})[id];
+    row.classList.toggle('stt-progress-error', !!failure);
+    status.setAttribute('role', failure ? 'alert' : 'status');
+    dismiss.hidden = !failure;
+    account.hidden = !failure || !provider;
+    if (provider) account.href = provider.url;
+    if (model.provider === 'browser' && !speechConstructor()) message = 'Browser voice input is unavailable. Choose a connected speech provider or type your message.';
+    if (!window.isSecureContext) message = 'Voice input requires a secure HTTPS connection.';
+    status.textContent = message;
+    timer.textContent = own && active.startTime ? Math.floor((Date.now() - active.startTime) / 60000) + ':' + String(Math.floor((Date.now() - active.startTime) / 1000) % 60).padStart(2, '0') : '';
+    cancel.hidden = !own;
+    var retryState = store.get('speechRetry'); retry.hidden = !retryState || retryState.id !== id;
+    row.hidden = !own && !failure && retry.hidden;
   }
+  store.subscribe(render);
+  store.subscribe(function (state, previous) {
+    var failure = (state.speechErrors || {})[id];
+    if (failure && failure !== (previous.speechErrors || {})[id]) showToast(failure.message, 'warn');
+  });
+  var observer = new MutationObserver(render);
+  observer.observe(input, { attributes: true, attributeFilter: ['disabled', 'readonly'] });
+  render();
 }
-
-function stopRecording() {
-  if (!recording) return;
-  recording = false;
-
-  if (recognition) {
-    try { recognition.stop(); } catch (e) { /* ignore */ }
-    recognition = null;
-  }
-
-  sttBtn.classList.remove('stt-active');
-  sttBtn.innerHTML = iconHtml('mic');
-  refreshIcons();
-  if (document.body.classList.contains("mate-dm-active") && document.body.dataset.mateName) {
-    ctx.inputEl.setAttribute('placeholder', 'Message ' + document.body.dataset.mateName + '...');
-  } else {
-    var _v = store.get('currentVendor') || "claude";
-    ctx.inputEl.setAttribute('placeholder', 'Message ' + (VENDOR_NAMES[_v] || VENDOR_NAMES.claude) + '...');
-  }
-}
-
-// --- External lang setter (used by profile module) ---
-export function setSTTLang(code) {
-  selectedLang = code;
-}
-
-export function getSTTLang() {
-  return selectedLang;
-}
-
-// --- Exports ---
-export function isSTTRecording() {
-  return recording;
-}
-
-export function isSTTInitializing() {
-  return false;
+export function initSTT() {
+  initSpeechLifecycle(); initSpeechPicker();
+  bindComposer('stt-btn', 'input', 'input-wrapper');
+  bindComposer('home-stt-btn', 'home-mate-chat-input', 'home-mate-chat-composer-frame');
+  loadSpeechSettings();
+  window.addEventListener('focus', loadSpeechSettings);
+  try { window.top.document.addEventListener('clay:speech-settings-saved', loadSpeechSettings); } catch (error) { /* Cross-origin host. */ }
 }

package/lib/public/modules/user-settings.js

@@ -6,7 +6,7 @@ import { showToast } from './utils.js';
 import { getChatLayout, mountThemePicker, setChatLayout } from './theme.js';
 import { applyTerminalFont } from './terminal-prefs.js';
 import { showEmailSetupModal, getEmailAccountListCache } from './context-sources.js';
-import { setSTTLang } from './stt.js';
+import { saveSpeechLanguage, speechLanguages } from './stt-settings.js';
 import { userAvatarUrl } from './avatar.js';
 import { store } from './store.js';
 import { getWs } from './ws-ref.js';
@@ -27,6 +27,11 @@ var currentProfile = null;
 
 
 export function initUserSettings(appCtx) {
+  window.addEventListener('clay:speech-language-saved', function (event) {
+    var select = document.getElementById('us-lang-select');
+    if (select) select.value = event.detail;
+  });
+
   ctx = appCtx;
   settingsEl = document.getElementById('user-settings');
   openBtn = document.getElementById('user-settings-btn');
@@ -153,8 +158,15 @@ export function initUserSettings(appCtx) {
 
   var langSelect = document.getElementById('us-lang-select');
   if (langSelect) {
+    langSelect.replaceChildren();
+    speechLanguages.forEach(function (language) {
+      var option = document.createElement('option'); option.value = language[0]; option.textContent = language[1]; langSelect.appendChild(option);
+    });
     langSelect.addEventListener('change', function () {
-      saveProfileChange({ lang: this.value });
+      var select = this; select.disabled = true;
+      saveSpeechLanguage(select.value).catch(function () {
+        select.value = store.get('speechLang') || 'en-US'; showToast('Could not save voice language');
+      }).finally(function () { select.disabled = false; });
     });
   }
 
@@ -422,10 +434,7 @@ function populateAccount() {
       pinFormTitle.textContent = pinEnabled ? 'Change PIN' : 'Set PIN';
     }
     if (langSelect) {
-      langSelect.value = data.lang || 'en-US';
-    }
-    if (data.lang) {
-      setSTTLang(data.lang);
+      langSelect.value = store.get('speechLang') || 'en-US';
     }
     if (avatarImg && avatarFallback) {
       var avatarUrl = getCurrentUserAvatarSrc(data);
@@ -500,9 +509,6 @@ function saveProfileChange(updates) {
     }
   }
   currentProfile = nextProfile;
-  if (updates.lang) {
-    setSTTLang(updates.lang);
-  }
   payload.name = nextProfile.name;
   payload.lang = nextProfile.lang;
   payload.avatarColor = nextProfile.avatarColor;

package/lib/public/modules/worker-pane-lock.js

@@ -49,6 +49,7 @@ var LOCKED_CONTROL_IDS = [
   "input-more-btn",
   "shell-command-btn",
   "stt-btn",
+  "stt-btn-picker",
   "schedule-btn",
   "ask-mate-btn",
   "context-sources-add",

package/lib/server-speech.js

@@ -0,0 +1,166 @@
+var WebSocketServer = require('ws').WebSocketServer;
+var catalog = require('./speech-catalog');
+var attachSpeechStore = require('./speech-store').attachSpeechStore;
+var transcribe = require('./speech-transcribe').transcribe;
+var openSpeechStream = require('./speech-stream').openSpeechStream;
+var MAX_AUDIO = 10 * 1024 * 1024;
+
+function attachSpeech(ctx) {
+  var storage = attachSpeechStore(ctx.CONFIG_DIR);
+  var sockets = new WebSocketServer({ noServer: true, maxPayload: 65536 });
+  var busy = new Map();
+  var destroyed = false;
+  function owner(req) {
+    if (!ctx.isRequestAuthed(req)) return null;
+    if (!ctx.users.isMultiUser()) return 'default';
+    var user = ctx.getMultiUserFromReq(req);
+    return user && user.id;
+  }
+  function sameOrigin(req) {
+    if (req.headers['sec-fetch-site'] === 'cross-site') return false;
+    if (!req.headers.origin) return true;
+    try { return new URL(req.headers.origin).host === req.headers.host; } catch (error) { return false; }
+  }
+  function json(res, status, body) {
+    if (res.destroyed || res.writableEnded) return;
+    res.writeHead(status, { 'Content-Type': 'application/json', 'Cache-Control': 'no-store' }); res.end(JSON.stringify(body));
+  }
+  function collect(req, limit, signal) {
+    return new Promise(function (resolve, reject) {
+      var chunks = []; var size = 0;
+      function abort() { chunks = []; reject(new Error('Request cancelled')); }
+      if (signal) {
+        if (signal.aborted) { abort(); return; }
+        signal.addEventListener('abort', abort, { once: true });
+        req.once('end', function () { signal.removeEventListener('abort', abort); });
+      }
+      req.on('data', function (chunk) { size += chunk.length; if (size <= limit) chunks.push(chunk); else { chunks = []; reject(new Error('Request is too large')); } });
+      req.on('end', function () { if (size <= limit) resolve(Buffer.concat(chunks)); });
+      req.on('error', reject); req.on('aborted', function () { reject(new Error('Request cancelled')); });
+    });
+  }
+  function language(value) {
+    if (typeof value !== 'string' || !/^[a-z]{2,3}(?:-[a-zA-Z]{2,8})?$/.test(value)) throw new Error('Choose a valid speech language');
+    return value;
+  }
+  function selection(id, mode, userId) {
+    var model = catalog.findModel(id);
+    if (!model || model.mode !== mode || model.provider === 'browser') throw new Error('Invalid transcription model');
+    var key = storage.key(userId, model.provider);
+    if (!key) throw new Error('Connect this provider in Voice input settings first');
+    return { model: model, key: key };
+  }
+  function handleRequest(req, res, fullUrl) {
+    if (fullUrl !== '/api/speech/settings' && fullUrl !== '/api/speech/transcribe') return false;
+    var userId = owner(req);
+    if (!userId) { json(res, 401, { error: 'Sign in to use voice input' }); return true; }
+    if (!sameOrigin(req)) { json(res, 403, { error: 'Invalid request origin' }); return true; }
+    if (fullUrl === '/api/speech/settings') {
+      if (req.method === 'GET') {
+        try { json(res, 200, storage.view(userId)); } catch (error) { json(res, 500, { error: 'Voice settings could not be loaded' }); }
+      } else if (req.method === 'PUT') {
+        collect(req, 8192).then(function (bytes) {
+          if (owner(req) !== userId) { json(res, 401, { error: 'Please sign in again' }); return; }
+          var data = JSON.parse(bytes.toString());
+          if (!data || typeof data !== 'object') throw new Error('Invalid settings');
+          var value = storage.update(userId, data);
+          if (busy.has(userId)) busy.get(userId)();
+          json(res, 200, value);
+        }).catch(function () { json(res, 400, { error: 'Could not save voice settings. Check the model and API key.' }); });
+      } else json(res, 405, { error: 'Method not allowed' });
+      return true;
+    }
+    if (req.method !== 'POST') { json(res, 405, { error: 'Method not allowed' }); return true; }
+    if (busy.has(userId)) { json(res, 409, { error: 'Another voice request is already active' }); return true; }
+    var controller = new AbortController();
+    function cancel() { controller.abort(); }
+    busy.set(userId, cancel);
+    var timer = setTimeout(cancel, 120000);
+    res.on('close', cancel);
+    (async function () {
+      try {
+        var selected = selection(req.headers['x-speech-model'], 'recorded', userId);
+        var lang = language(req.headers['x-speech-language']);
+        var bytes = await collect(req, MAX_AUDIO, controller.signal);
+        if (owner(req) !== userId) throw new Error('Please sign in again');
+        if (bytes.length < 44 || bytes.toString('ascii', 0, 4) !== 'RIFF' || bytes.toString('ascii', 8, 12) !== 'WAVE') throw new Error('Invalid voice recording');
+        var text = await transcribe(selected.model, selected.key, bytes, lang, controller.signal);
+        if (typeof text !== 'string') throw new Error('No transcript returned');
+        json(res, 200, { text: text });
+      } catch (error) {
+        json(res, error.status || 400, { error: controller.signal.aborted ? 'Transcription cancelled or timed out. Please try again.' :
+          (error.status ? error.message : 'Could not transcribe this recording. Check your model, key and connection.') });
+      } finally { clearTimeout(timer); res.removeListener('close', cancel); if (busy.get(userId) === cancel) busy.delete(userId); }
+    })();
+    return true;
+  }
+  function handleUpgrade(req, socket, head) {
+    if (req.url.split('?')[0] !== '/api/speech/live') return false;
+    var userId = owner(req);
+    if (!userId || !sameOrigin(req) || busy.has(userId)) {
+      socket.write('HTTP/1.1 403 Forbidden\r\nConnection: close\r\n\r\n'); socket.destroy(); return true;
+    }
+    sockets.handleUpgrade(req, socket, head, function (ws) {
+      var upstream; var ready = false; var finishing = false; var total = 0; var ended = false;
+      function send(message) { if (ws.readyState === 1) ws.send(JSON.stringify(message)); }
+      function diagnostic(stage, values) { send(Object.assign({ type: 'diagnostic', stage: stage }, values || {})); }
+      diagnostic('clay-websocket-accepted');
+      function close() {
+        if (ended) return; ended = true;
+        clearTimeout(timer); clearInterval(authTimer);
+        if (upstream) upstream.abort();
+        if (busy.get(userId) === cancel) busy.delete(userId);
+        ws.close();
+      }
+      function cancel() { send({ type: 'error', error: 'Voice input cancelled. Your draft has been preserved.' }); close(); }
+      function fail(error) { diagnostic('clay-session-failed'); send({ type: 'error', error: error.message }); close(); }
+      busy.set(userId, cancel);
+      var timer = setTimeout(function () { fail(new Error('Voice input timed out')); }, 15000);
+      var authTimer = setInterval(function () { if (owner(req) !== userId) cancel(); }, 5000);
+      ws.on('message', function (raw, binary) {
+        if (ended) return;
+        try {
+          if (binary) {
+            if (!ready || finishing) throw new Error('Microphone session is not ready');
+            total += raw.length;
+            if (total === raw.length) diagnostic('clay-first-audio', { bytes: raw.length });
+            if (total > MAX_AUDIO || raw.length % 2) throw new Error('Recording limit reached');
+            upstream.audio(raw); return;
+          }
+          var message = JSON.parse(raw.toString());
+          if (message.type === 'start' && !upstream) {
+            diagnostic('clay-selecting-provider');
+            var selected = selection(message.model, 'live', userId);
+            var lang = language(message.language);
+            diagnostic('clay-provider-connecting');
+            upstream = openSpeechStream(selected.model, selected.key, lang, {
+              diagnostic: diagnostic,
+              ready: function () {
+                diagnostic('provider-ready');
+                ready = true; clearTimeout(timer);
+                timer = setTimeout(function () { fail(new Error('Recording limit reached. Stop within five minutes.')); }, 310000);
+                send({ type: 'ready' });
+              },
+              text: function (text) { send({ type: 'text', text: text }); },
+              done: function (text) { send({ type: 'done', text: text }); close(); },
+              error: fail
+            });
+          } else if (message.type === 'finish' && ready && !finishing) { finishing = true; upstream.finish(); }
+          else if (message.type === 'cancel') close();
+          else throw new Error('Invalid voice message');
+        } catch (error) { fail(new Error('Could not start or continue voice input. Check your key and connection.')); }
+      });
+      ws.on('close', close); ws.on('error', close);
+    });
+    return true;
+  }
+  function destroy() {
+    if (destroyed) return;
+    destroyed = true;
+    busy.forEach(function (cancel) { cancel(); });
+    sockets.clients.forEach(function (socket) { socket.terminate(); });
+    sockets.close();
+  }
+  return { handleRequest: handleRequest, handleUpgrade: handleUpgrade, destroy: destroy };
+}
+module.exports = { attachSpeech: attachSpeech };

package/lib/server.js

@@ -23,6 +23,7 @@ var serverCursorSharing = require("./server-cursor-sharing");
 var serverTools = require("./server-tools");
 var serverAdmin = require("./server-admin");
 var serverSettings = require("./server-settings");
+var serverSpeech = require("./server-speech");
 var serverPalette = require("./server-palette");
 var workspaceQueryModule = require("./workspace-query-service");
 var { attachWorkspaceQueryService } = workspaceQueryModule;
@@ -439,6 +440,9 @@ function createServer(opts) {
     getMultiUserFromReq: getMultiUserFromReq,
   });
 
+  var speech = serverSpeech.attachSpeech({ users: users, getMultiUserFromReq: getMultiUserFromReq,
+    isRequestAuthed: isRequestAuthed, CONFIG_DIR: CONFIG_DIR });
+
   var settings = serverSettings.attachSettings({
     users: users,
     mates: mates,
@@ -775,6 +779,7 @@ function createServer(opts) {
       return;
     }
 
+    if (speech.handleRequest(req, res, fullUrl)) return;
     if (settings.handleRequest(req, res, fullUrl)) return;
 
     // --- Admin API endpoints (multi-user mode only) ---
@@ -1104,7 +1109,9 @@ function createServer(opts) {
     },
   });
 
+  server.once("close", function () { speech.destroy(); });
   server.on("upgrade", function (req, socket, head) {
+    if (speech.handleUpgrade(req, socket, head)) return;
     var wsRequest = parseWsRequestUrl(req.url);
     var wsPathname = wsRequest.path;
     // Origin validation (CSRF prevention)
@@ -1904,6 +1911,7 @@ function createServer(opts) {
   });
 
   function destroyAll() {
+    speech.destroy();
     var shutdowns = [];
     shutdowns.push(durableSchedulerLifecycle.shutdown());
     projects.forEach(function (ctx, slug) {