agentgui 1.0.728 → 1.0.730

package/static/js/voice-machine.js

@@ -0,0 +1,145 @@
+(function() {
+  const { createMachine, createActor, assign } = XState;
+
+  const MAX_FAILURES = 3;
+
+  const voiceMachine = createMachine({
+    id: 'voice',
+    initial: 'idle',
+    context: {
+      queue: [],
+      consecutiveFailures: 0,
+      currentText: null,
+      enabled: true,
+    },
+    states: {
+      idle: {
+        on: {
+          SPEAK: {
+            target: 'queued',
+            guard: ({ context }) => context.enabled,
+            actions: assign(({ context, event }) => ({
+              queue: [...context.queue, event.text],
+            })),
+          },
+          SET_ENABLED: {
+            actions: assign(({ event }) => ({ enabled: event.enabled })),
+          },
+        },
+      },
+      queued: {
+        always: [
+          {
+            guard: ({ context }) => context.queue.length > 0,
+            target: 'speaking',
+            actions: assign(({ context }) => ({
+              currentText: context.queue[0],
+              queue: context.queue.slice(1),
+            })),
+          },
+          { target: 'idle' },
+        ],
+        on: {
+          SPEAK: {
+            guard: ({ context }) => context.enabled,
+            actions: assign(({ context, event }) => ({
+              queue: [...context.queue, event.text],
+            })),
+          },
+          STOP: {
+            target: 'idle',
+            actions: assign({ queue: [], currentText: null }),
+          },
+        },
+      },
+      speaking: {
+        on: {
+          SPEAK: {
+            guard: ({ context }) => context.enabled,
+            actions: assign(({ context, event }) => ({
+              queue: [...context.queue, event.text],
+            })),
+          },
+          PLAYBACK_DONE: [
+            {
+              guard: ({ context }) => context.queue.length > 0,
+              target: 'speaking',
+              actions: assign(({ context }) => ({
+                currentText: context.queue[0],
+                queue: context.queue.slice(1),
+                consecutiveFailures: 0,
+              })),
+            },
+            {
+              target: 'idle',
+              actions: assign({ currentText: null, consecutiveFailures: 0 }),
+            },
+          ],
+          PLAYBACK_ERROR: [
+            {
+              guard: ({ context }) => context.consecutiveFailures + 1 >= MAX_FAILURES,
+              target: 'disabled',
+              actions: assign({ queue: [], currentText: null, consecutiveFailures: MAX_FAILURES }),
+            },
+            {
+              target: 'idle',
+              actions: assign(({ context }) => ({
+                consecutiveFailures: context.consecutiveFailures + 1,
+                currentText: null,
+              })),
+            },
+          ],
+          STOP: {
+            target: 'idle',
+            actions: assign({ queue: [], currentText: null }),
+          },
+          SET_ENABLED: {
+            actions: assign(({ event }) => ({ enabled: event.enabled })),
+          },
+        },
+      },
+      disabled: {
+        on: {
+          RESET: {
+            target: 'idle',
+            actions: assign({ consecutiveFailures: 0, queue: [], currentText: null }),
+          },
+          SET_ENABLED: {
+            actions: assign(({ event }) => ({ enabled: event.enabled })),
+          },
+        },
+      },
+    },
+  });
+
+  const actor = createActor(voiceMachine);
+  actor.start();
+
+  function sendEvent(event) {
+    actor.send(event);
+    return actor.getSnapshot();
+  }
+
+  function getState() {
+    return actor.getSnapshot().value;
+  }
+
+  function isActive() {
+    return getState() === 'speaking' || getState() === 'queued';
+  }
+
+  function isDisabled() {
+    return getState() === 'disabled';
+  }
+
+  function getContext() {
+    return actor.getSnapshot().context;
+  }
+
+  function subscribe(fn) {
+    return actor.subscribe(fn);
+  }
+
+  window.__voiceMachine = actor;
+  window.voiceMachineAPI = { send: sendEvent, getState, isActive, isDisabled, getContext, subscribe };
+})();

package/static/js/voice.js

@@ -1,119 +1,132 @@
-(function() {
-  var BASE = window.__BASE_URL || '';
-  var ttsEnabled = true;
-  var speechQueue = [];
-  var isSpeaking = false;
-  var currentAudio = null;
-  var selectedVoiceId = localStorage.getItem('gmgui-voice-selection') || 'default';
-  var ttsAudioCache = new Map();
-  var TTS_CLIENT_CACHE_MAX = 50;
-  var audioChunkQueue = [];
-  var isPlayingChunk = false;
-  var streamDone = false;
-  var ttsConsecutiveFailures = 0;
-  var TTS_MAX_FAILURES = 3;
-  var ttsDisabledUntilReset = false;
-  var streamingSupported = true;
-
-  window.addEventListener('ws-message', function(e) {
-    var data = e.detail;
-    if (!data) return;
-    if (data.type === 'tts_audio' && data.audio && data.voiceId === selectedVoiceId) cacheTTSAudio(data.cacheKey, data.audio);
-    if (data.type === 'sync_connected') sendVoiceToServer();
-  });
-
-  function cacheTTSAudio(cacheKey, b64) {
-    if (ttsAudioCache.size >= TTS_CLIENT_CACHE_MAX) ttsAudioCache.delete(ttsAudioCache.keys().next().value);
-    var binary = atob(b64), bytes = new Uint8Array(binary.length);
-    for (var i = 0; i < binary.length; i++) bytes[i] = binary.charCodeAt(i);
-    ttsAudioCache.set(cacheKey, new Blob([bytes], { type: 'audio/wav' }));
-  }
-
-  function speakDirect(text) {
-    var clean = text.replace(/<[^>]*>/g, '').trim();
-    if (!clean) return;
-    var parts = (typeof agentGUIClient !== 'undefined' && agentGUIClient && typeof agentGUIClient.parseMarkdownCodeBlocks === 'function')
-      ? agentGUIClient.parseMarkdownCodeBlocks(clean) : [{ type: 'text', content: clean }];
-    parts.forEach(function(p) { if (p.type !== 'code' && p.content.trim()) speechQueue.push(p.content.trim()); });
-    processQueue();
-  }
-
-  function playNextChunk() {
-    if (!audioChunkQueue.length) {
-      isPlayingChunk = false;
-      if (streamDone) { isSpeaking = false; processQueue(); }
-      return;
-    }
-    isPlayingChunk = true;
-    var blob = audioChunkQueue.shift(), url = URL.createObjectURL(blob);
-    currentAudio = new Audio(url);
-    var next = function() { URL.revokeObjectURL(url); currentAudio = null; playNextChunk(); };
-    currentAudio.onended = next; currentAudio.onerror = next;
-    currentAudio.play().catch(next);
-  }
-
-  function processQueue() {
-    if (isSpeaking || !speechQueue.length) return;
-    if (ttsDisabledUntilReset) { speechQueue = []; return; }
-    isSpeaking = true; streamDone = false;
-    var text = speechQueue.shift();
-    audioChunkQueue = []; isPlayingChunk = false;
-    var cached = ttsAudioCache.get(selectedVoiceId + ':' + text);
-    if (cached) { ttsConsecutiveFailures = 0; audioChunkQueue.push(cached); streamDone = true; if (!isPlayingChunk) playNextChunk(); return; }
-    function ok() { ttsConsecutiveFailures = 0; }
-    function fail() {
-      if (++ttsConsecutiveFailures >= TTS_MAX_FAILURES) { ttsDisabledUntilReset = true; speechQueue = []; }
-      streamDone = true; isSpeaking = false;
-      if (!ttsDisabledUntilReset) processQueue();
-    }
-    function stream() {
-      if (!streamingSupported) { nonStream(text); return; }
-      fetch(BASE + '/api/tts-stream', { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify({ text: text, voiceId: selectedVoiceId }) })
-        .then(function(r) {
-          if (!r.ok) { streamingSupported = false; throw 0; }
-          var reader = r.body.getReader(), buf = new Uint8Array(0);
-          function cat(a, b) { var c = new Uint8Array(a.length + b.length); c.set(a); c.set(b, a.length); return c; }
-          function pump() { return reader.read().then(function(res) {
-            if (res.done) { ok(); streamDone = true; if (!isPlayingChunk && !audioChunkQueue.length) { isSpeaking = false; processQueue(); } return; }
-            buf = cat(buf, res.value);
-            while (buf.length >= 4) {
-              var len = new DataView(buf.buffer, buf.byteOffset, 4).getUint32(0, false);
-              if (buf.length < 4 + len) break;
-              audioChunkQueue.push(new Blob([buf.slice(4, 4 + len)], { type: 'audio/wav' }));
-              buf = buf.slice(4 + len);
-              if (!isPlayingChunk) playNextChunk();
-            }
-            return pump();
-          }); }
-          return pump();
-        }).catch(function() { nonStream(text); });
-    }
-    function nonStream(txt) {
-      fetch(BASE + '/api/tts', { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify({ text: txt, voiceId: selectedVoiceId }) })
-        .then(function(r) { if (!r.ok) throw 0; return r.arrayBuffer(); })
-        .then(function(b) { ok(); audioChunkQueue.push(new Blob([b], { type: 'audio/wav' })); streamDone = true; if (!isPlayingChunk) playNextChunk(); })
-        .catch(fail);
-    }
-    stream();
-  }
-
-  function stopSpeaking() {
-    speechQueue = []; audioChunkQueue = []; isPlayingChunk = false; isSpeaking = false;
-    ttsConsecutiveFailures = 0; ttsDisabledUntilReset = false;
-    if (currentAudio) { currentAudio.pause(); currentAudio = null; }
-  }
-
-  function sendVoiceToServer() {
-    if (typeof agentGUIClient !== 'undefined' && agentGUIClient && agentGUIClient.wsManager && agentGUIClient.wsManager.isConnected)
-      agentGUIClient.wsManager.sendMessage({ type: 'set_voice', voiceId: selectedVoiceId });
-  }
-
-  window.voiceModule = {
-    getAutoSpeak: function() { return ttsEnabled; },
-    setAutoSpeak: function(v) { ttsEnabled = Boolean(v); localStorage.setItem('gmgui-auto-speak', ttsEnabled); if (!ttsEnabled) stopSpeaking(); },
-    getVoice: function() { return selectedVoiceId; },
-    setVoice: function(id) { selectedVoiceId = String(id); localStorage.setItem('gmgui-voice-selection', selectedVoiceId); sendVoiceToServer(); },
-    speakText: speakDirect,
-    stopSpeaking: stopSpeaking
-  };
-})();
+(function() {
+  var BASE = window.__BASE_URL || '';
+  var ttsEnabled = true;
+  var currentAudio = null;
+  var selectedVoiceId = localStorage.getItem('gmgui-voice-selection') || 'default';
+  var ttsAudioCache = new Map();
+  var TTS_CLIENT_CACHE_MAX = 50;
+  var audioChunkQueue = [];
+  var streamingSupported = true;
+
+  window.addEventListener('ws-message', function(e) {
+    var data = e.detail;
+    if (!data) return;
+    if (data.type === 'tts_audio' && data.audio && data.voiceId === selectedVoiceId) cacheTTSAudio(data.cacheKey, data.audio);
+    if (data.type === 'sync_connected') sendVoiceToServer();
+  });
+
+  var api = function() { return window.voiceMachineAPI; };
+
+  if (window.voiceMachineAPI) {
+    window.voiceMachineAPI.subscribe(function(snapshot) {
+      if (snapshot.value === 'speaking' && snapshot.context.currentText) {
+        _doSpeak(snapshot.context.currentText);
+      }
+    });
+  } else {
+    document.addEventListener('DOMContentLoaded', function() {
+      if (window.voiceMachineAPI) {
+        window.voiceMachineAPI.subscribe(function(snapshot) {
+          if (snapshot.value === 'speaking' && snapshot.context.currentText) {
+            _doSpeak(snapshot.context.currentText);
+          }
+        });
+      }
+    });
+  }
+
+  function cacheTTSAudio(cacheKey, b64) {
+    if (ttsAudioCache.size >= TTS_CLIENT_CACHE_MAX) ttsAudioCache.delete(ttsAudioCache.keys().next().value);
+    var binary = atob(b64), bytes = new Uint8Array(binary.length);
+    for (var i = 0; i < binary.length; i++) bytes[i] = binary.charCodeAt(i);
+    ttsAudioCache.set(cacheKey, new Blob([bytes], { type: 'audio/wav' }));
+  }
+
+  function playNextChunk() {
+    if (!audioChunkQueue.length) {
+      if (api()) api().send({ type: 'PLAYBACK_DONE' });
+      return;
+    }
+    var blob = audioChunkQueue.shift(), url = URL.createObjectURL(blob);
+    currentAudio = new Audio(url);
+    var next = function() { URL.revokeObjectURL(url); currentAudio = null; playNextChunk(); };
+    currentAudio.onended = next;
+    currentAudio.onerror = function() { URL.revokeObjectURL(url); currentAudio = null; if (api()) api().send({ type: 'PLAYBACK_ERROR' }); };
+    currentAudio.play().catch(function() { URL.revokeObjectURL(url); currentAudio = null; if (api()) api().send({ type: 'PLAYBACK_ERROR' }); });
+  }
+
+  function _doSpeak(text) {
+    audioChunkQueue = [];
+    var cached = ttsAudioCache.get(selectedVoiceId + ':' + text);
+    if (cached) { audioChunkQueue.push(cached); playNextChunk(); return; }
+    function stream() {
+      if (!streamingSupported) { nonStream(text); return; }
+      fetch(BASE + '/api/tts-stream', { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify({ text: text, voiceId: selectedVoiceId }) })
+        .then(function(r) {
+          if (!r.ok) { streamingSupported = false; throw 0; }
+          var reader = r.body.getReader(), buf = new Uint8Array(0);
+          function cat(a, b) { var c = new Uint8Array(a.length + b.length); c.set(a); c.set(b, a.length); return c; }
+          function pump() { return reader.read().then(function(res) {
+            if (res.done) { if (!audioChunkQueue.length) { if (api()) api().send({ type: 'PLAYBACK_DONE' }); } return; }
+            buf = cat(buf, res.value);
+            while (buf.length >= 4) {
+              var len = new DataView(buf.buffer, buf.byteOffset, 4).getUint32(0, false);
+              if (buf.length < 4 + len) break;
+              audioChunkQueue.push(new Blob([buf.slice(4, 4 + len)], { type: 'audio/wav' }));
+              buf = buf.slice(4 + len);
+              if (audioChunkQueue.length === 1) playNextChunk();
+            }
+            return pump();
+          }); }
+          return pump();
+        }).catch(function() { nonStream(text); });
+    }
+    function nonStream(txt) {
+      fetch(BASE + '/api/tts', { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify({ text: txt, voiceId: selectedVoiceId }) })
+        .then(function(r) { if (!r.ok) throw 0; return r.arrayBuffer(); })
+        .then(function(b) { audioChunkQueue.push(new Blob([b], { type: 'audio/wav' })); playNextChunk(); })
+        .catch(function() { if (api()) api().send({ type: 'PLAYBACK_ERROR' }); });
+    }
+    stream();
+  }
+
+  function speakDirect(text) {
+    if (!ttsEnabled) return;
+    var vmApi = api();
+    if (vmApi && vmApi.isDisabled()) return;
+    var clean = text.replace(/<[^>]*>/g, '').trim();
+    if (!clean) return;
+    var parts = (typeof agentGUIClient !== 'undefined' && agentGUIClient && typeof agentGUIClient.parseMarkdownCodeBlocks === 'function')
+      ? agentGUIClient.parseMarkdownCodeBlocks(clean) : [{ type: 'text', content: clean }];
+    parts.forEach(function(p) {
+      if (p.type !== 'code' && p.content.trim()) {
+        if (vmApi) vmApi.send({ type: 'SPEAK', text: p.content.trim() });
+      }
+    });
+  }
+
+  function stopSpeaking() {
+    audioChunkQueue = [];
+    if (currentAudio) { currentAudio.pause(); currentAudio = null; }
+    if (api()) api().send({ type: 'STOP' });
+  }
+
+  function sendVoiceToServer() {
+    if (typeof agentGUIClient !== 'undefined' && agentGUIClient && agentGUIClient.wsManager && agentGUIClient.wsManager.isConnected)
+      agentGUIClient.wsManager.sendMessage({ type: 'set_voice', voiceId: selectedVoiceId });
+  }
+
+  window.voiceModule = {
+    getAutoSpeak: function() { return ttsEnabled; },
+    setAutoSpeak: function(v) {
+      ttsEnabled = Boolean(v);
+      localStorage.setItem('gmgui-auto-speak', ttsEnabled);
+      if (!ttsEnabled) stopSpeaking();
+      if (api()) api().send({ type: 'SET_ENABLED', enabled: ttsEnabled });
+    },
+    getVoice: function() { return selectedVoiceId; },
+    setVoice: function(id) { selectedVoiceId = String(id); localStorage.setItem('gmgui-voice-selection', selectedVoiceId); sendVoiceToServer(); },
+    speakText: speakDirect,
+    stopSpeaking: stopSpeaking
+  };
+})();