agentgui 1.0.174 → 1.0.175

package/lib/speech.js

@@ -115,6 +115,8 @@ let speakerEmbeddingPipeline = null;
 let sttLoading = false;
 let ttsLoading = false;
 let speakerEmbeddingLoading = false;
+let ttsLoadError = null;
+let sttLoadError = null;
 const voiceEmbeddingsCache = new Map();
 const SAMPLE_RATE_STT = 16000;
 const SAMPLE_RATE_TTS = 16000;
@@ -261,8 +263,10 @@ async function generateEmbeddingFromCustomVoice(voiceId) {
 
 async function getSTT() {
   if (sttPipeline) return sttPipeline;
+  if (sttLoadError) throw sttLoadError;
   if (sttLoading) {
     while (sttLoading) await new Promise(r => setTimeout(r, 100));
+    if (sttLoadError) throw sttLoadError;
     if (!sttPipeline) throw new Error('STT pipeline failed to load');
     return sttPipeline;
   }
@@ -278,10 +282,12 @@ async function getSTT() {
       device: 'cpu',
       local_files_only: isLocal,
     });
+    sttLoadError = null;
     return sttPipeline;
   } catch (err) {
     sttPipeline = null;
-    throw new Error('STT model load failed: ' + err.message);
+    sttLoadError = new Error('STT model load failed: ' + err.message);
+    throw sttLoadError;
   } finally {
     sttLoading = false;
   }
@@ -289,8 +295,10 @@ async function getSTT() {
 
 async function getTTS() {
   if (ttsPipeline) return ttsPipeline;
+  if (ttsLoadError) throw ttsLoadError;
   if (ttsLoading) {
     while (ttsLoading) await new Promise(r => setTimeout(r, 100));
+    if (ttsLoadError) throw ttsLoadError;
     if (!ttsPipeline) throw new Error('TTS pipeline failed to load');
     return ttsPipeline;
   }
@@ -303,10 +311,12 @@ async function getTTS() {
       dtype: 'fp32',
     });
     await ensureSpeakerEmbeddings();
+    ttsLoadError = null;
     return ttsPipeline;
   } catch (err) {
     ttsPipeline = null;
-    throw new Error('TTS model load failed: ' + err.message);
+    ttsLoadError = new Error('TTS model load failed: ' + err.message);
+    throw ttsLoadError;
   } finally {
     ttsLoading = false;
   }
@@ -482,6 +492,8 @@ function getStatus() {
     ttsReady: !!ttsPipeline,
     sttLoading,
     ttsLoading,
+    sttError: sttLoadError ? sttLoadError.message : null,
+    ttsError: ttsLoadError ? ttsLoadError.message : null,
   };
 }
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentgui",
-  "version": "1.0.174",
+  "version": "1.0.175",
   "description": "Multi-agent ACP client with real-time communication",
   "type": "module",
   "main": "server.js",

package/server.js

@@ -554,13 +554,20 @@ const server = http.createServer(async (req, res) => {
           sendJSON(req, res, 400, { error: 'No text provided' });
           return;
         }
-        const { synthesize } = await getSpeech();
-        const wavBuffer = await synthesize(text, voiceId);
+        const speech = await getSpeech();
+        const status = speech.getStatus();
+        if (status.ttsError) {
+          sendJSON(req, res, 503, { error: status.ttsError, retryable: false });
+          return;
+        }
+        const wavBuffer = await speech.synthesize(text, voiceId);
         res.writeHead(200, { 'Content-Type': 'audio/wav', 'Content-Length': wavBuffer.length });
         res.end(wavBuffer);
       } catch (err) {
         debugLog('[TTS] Error: ' + err.message);
-        if (!res.headersSent) sendJSON(req, res, 500, { error: err.message || 'TTS failed' });
+        const isModelError = /model.*load|pipeline.*failed|failed to load/i.test(err.message);
+        const statusCode = isModelError ? 503 : 500;
+        if (!res.headersSent) sendJSON(req, res, statusCode, { error: err.message || 'TTS failed', retryable: !isModelError });
       }
       return;
     }
@@ -574,14 +581,19 @@ const server = http.createServer(async (req, res) => {
           sendJSON(req, res, 400, { error: 'No text provided' });
           return;
         }
-        const { synthesizeStream } = await getSpeech();
+        const speech = await getSpeech();
+        const status = speech.getStatus();
+        if (status.ttsError) {
+          sendJSON(req, res, 503, { error: status.ttsError, retryable: false });
+          return;
+        }
         res.writeHead(200, {
           'Content-Type': 'application/octet-stream',
           'Transfer-Encoding': 'chunked',
           'X-Content-Type': 'audio/wav-stream',
           'Cache-Control': 'no-cache'
         });
-        for await (const wavChunk of synthesizeStream(text, voiceId)) {
+        for await (const wavChunk of speech.synthesizeStream(text, voiceId)) {
           const lenBuf = Buffer.alloc(4);
           lenBuf.writeUInt32BE(wavChunk.length, 0);
           res.write(lenBuf);
@@ -590,7 +602,9 @@ const server = http.createServer(async (req, res) => {
         res.end();
       } catch (err) {
         debugLog('[TTS-STREAM] Error: ' + err.message);
-        if (!res.headersSent) sendJSON(req, res, 500, { error: err.message || 'TTS stream failed' });
+        const isModelError = /model.*load|pipeline.*failed|failed to load/i.test(err.message);
+        const statusCode = isModelError ? 503 : 500;
+        if (!res.headersSent) sendJSON(req, res, statusCode, { error: err.message || 'TTS stream failed', retryable: !isModelError });
         else res.end();
       }
       return;

package/static/js/voice.js

@@ -298,6 +298,9 @@
   var audioChunkQueue = [];
   var isPlayingChunk = false;
   var streamDone = false;
+  var ttsConsecutiveFailures = 0;
+  var TTS_MAX_FAILURES = 3;
+  var ttsDisabledUntilReset = false;
 
   function playNextChunk() {
     if (audioChunkQueue.length === 0) {
@@ -331,19 +334,41 @@
 
   function processQueue() {
     if (isSpeaking || speechQueue.length === 0) return;
+    if (ttsDisabledUntilReset) {
+      speechQueue = [];
+      return;
+    }
     isSpeaking = true;
     streamDone = false;
     var text = speechQueue.shift();
     audioChunkQueue = [];
     isPlayingChunk = false;
-    
+
+    function onTtsSuccess() {
+      ttsConsecutiveFailures = 0;
+    }
+
+    function onTtsFailed() {
+      ttsConsecutiveFailures++;
+      if (ttsConsecutiveFailures >= TTS_MAX_FAILURES) {
+        console.warn('[Voice] TTS failed ' + ttsConsecutiveFailures + ' times consecutively, disabling until reset');
+        ttsDisabledUntilReset = true;
+        speechQueue = [];
+      }
+      streamDone = true;
+      isSpeaking = false;
+      if (!ttsDisabledUntilReset) {
+        processQueue();
+      }
+    }
+
     function tryStreaming() {
       fetch(BASE + '/api/tts-stream', {
         method: 'POST',
         headers: { 'Content-Type': 'application/json' },
         body: JSON.stringify({ text: text, voiceId: selectedVoiceId })
       }).then(function(resp) {
-        if (!resp.ok) throw new Error('TTS stream failed');
+        if (!resp.ok) throw new Error('TTS stream failed: ' + resp.status);
         var reader = resp.body.getReader();
         var buffer = new Uint8Array(0);
 
@@ -357,6 +382,7 @@
         function pump() {
           return reader.read().then(function(result) {
             if (result.done) {
+              onTtsSuccess();
               streamDone = true;
               if (!isPlayingChunk && audioChunkQueue.length === 0) {
                 isSpeaking = false;
@@ -384,16 +410,17 @@
         tryNonStreaming(text);
       });
     }
-    
+
     function tryNonStreaming(txt) {
       fetch(BASE + '/api/tts', {
         method: 'POST',
         headers: { 'Content-Type': 'application/json' },
         body: JSON.stringify({ text: txt, voiceId: selectedVoiceId })
       }).then(function(resp) {
-        if (!resp.ok) throw new Error('TTS failed');
+        if (!resp.ok) throw new Error('TTS failed: ' + resp.status);
         return resp.arrayBuffer();
       }).then(function(buf) {
+        onTtsSuccess();
         var blob = new Blob([buf], { type: 'audio/wav' });
         audioChunkQueue.push(blob);
         if (!isPlayingChunk) playNextChunk();
@@ -401,12 +428,10 @@
         isSpeaking = false;
         processQueue();
       }).catch(function() {
-        streamDone = true;
-        isSpeaking = false;
-        processQueue();
+        onTtsFailed();
       });
     }
-    
+
     tryStreaming();
   }
 
@@ -415,6 +440,8 @@
     audioChunkQueue = [];
     isPlayingChunk = false;
     isSpeaking = false;
+    ttsConsecutiveFailures = 0;
+    ttsDisabledUntilReset = false;
     if (currentAudio) {
       currentAudio.pause();
       currentAudio = null;