agentgui 1.0.215 → 1.0.217

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentgui",
-  "version": "1.0.215",
+  "version": "1.0.217",
   "description": "Multi-agent ACP client with real-time communication",
   "type": "module",
   "main": "server.js",

package/server.js

@@ -99,7 +99,7 @@ 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. Write as if you are speaking to someone in a casual conversation.`;
+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.`;
 
 const activeExecutions = new Map();
 const activeScripts = new Map();

package/static/js/voice.js

@@ -13,6 +13,7 @@
   var recordedChunks = [];
   var TARGET_SAMPLE_RATE = 16000;
   var spokenChunks = new Set();
+  var renderedSeqs = new Set();
   var isLoadingHistory = false;
   var selectedVoiceId = localStorage.getItem('voice-selected-id') || 'default';
   var ttsAudioCache = new Map();
@@ -313,6 +314,12 @@
     return ttsAudioCache.get(key) || null;
   }
 
+  function splitSentences(text) {
+    var raw = text.match(/[^.!?]+[.!?]+[\s]?|[^.!?]+$/g);
+    if (!raw) return [text];
+    return raw.map(function(s) { return s.trim(); }).filter(function(s) { return s.length > 0; });
+  }
+
   var audioChunkQueue = [];
   var isPlayingChunk = false;
   var streamDone = false;
@@ -373,6 +380,38 @@
       return;
     }
 
+    var sentences = splitSentences(text);
+    var cachedSentences = [];
+    var uncachedText = [];
+    for (var i = 0; i < sentences.length; i++) {
+      var blob = getCachedTTSBlob(sentences[i]);
+      if (blob) {
+        cachedSentences.push({ idx: i, blob: blob });
+      } else {
+        uncachedText.push(sentences[i]);
+      }
+    }
+
+    if (cachedSentences.length === sentences.length) {
+      ttsConsecutiveFailures = 0;
+      for (var j = 0; j < cachedSentences.length; j++) {
+        audioChunkQueue.push(cachedSentences[j].blob);
+      }
+      streamDone = true;
+      if (!isPlayingChunk) playNextChunk();
+      return;
+    }
+
+    if (cachedSentences.length > 0) {
+      ttsConsecutiveFailures = 0;
+      for (var k = 0; k < cachedSentences.length; k++) {
+        audioChunkQueue.push(cachedSentences[k].blob);
+      }
+      if (!isPlayingChunk) playNextChunk();
+    }
+
+    var remainingText = uncachedText.join(' ');
+
     function onTtsSuccess() {
       ttsConsecutiveFailures = 0;
     }
@@ -392,11 +431,11 @@
     }
 
     function tryStreaming() {
-      if (!streamingSupported) { tryNonStreaming(text); return; }
+      if (!streamingSupported) { tryNonStreaming(remainingText); return; }
       fetch(BASE + '/api/tts-stream', {
         method: 'POST',
         headers: { 'Content-Type': 'application/json' },
-        body: JSON.stringify({ text: text, voiceId: selectedVoiceId })
+        body: JSON.stringify({ text: remainingText, voiceId: selectedVoiceId })
       }).then(function(resp) {
         if (!resp.ok) {
           streamingSupported = false;
@@ -441,7 +480,7 @@
 
         return pump();
       }).catch(function() {
-        tryNonStreaming(text);
+        tryNonStreaming(remainingText);
       });
     }
 
@@ -583,17 +622,21 @@
       if (!voiceActive) return;
       if (data.type === 'streaming_progress' && data.block) {
         if (data.conversationId && data.conversationId !== currentConversationId) return;
+        if (data.seq !== undefined && renderedSeqs.has(data.seq)) return;
+        if (data.seq !== undefined) renderedSeqs.add(data.seq);
         handleVoiceBlock(data.block, true);
       }
       if (data.type === 'streaming_start') {
         if (data.conversationId && data.conversationId !== currentConversationId) return;
         spokenChunks = new Set();
+        renderedSeqs = new Set();
       }
     });
     window.addEventListener('conversation-selected', function(e) {
       currentConversationId = e.detail.conversationId;
       stopSpeaking();
       spokenChunks = new Set();
+      renderedSeqs = new Set();
       if (voiceActive) {
         loadVoiceBlocks(currentConversationId);
       }
@@ -633,6 +676,7 @@
         }
         var hasContent = false;
         data.chunks.forEach(function(chunk) {
+          if (chunk.sequence !== undefined) renderedSeqs.add(chunk.sequence);
           var block = typeof chunk.data === 'string' ? JSON.parse(chunk.data) : chunk.data;
           if (!block) return;
           if (block.type === 'text' && block.text) {