@eshal-bot/chat-widget 0.1.44 → 0.1.46

package/dist/chat-widget.js

@@ -8589,7 +8589,10 @@
    * DEFAULT_SPEECH_GATE_CONFIG.
    *
    * @param {{ sampleRate: number, energyThreshold?: number, speechStartMs?: number, hangoverMs?: number, preRollMs?: number }} config
-   * @param {{ onVoiceFrame: (pcm: ArrayBuffer) => void, onSpeechEnd: () => void }} callbacks
+   * @param {{ onSpeechStart?: () => void, onVoiceFrame: (pcm: ArrayBuffer) => void, onSpeechEnd: () => void }} callbacks
+   *   onSpeechStart fires the instant a turn opens — use it for barge-in
+   *   (stop any assistant TTS still playing). onVoiceFrame emits each voiced
+   *   PCM frame; onSpeechEnd fires when the turn closes after the hangover.
    */
   function createSpeechGate(config, callbacks) {
     var _config$energyThresho, _config$speechStartMs, _config$hangoverMs, _config$preRollMs;
@@ -8637,10 +8640,14 @@
         if (voiced) {
           voicedRunMs += durationMs;
           if (voicedRunMs >= cfg.speechStartMs) {
-            // Confirmed speech — open the turn and flush the pre-roll (which
-            // contains the voiced onset) before going live.
+            var _callbacks$onSpeechSt;
+            // Confirmed speech — open the turn. Fire onSpeechStart FIRST so the
+            // caller can barge in (stop assistant TTS still playing) before we
+            // start streaming this utterance, then flush the pre-roll (which
+            // contains the voiced onset) and go live.
             speaking = true;
             silenceRunMs = 0;
+            (_callbacks$onSpeechSt = callbacks.onSpeechStart) === null || _callbacks$onSpeechSt === void 0 || _callbacks$onSpeechSt.call(callbacks);
             for (const frame of preRoll) callbacks.onVoiceFrame(frame.pcm);
             preRoll = [];
             preRollMsBuffered = 0;
@@ -9055,6 +9062,27 @@
     };
   };
 
+  // Merge the text (SSE) and voice (Bidi) message arrays into one chronological
+  // transcript: tag text messages with a default source, dedupe by id, and sort
+  // by timestamp. Used for BOTH live display (activeMessages) and the
+  // localStorage snapshot — so voice turns persist across reload exactly like
+  // text turns. (Previously the snapshot saved only the text array, so voice
+  // vanished on reload even though it rendered live and was saved server-side
+  // in the Inbox.)
+  const mergeTranscript = (messages, bidiMessages) => {
+    const allMessages = [...messages.map(msg => _objectSpread2(_objectSpread2({}, msg), {}, {
+      metadata: msg.metadata || {
+        source: "sse"
+      }
+    })), ...bidiMessages];
+    const uniqueMessages = Array.from(new Map(allMessages.map(msg => [msg.id, msg])).values());
+    return uniqueMessages.sort((a, b) => {
+      const timeA = a.timestamp instanceof Date ? a.timestamp.getTime() : new Date(a.timestamp).getTime();
+      const timeB = b.timestamp instanceof Date ? b.timestamp.getTime() : new Date(b.timestamp).getTime();
+      return timeA - timeB;
+    });
+  };
+
   // Helper functions for onboarding
   const getOnboardingQuestions = onboardingQuestions => {
     if (onboardingQuestions && onboardingQuestions.length > 0) {
@@ -9450,8 +9478,12 @@
     reactExports.useEffect(() => {
       if (!organizationId || !bidiSessionId) return;
       if (isConversationLoading || isLoading) return;
-      saveSessionMessages(organizationId, bidiSessionId, messages, resolvedMessageLimits);
-    }, [messages, organizationId, bidiSessionId, isConversationLoading, isLoading, resolvedMessageLimits]);
+      // Persist the MERGED transcript (text + voice). Previously only `messages`
+      // (text) was saved, so bidi/voice turns — which live in `bidiMessages` —
+      // were dropped on reload even though they render live and are saved
+      // server-side (Inbox). Merging here makes voice survive reload like text.
+      saveSessionMessages(organizationId, bidiSessionId, mergeTranscript(messages, bidiMessages), resolvedMessageLimits);
+    }, [messages, bidiMessages, organizationId, bidiSessionId, isConversationLoading, isLoading, resolvedMessageLimits]);
 
     // Show the onboarding form on first load; bypass it if already completed (persisted in session)
     reactExports.useEffect(() => {
@@ -10581,6 +10613,25 @@
         speechGateRef.current = createSpeechGate({
           sampleRate: 16000
         }, {
+          // Barge-in: the instant the user starts talking, stop any assistant
+          // TTS still playing so we don't talk over them. endOfAudio drops the
+          // player's buffered audio immediately (readIndex → writeIndex). No-op
+          // when nothing is playing (e.g. the user's first utterance).
+          onSpeechStart: () => {
+            var _audioPlayerNodeRef$c2;
+            // Barge-in: user started talking → (1) stop assistant TTS already
+            // buffered locally, and (2) tell the backend to cancel the in-flight
+            // reply so its remaining TTS doesn't clobber the next turn's audio.
+            (_audioPlayerNodeRef$c2 = audioPlayerNodeRef.current) === null || _audioPlayerNodeRef$c2 === void 0 || _audioPlayerNodeRef$c2.port.postMessage({
+              command: "endOfAudio"
+            });
+            const ws = websocketRef.current;
+            if (ws && ws.readyState === WebSocket.OPEN) {
+              ws.send(JSON.stringify({
+                type: "interrupt"
+              }));
+            }
+          },
           onVoiceFrame: sendVoicePcm,
           onSpeechEnd: () => {
             const ws = websocketRef.current;
@@ -10753,11 +10804,11 @@
     }, [isVoiceSessionActive]);
     reactExports.useEffect(() => {
       return () => {
-        var _audioPlayerNodeRef$c2, _audioPlayerContextRe2, _audioRecorderNodeRef2, _audioRecorderContext2;
+        var _audioPlayerNodeRef$c3, _audioPlayerContextRe2, _audioRecorderNodeRef2, _audioRecorderContext2;
         if (websocketRef.current && websocketRef.current.readyState === WebSocket.OPEN) {
           websocketRef.current.close();
         }
-        (_audioPlayerNodeRef$c2 = audioPlayerNodeRef.current) === null || _audioPlayerNodeRef$c2 === void 0 || _audioPlayerNodeRef$c2.disconnect();
+        (_audioPlayerNodeRef$c3 = audioPlayerNodeRef.current) === null || _audioPlayerNodeRef$c3 === void 0 || _audioPlayerNodeRef$c3.disconnect();
         (_audioPlayerContextRe2 = audioPlayerContextRef.current) === null || _audioPlayerContextRe2 === void 0 || _audioPlayerContextRe2.close();
         (_audioRecorderNodeRef2 = audioRecorderNodeRef.current) === null || _audioRecorderNodeRef2 === void 0 || _audioRecorderNodeRef2.disconnect();
         (_audioRecorderContext2 = audioRecorderContextRef.current) === null || _audioRecorderContext2 === void 0 || _audioRecorderContext2.close();
@@ -10924,28 +10975,7 @@
 
     // Merge and sort messages from both sources (text and voice) by timestamp
     // This ensures all messages persist regardless of voice session state
-    const activeMessages = reactExports.useMemo(() => {
-      // Merge both message arrays
-      const allMessages = [
-      // Convert regular messages to include metadata if not present
-      ...messages.map(msg => _objectSpread2(_objectSpread2({}, msg), {}, {
-        metadata: msg.metadata || {
-          source: "sse"
-        }
-      })),
-      // Include bidi messages
-      ...bidiMessages];
-
-      // Remove duplicates based on message ID (in case same message exists in both arrays)
-      const uniqueMessages = Array.from(new Map(allMessages.map(msg => [msg.id, msg])).values());
-
-      // Sort by timestamp to maintain chronological order
-      return uniqueMessages.sort((a, b) => {
-        const timeA = a.timestamp instanceof Date ? a.timestamp.getTime() : new Date(a.timestamp).getTime();
-        const timeB = b.timestamp instanceof Date ? b.timestamp.getTime() : new Date(b.timestamp).getTime();
-        return timeA - timeB;
-      });
-    }, [bidiMessages, messages]);
+    const activeMessages = reactExports.useMemo(() => mergeTranscript(messages, bidiMessages), [bidiMessages, messages]);
 
     // Log message state changes for debugging
     reactExports.useEffect(() => {