@eshal-bot/chat-widget 0.1.42 → 0.1.44

package/dist/chat-widget.js

@@ -8537,6 +8537,144 @@
     return pcm16Array.buffer;
   }
 
+  /**
+   * Client-side Voice-Activity gate for the widget's voice (Bidi) session.
+   *
+   * Ported from packages/agent-sdk/src/voice/audio/vad.ts (`createSpeechGate`),
+   * which was itself originally derived from this widget. The gate sits between
+   * the mic worklet and the WebSocket: it forwards PCM ONLY inside a confirmed
+   * speech turn and fires `onSpeechEnd` when the turn closes, so the caller can
+   * tell the backend the utterance is done (an `audio_end` signal).
+   *
+   * Why the widget needs this: knowledge-api's non-streaming voice path only
+   * transcribes a turn when it receives `audio_end` (or its flaky energy-VAD
+   * backstop trips). Without a client gate the widget streamed mic audio forever
+   * and the backend just buffered it ("Collecting Audio…") without ever
+   * transcribing — so voice connected but never replied. The agent-sdk already
+   * gates this way; this brings the widget to parity.
+   *
+   * Framework-free and side-effect-free (only calls the two callbacks) so it can
+   * be reasoned about / tested in isolation.
+   */
+
+  const DEFAULT_SPEECH_GATE_CONFIG = {
+    // RMS energy (0..1) at/above which a frame counts as voiced. Speech is
+    // typically 0.05–0.3; quiet room tone sits well under 0.02.
+    energyThreshold: 0.025,
+    // Continuous voiced audio (ms) required to OPEN a turn — rejects coughs/clicks.
+    speechStartMs: 200,
+    // Continuous silence (ms) after speech that CLOSES the turn. Long enough to
+    // ride over natural mid-sentence pauses, short enough to feel responsive.
+    hangoverMs: 1000,
+    // How much audio (ms) preceding turn-open to retain + flush so the leading
+    // phoneme of the first word survives the start-detection delay.
+    preRollMs: 300
+  };
+
+  /** RMS amplitude of an Int16 PCM frame (ArrayBuffer), normalised to 0..1. */
+  function computeRms(pcm) {
+    const samples = new Int16Array(pcm);
+    if (samples.length === 0) return 0;
+    let sumSquares = 0;
+    for (let i = 0; i < samples.length; i += 1) {
+      const v = samples[i] / 0x8000; // normalise Int16 → -1..1
+      sumSquares += v * v;
+    }
+    return Math.sqrt(sumSquares / samples.length);
+  }
+
+  /**
+   * Create a stateful speech gate. Returns `{ process, reset, isSpeaking }`.
+   * `config.sampleRate` is required; everything else falls back to
+   * DEFAULT_SPEECH_GATE_CONFIG.
+   *
+   * @param {{ sampleRate: number, energyThreshold?: number, speechStartMs?: number, hangoverMs?: number, preRollMs?: number }} config
+   * @param {{ onVoiceFrame: (pcm: ArrayBuffer) => void, onSpeechEnd: () => void }} callbacks
+   */
+  function createSpeechGate(config, callbacks) {
+    var _config$energyThresho, _config$speechStartMs, _config$hangoverMs, _config$preRollMs;
+    const cfg = {
+      sampleRate: config.sampleRate,
+      energyThreshold: (_config$energyThresho = config.energyThreshold) !== null && _config$energyThresho !== void 0 ? _config$energyThresho : DEFAULT_SPEECH_GATE_CONFIG.energyThreshold,
+      speechStartMs: (_config$speechStartMs = config.speechStartMs) !== null && _config$speechStartMs !== void 0 ? _config$speechStartMs : DEFAULT_SPEECH_GATE_CONFIG.speechStartMs,
+      hangoverMs: (_config$hangoverMs = config.hangoverMs) !== null && _config$hangoverMs !== void 0 ? _config$hangoverMs : DEFAULT_SPEECH_GATE_CONFIG.hangoverMs,
+      preRollMs: (_config$preRollMs = config.preRollMs) !== null && _config$preRollMs !== void 0 ? _config$preRollMs : DEFAULT_SPEECH_GATE_CONFIG.preRollMs
+    };
+    let speaking = false;
+    // While idle: how much consecutive voiced audio we've seen (to open).
+    let voicedRunMs = 0;
+    // While speaking: how much consecutive silence we've seen (to close).
+    let silenceRunMs = 0;
+    // Ring of recent frames kept only while idle, for pre-roll on open.
+    let preRoll = [];
+    let preRollMsBuffered = 0;
+
+    // Int16 → 2 bytes/sample; ms = samples / sampleRate * 1000.
+    const frameDurationMs = pcm => pcm.byteLength / 2 / cfg.sampleRate * 1000;
+    const reset = () => {
+      speaking = false;
+      voicedRunMs = 0;
+      silenceRunMs = 0;
+      preRoll = [];
+      preRollMsBuffered = 0;
+    };
+    const process = pcm => {
+      const durationMs = frameDurationMs(pcm);
+      const voiced = computeRms(pcm) >= cfg.energyThreshold;
+      if (!speaking) {
+        // Keep a bounded pre-roll of the most recent audio so we don't clip the
+        // onset once we decide this is speech.
+        preRoll.push({
+          pcm,
+          durationMs
+        });
+        preRollMsBuffered += durationMs;
+        while (preRollMsBuffered > cfg.preRollMs && preRoll.length > 1) {
+          var _dropped$durationMs;
+          const dropped = preRoll.shift();
+          preRollMsBuffered -= (_dropped$durationMs = dropped === null || dropped === void 0 ? void 0 : dropped.durationMs) !== null && _dropped$durationMs !== void 0 ? _dropped$durationMs : 0;
+        }
+        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.
+            speaking = true;
+            silenceRunMs = 0;
+            for (const frame of preRoll) callbacks.onVoiceFrame(frame.pcm);
+            preRoll = [];
+            preRollMsBuffered = 0;
+          }
+        } else {
+          // Need *sustained* energy — reset on any silent frame so a brief
+          // transient can't accumulate its way over the threshold.
+          voicedRunMs = 0;
+        }
+        return;
+      }
+
+      // Inside a confirmed turn: forward everything, including short internal
+      // pauses, until the hangover elapses.
+      callbacks.onVoiceFrame(pcm);
+      if (voiced) {
+        silenceRunMs = 0;
+      } else {
+        silenceRunMs += durationMs;
+        if (silenceRunMs >= cfg.hangoverMs) {
+          speaking = false;
+          voicedRunMs = 0;
+          silenceRunMs = 0;
+          callbacks.onSpeechEnd();
+        }
+      }
+    };
+    return {
+      process,
+      reset,
+      isSpeaking: () => speaking
+    };
+  }
+
   /**
    * Clean CJK (Chinese, Japanese, Korean) spaces
    */
@@ -8935,6 +9073,11 @@
       openDelay,
       darkMode,
       enableVoiceInteraction = false,
+      // behavior.provideCitations — forwarded as enable_citation on the chat
+      // request so the backend skips citation generation when disabled. The
+      // Sources render is also gated in MessageBubble (defence-in-depth, same
+      // as the dashboard testing lab). Defaults to true.
+      provideCitations = true,
       onboardingQuestions = [],
       onboardingEnabled = false,
       collectionPrompt,
@@ -9054,6 +9197,11 @@
     const audioRecorderNodeRef = reactExports.useRef(null);
     const audioRecorderContextRef = reactExports.useRef(null);
     const micStreamRef = reactExports.useRef(null);
+    // Client-side VAD gate. Forwards only voiced PCM upstream and fires
+    // `audio_end` on end-of-speech so the backend's non-streaming voice path
+    // transcribes the turn (mirrors the agent-sdk). Without it the backend
+    // buffers audio forever and never responds.
+    const speechGateRef = reactExports.useRef(null);
     const currentInputMessageIdRef = reactExports.useRef(null);
     const currentAssistantMessageIdRef = reactExports.useRef(null);
     const isAudioReadyRef = reactExports.useRef(false);
@@ -9458,6 +9606,11 @@
           message: trimmed,
           type: apiType,
           conversationId: bidiSessionId,
+          // Citation toggle (behavior.provideCitations). Mirrors the testing
+          // lab's useChat.ts so knowledge-api skips citation generation when
+          // the admin disabled them. The Sources panel is also gated client-
+          // side in MessageBubble (defence-in-depth). Defaults to on.
+          enable_citation: provideCitations !== null && provideCitations !== void 0 ? provideCitations : true,
           // Identify this as a widget call so the knowledge-api side
           // (chat_utils.EXTERNAL_CHAT_CHANNELS in services/
           // knowledge-api) can enforce the visibility ACL — private
@@ -9562,7 +9715,7 @@
           setIsLoading(false);
         }
       }
-    }, [apiBaseUrl, apiKey, getNextMessageId, organizationId, messages, bidiSessionId, routingChoice, widgetFilter]);
+    }, [apiBaseUrl, apiKey, getNextMessageId, organizationId, messages, bidiSessionId, routingChoice, widgetFilter, provideCitations]);
     const handleOnboardingFormSubmit = reactExports.useCallback(async answers => {
       const questions = getOnboardingQuestions(onboardingQuestions);
       if (questions.length === 0) return;
@@ -10183,7 +10336,7 @@
       }));
     }, []);
     const stopVoiceSession = reactExports.useCallback(() => {
-      var _audioPlayerNodeRef$c, _audioPlayerContextRe, _audioRecorderNodeRef, _audioRecorderContext;
+      var _audioPlayerNodeRef$c, _audioPlayerContextRe, _audioRecorderNodeRef, _audioRecorderContext, _speechGateRef$curren;
       setIsVoiceSessionActive(false);
       setIsLoading(false);
       isVoiceSessionActiveRef.current = false;
@@ -10207,6 +10360,8 @@
       audioRecorderContextRef.current = null;
       stopMicrophone(micStreamRef.current);
       micStreamRef.current = null;
+      (_speechGateRef$curren = speechGateRef.current) === null || _speechGateRef$curren === void 0 || _speechGateRef$curren.reset();
+      speechGateRef.current = null;
       isAudioReadyRef.current = false;
     }, []);
     const handleBidiEvent = reactExports.useCallback(event => {
@@ -10386,11 +10541,22 @@
       }
     }, [appendAssistantContent, appendUserTranscription, handleContentParts, handleInterrupted, handleTurnComplete, stopVoiceSession]);
     const handleRecorderAudio = reactExports.useCallback(pcmData => {
-      const ws = websocketRef.current;
-      if (!isVoiceSessionActiveRef.current || !ws || ws.readyState !== WebSocket.OPEN) {
+      if (!isVoiceSessionActiveRef.current) {
+        return;
+      }
+      // Route mic frames through the VAD gate: it forwards only voiced frames
+      // (onVoiceFrame → ws.send) and fires onSpeechEnd → audio_end so the
+      // backend finalizes the turn. Fall back to raw streaming only if the
+      // gate somehow wasn't built.
+      const gate = speechGateRef.current;
+      if (gate) {
+        gate.process(pcmData);
         return;
       }
-      ws.send(pcmData);
+      const ws = websocketRef.current;
+      if (ws && ws.readyState === WebSocket.OPEN) {
+        ws.send(pcmData);
+      }
     }, []);
     const initializeAudioPipeline = reactExports.useCallback(async () => {
       if (isAudioReadyRef.current || typeof window === "undefined") {
@@ -10401,6 +10567,30 @@
         audioPlayerNodeRef.current = playerNode;
         audioPlayerContextRef.current = playerContext;
         await playerContext.resume();
+
+        // Build the VAD gate before the recorder starts producing frames.
+        // onVoiceFrame streams voiced PCM upstream; onSpeechEnd tells the backend
+        // the utterance is complete so it transcribes + responds. The recorder
+        // runs at 16 kHz (see audio-recorder.js).
+        const sendVoicePcm = pcm => {
+          const ws = websocketRef.current;
+          if (ws && ws.readyState === WebSocket.OPEN) {
+            ws.send(pcm);
+          }
+        };
+        speechGateRef.current = createSpeechGate({
+          sampleRate: 16000
+        }, {
+          onVoiceFrame: sendVoicePcm,
+          onSpeechEnd: () => {
+            const ws = websocketRef.current;
+            if (ws && ws.readyState === WebSocket.OPEN) {
+              ws.send(JSON.stringify({
+                type: "audio_end"
+              }));
+            }
+          }
+        });
         const [recorderNode, recorderContext, micStream] = await startAudioRecorderWorklet(handleRecorderAudio);
         audioRecorderNodeRef.current = recorderNode;
         audioRecorderContextRef.current = recorderContext;
@@ -54632,6 +54822,7 @@
       organizationId,
       onPromptSuggestionClick,
       allowFeedback = true,
+      provideCitations = true,
       onFeedback,
       headerTextColor
     } = _ref17;
@@ -54740,7 +54931,7 @@
                 }), message.decisionResponse && /*#__PURE__*/jsxRuntimeExports.jsxs("div", {
                   className: "mt-2 text-xs ".concat(isDark ? "text-gray-400" : "text-gray-500"),
                   children: ["Selected: ", message.decisionResponse.selectedOption]
-                }), ((_message$sources = message.sources) === null || _message$sources === void 0 ? void 0 : _message$sources.length) > 0 && /*#__PURE__*/jsxRuntimeExports.jsxs(jsxRuntimeExports.Fragment, {
+                }), provideCitations !== false && ((_message$sources = message.sources) === null || _message$sources === void 0 ? void 0 : _message$sources.length) > 0 && /*#__PURE__*/jsxRuntimeExports.jsxs(jsxRuntimeExports.Fragment, {
                   children: [/*#__PURE__*/jsxRuntimeExports.jsx("div", {
                     className: "mt-3",
                     style: {
@@ -54765,7 +54956,7 @@
                   textColor: messageTextColor,
                   fontFamily: messageFontFamily,
                   fontSize: messageFontSize
-                }), ((_message$sources2 = message.sources) === null || _message$sources2 === void 0 ? void 0 : _message$sources2.length) > 0 && /*#__PURE__*/jsxRuntimeExports.jsxs(jsxRuntimeExports.Fragment, {
+                }), provideCitations !== false && ((_message$sources2 = message.sources) === null || _message$sources2 === void 0 ? void 0 : _message$sources2.length) > 0 && /*#__PURE__*/jsxRuntimeExports.jsxs(jsxRuntimeExports.Fragment, {
                   children: [/*#__PURE__*/jsxRuntimeExports.jsx("div", {
                     className: "mt-3",
                     style: {
@@ -54908,6 +55099,7 @@
       organizationId,
       onPromptSuggestionClick,
       allowFeedback,
+      provideCitations,
       onFeedback,
       headerTextColor
     } = _ref19;
@@ -54951,6 +55143,7 @@
       organizationId: organizationId,
       onPromptSuggestionClick: onPromptSuggestionClick,
       allowFeedback: allowFeedback,
+      provideCitations: provideCitations,
       onFeedback: onFeedback,
       headerTextColor: headerTextColor
     });
@@ -55806,6 +55999,7 @@
       onPromptSuggestionClick,
       conversationResetKey,
       allowFeedback,
+      provideCitations,
       onFeedback,
       headerTextColor,
       // Inline filter slots — rendered in the initial-state region
@@ -55961,6 +56155,7 @@
         apiBaseUrl: apiBaseUrl,
         organizationId: organizationId,
         allowFeedback: allowFeedback,
+        provideCitations: provideCitations,
         onFeedback: onFeedback,
         headerTextColor: headerTextColor
       }, message.id)), (() => {
@@ -59582,6 +59777,7 @@
       onResetConversation,
       autoPromptEnabled = true,
       allowFeedback = true,
+      provideCitations = true,
       onFeedback,
       onRefreshChat,
       onCopyChat,
@@ -60120,6 +60316,7 @@
               organizationId: organizationId,
               conversationResetKey: conversationId,
               allowFeedback: allowFeedback,
+              provideCitations: provideCitations,
               onFeedback: onFeedback,
               headerTextColor: headerTextColor,
               beforePromptsSlot: showInlineFilter ? /*#__PURE__*/jsxRuntimeExports.jsx(CategoryFilterInline, {
@@ -60457,7 +60654,7 @@
 
     // Map API response to widget props (use props as overrides if provided)
     const widgetConfig = reactExports.useMemo(() => {
-      var _agentConfig$config, _cfg$identity, _cfg$branding, _cfg$typography, _cfg$messages, _cfg$voice, _cfg$onboarding, _cfg$disclaimer, _cfg$csat, _cfg$inactivity, _cfg$behavior, _cfg$security, _cfg$layout, _ref2, _messages$bundles$loc, _messages$bundles, _messages$bundles$en, _messages$bundles2, _ref3, _ref4, _ref5, _ref6, _branding$chatbotAvat, _branding$chatbotAvat2, _ref7, _ref8, _ref9, _ref0, _ref1, _ref10, _ref11, _ref12, _ref13, _ref14, _ref15, _ref16, _ref17, _ref18, _inactivity$timeoutVa, _inactivity$timeoutUn, _ref19, _ref20, _ref21, _ref22, _ref23, _ref24, _ref25, _ref26, _csat$enabled, _csat$format, _csat$prompt, _csat$triggerType, _csat$idleTimeoutMins, _csat$followUpEnabled, _csat$followUpPrompt, _behavior$autoPromptE, _behavior$allowFeedba;
+      var _agentConfig$config, _cfg$identity, _cfg$branding, _cfg$typography, _cfg$messages, _cfg$voice, _cfg$onboarding, _cfg$disclaimer, _cfg$csat, _cfg$inactivity, _cfg$behavior, _cfg$security, _cfg$layout, _ref2, _messages$bundles$loc, _messages$bundles, _messages$bundles$en, _messages$bundles2, _ref3, _ref4, _ref5, _ref6, _branding$chatbotAvat, _branding$chatbotAvat2, _ref7, _ref8, _ref9, _ref0, _ref1, _ref10, _ref11, _ref12, _ref13, _ref14, _ref15, _ref16, _ref17, _ref18, _inactivity$timeoutVa, _inactivity$timeoutUn, _ref19, _ref20, _ref21, _ref22, _ref23, _ref24, _ref25, _ref26, _csat$enabled, _csat$format, _csat$prompt, _csat$triggerType, _csat$idleTimeoutMins, _csat$followUpEnabled, _csat$followUpPrompt, _behavior$autoPromptE, _behavior$allowFeedba, _behavior$provideCita;
       if (!agentConfig) {
         return null;
       }
@@ -60548,7 +60745,12 @@
         csatFollowUpEnabled: (_csat$followUpEnabled = csat.followUpEnabled) !== null && _csat$followUpEnabled !== void 0 ? _csat$followUpEnabled : false,
         csatFollowUpPrompt: (_csat$followUpPrompt = csat.followUpPrompt) !== null && _csat$followUpPrompt !== void 0 ? _csat$followUpPrompt : null,
         autoPromptEnabled: (_behavior$autoPromptE = behavior.autoPromptEnabled) !== null && _behavior$autoPromptE !== void 0 ? _behavior$autoPromptE : true,
-        allowFeedback: (_behavior$allowFeedba = behavior.allowFeedback) !== null && _behavior$allowFeedba !== void 0 ? _behavior$allowFeedba : true
+        allowFeedback: (_behavior$allowFeedba = behavior.allowFeedback) !== null && _behavior$allowFeedba !== void 0 ? _behavior$allowFeedba : true,
+        // behavior.provideCitations — when false, hide the Sources panel on
+        // every message and tell the backend to skip citation generation.
+        // Defaults to true so agents that haven't set the field keep showing
+        // citations (matches the testing-lab / chatbot-preview default).
+        provideCitations: (_behavior$provideCita = behavior.provideCitations) !== null && _behavior$provideCita !== void 0 ? _behavior$provideCita : true
       };
     }, [agentConfig, darkMode, primaryColor, position, companyName, conciergeName, companyLogo, welcomeMessage, quickQuestions, quickQuestionsEnabled, textColor, agentMessageBubbleColor, userMessageBoxColor, assistantTextColor, userTextColor, fontFamily, fontSize, defaultLanguage, orgId, headerTextBold, headerTextItalic, enableVoiceInteraction, disclaimerEnabled, disclaimerText, disclaimerPosition, disclaimerLinkLabel, disclaimerLinkUrl, showPoweredBy, onboardingQuestions, onboardingEnabled, collectionPrompt, allowedDomains, agentMessageBubbleFontFamily, agentMessageBubbleFontSize, userMessageBubbleFontFamily, userMessageBubbleFontSize
     // CSAT dependencies
@@ -60632,6 +60834,9 @@
       onboardingQuestions: widgetConfig.onboardingQuestions,
       onboardingEnabled: widgetConfig.onboardingEnabled,
       collectionPrompt: widgetConfig.collectionPrompt,
+      // Gates the chat request's enable_citation flag (text path). Voice
+      // already reads provideCitations from the DB in voice_chat.py.
+      provideCitations: widgetConfig.provideCitations,
       inactivityTimeoutValue: widgetConfig.inactivityTimeoutValue,
       inactivityTimeoutUnit: widgetConfig.inactivityTimeoutUnit,
       // Sliding-window cap for the localStorage transcript. Comes from
@@ -60798,6 +61003,7 @@
           onResetConversation: resetConversation,
           autoPromptEnabled: widgetConfig.autoPromptEnabled,
           allowFeedback: widgetConfig.allowFeedback,
+          provideCitations: widgetConfig.provideCitations,
           onFeedback: handleMessageFeedback,
           onRefreshChat: handleRefreshChat,
           onCopyChat: handleCopyChat