@absolutejs/voice 0.0.22-beta.92 → 0.0.22-beta.93

package/dist/angular/index.js

@@ -1287,7 +1287,13 @@ var createVoiceController = (path, options = {}) => {
     capture = createMicrophoneCapture({
       channelCount: options.capture?.channelCount ?? preset.capture.channelCount,
       onLevel: options.capture?.onLevel,
-      onAudio: (audio) => stream.sendAudio(audio),
+      onAudio: (audio) => {
+        if (options.capture?.onAudio) {
+          options.capture.onAudio(audio, stream.sendAudio);
+          return;
+        }
+        stream.sendAudio(audio);
+      },
       sampleRateHz: options.capture?.sampleRateHz ?? preset.capture.sampleRateHz
     });
     return capture;

package/dist/client/htmxBootstrap.js

@@ -994,7 +994,13 @@ var createVoiceController = (path, options = {}) => {
     capture = createMicrophoneCapture({
       channelCount: options.capture?.channelCount ?? preset.capture.channelCount,
       onLevel: options.capture?.onLevel,
-      onAudio: (audio) => stream.sendAudio(audio),
+      onAudio: (audio) => {
+        if (options.capture?.onAudio) {
+          options.capture.onAudio(audio, stream.sendAudio);
+          return;
+        }
+        stream.sendAudio(audio);
+      },
       sampleRateHz: options.capture?.sampleRateHz ?? preset.capture.sampleRateHz
     });
     return capture;
@@ -1104,6 +1110,77 @@ var createVoiceController = (path, options = {}) => {
   };
 };
 
+// src/client/bargeInMonitor.ts
+var DEFAULT_THRESHOLD_MS = 250;
+var createEventId = () => `barge-in:${Date.now()}:${crypto.randomUUID?.() ?? Math.random().toString(36).slice(2)}`;
+var summarize = (events, thresholdMs) => {
+  const stopped = events.filter((event) => event.status === "stopped");
+  const latencies = stopped.map((event) => event.latencyMs).filter((value) => typeof value === "number");
+  const failed = stopped.filter((event) => typeof event.latencyMs === "number" && event.latencyMs > thresholdMs).length;
+  const passed = stopped.length - failed;
+  return {
+    averageLatencyMs: latencies.length > 0 ? Math.round(latencies.reduce((total, value) => total + value, 0) / latencies.length) : undefined,
+    events: [...events],
+    failed,
+    lastEvent: events.at(-1),
+    passed,
+    status: events.length === 0 ? "empty" : failed > 0 ? "fail" : stopped.length === 0 ? "warn" : "pass",
+    thresholdMs,
+    total: stopped.length
+  };
+};
+var createVoiceBargeInMonitor = (options = {}) => {
+  const listeners = new Set;
+  const thresholdMs = options.thresholdMs ?? DEFAULT_THRESHOLD_MS;
+  const fetchImpl = options.fetch ?? globalThis.fetch;
+  const events = [];
+  const emit = () => {
+    for (const listener of listeners) {
+      listener();
+    }
+  };
+  const postEvent = (event) => {
+    if (!options.path || typeof fetchImpl !== "function") {
+      return;
+    }
+    fetchImpl(options.path, {
+      body: JSON.stringify(event),
+      headers: {
+        "Content-Type": "application/json"
+      },
+      method: "POST"
+    }).catch(() => {});
+  };
+  const record = (status, input) => {
+    const event = {
+      at: Date.now(),
+      id: createEventId(),
+      latencyMs: input.latencyMs,
+      playbackStopLatencyMs: input.playbackStopLatencyMs,
+      reason: input.reason,
+      sessionId: input.sessionId,
+      status,
+      thresholdMs
+    };
+    events.push(event);
+    postEvent(event);
+    emit();
+    return event;
+  };
+  return {
+    getSnapshot: () => summarize(events, thresholdMs),
+    recordRequested: (input) => record("requested", input),
+    recordSkipped: (input) => record("skipped", input),
+    recordStopped: (input) => record("stopped", input),
+    subscribe: (subscriber) => {
+      listeners.add(subscriber);
+      return () => {
+        listeners.delete(subscriber);
+      };
+    }
+  };
+};
+
 // src/client/htmxBootstrap.ts
 var VOICE_WAVE_POINTS = 48;
 var VOICE_WAVE_WIDTH = 320;
@@ -1216,6 +1293,13 @@ var parsePromptList = (value) => {
   } catch {}
   return DEFAULT_GUIDED_PROMPTS;
 };
+var parseOptionalNumber = (value) => {
+  if (!value) {
+    return;
+  }
+  const parsed = Number(value);
+  return Number.isFinite(parsed) ? parsed : undefined;
+};
 var requireElement = (root, selector, ctor, name) => {
   const value = selector ? document.querySelector(selector) : null;
   if (value instanceof ctor) {
@@ -1266,6 +1350,13 @@ var initVoiceHTMXRoot = (root) => {
   const guidedPrompts = parsePromptList(root.dataset.voiceGuidedPrompts);
   const guidedLabel = root.dataset.voiceGuidedLabel ?? DEFAULT_GUIDED_LABEL;
   const generalLabel = root.dataset.voiceGeneralLabel ?? DEFAULT_GENERAL_LABEL;
+  const bargeInPath = root.dataset.voiceBargeInPath;
+  const bargeInMonitor = bargeInPath ? createVoiceBargeInMonitor({
+    path: bargeInPath,
+    thresholdMs: parseOptionalNumber(root.dataset.voiceBargeInThresholdMs)
+  }) : null;
+  const bargeInRecentWindowMs = parseOptionalNumber(root.dataset.voiceBargeInRecentWindowMs) ?? 4000;
+  const bargeInSpeechThreshold = parseOptionalNumber(root.dataset.voiceBargeInSpeechThreshold) ?? 0.04;
   const syncElement = requireElement(document, root.dataset.voiceSync, HTMLElement, "voice-htmx-sync");
   const connectionMetric = requireElement(root, root.dataset.voiceConnection, HTMLElement, "metric-connection");
   const errorStatus = requireElement(root, root.dataset.voiceError, HTMLElement, "status-error");
@@ -1279,9 +1370,52 @@ var initVoiceHTMXRoot = (root) => {
   const voiceMonitorCopy = requireElement(root, root.dataset.voiceMonitorCopy, HTMLElement, "voice-monitor-copy");
   const voiceWaveGlow = requireElement(root, root.dataset.voiceWaveGlow, SVGPathElement, "voice-wave-glow");
   const voiceWavePath = requireElement(root, root.dataset.voiceWavePath, SVGPathElement, "voice-wave-path");
+  let activeMode = null;
+  let hasStartedModes = {
+    general: false,
+    guided: false
+  };
+  let isCapturing = false;
+  let micError = null;
+  let waveLevels = createInitialVoiceWaveLevels();
+  let lastInputLevel = 0;
+  let lastAssistantAt = 0;
+  let lastAssistantAudioCount = 0;
+  let lastAssistantTextCount = 0;
+  const syncBargeInOutput = () => {
+    if (!bargeInMonitor) {
+      return;
+    }
+    const voice = currentVoice();
+    const audioCount = voice.assistantAudio.length;
+    const textCount = voice.assistantTexts.length;
+    if (audioCount > lastAssistantAudioCount || textCount > lastAssistantTextCount) {
+      lastAssistantAt = Date.now();
+    }
+    lastAssistantAudioCount = audioCount;
+    lastAssistantTextCount = textCount;
+  };
+  const sendAudioWithBargeInEvidence = (audio, sendAudio) => {
+    syncBargeInOutput();
+    if (bargeInMonitor && Date.now() - lastAssistantAt <= bargeInRecentWindowMs && lastInputLevel >= bargeInSpeechThreshold) {
+      bargeInMonitor.recordRequested({
+        reason: "manual-audio",
+        sessionId: currentVoice().sessionId
+      });
+      bargeInMonitor.recordStopped({
+        latencyMs: 0,
+        playbackStopLatencyMs: 0,
+        reason: "manual-audio",
+        sessionId: currentVoice().sessionId
+      });
+    }
+    sendAudio(audio);
+  };
   const guidedVoice = createVoiceController(guidedPath, {
     capture: {
+      onAudio: sendAudioWithBargeInEvidence,
       onLevel: (level) => {
+        lastInputLevel = level;
         waveLevels = pushVoiceWaveLevel(waveLevels, level);
         renderWave();
       }
@@ -1290,7 +1424,9 @@ var initVoiceHTMXRoot = (root) => {
   });
   const generalVoice = createVoiceController(generalPath, {
     capture: {
+      onAudio: sendAudioWithBargeInEvidence,
       onLevel: (level) => {
+        lastInputLevel = level;
         waveLevels = pushVoiceWaveLevel(waveLevels, level);
         renderWave();
       }
@@ -1299,14 +1435,6 @@ var initVoiceHTMXRoot = (root) => {
   });
   const stopGuidedBinding = guidedVoice.bindHTMX({ element: syncElement });
   const stopGeneralBinding = generalVoice.bindHTMX({ element: syncElement });
-  let activeMode = null;
-  let hasStartedModes = {
-    general: false,
-    guided: false
-  };
-  let isCapturing = false;
-  let micError = null;
-  let waveLevels = createInitialVoiceWaveLevels();
   const currentVoice = () => activeMode === "general" ? generalVoice : guidedVoice;
   const renderWave = () => {
     const path = createVoiceWavePath(waveLevels);
@@ -1385,8 +1513,14 @@ var initVoiceHTMXRoot = (root) => {
       render();
     }
   };
-  guidedVoice.subscribe(render);
-  generalVoice.subscribe(render);
+  guidedVoice.subscribe(() => {
+    syncBargeInOutput();
+    render();
+  });
+  generalVoice.subscribe(() => {
+    syncBargeInOutput();
+    render();
+  });
   startGuidedButton.addEventListener("click", () => {
     startMode("guided");
   });

package/dist/client/index.js

@@ -1446,7 +1446,13 @@ var createVoiceController = (path, options = {}) => {
     capture = createMicrophoneCapture({
       channelCount: options.capture?.channelCount ?? preset.capture.channelCount,
       onLevel: options.capture?.onLevel,
-      onAudio: (audio) => stream.sendAudio(audio),
+      onAudio: (audio) => {
+        if (options.capture?.onAudio) {
+          options.capture.onAudio(audio, stream.sendAudio);
+          return;
+        }
+        stream.sendAudio(audio);
+      },
       sampleRateHz: options.capture?.sampleRateHz ?? preset.capture.sampleRateHz
     });
     return capture;

package/dist/react/index.js

@@ -3116,7 +3116,13 @@ var createVoiceController = (path, options = {}) => {
     capture = createMicrophoneCapture({
       channelCount: options.capture?.channelCount ?? preset.capture.channelCount,
       onLevel: options.capture?.onLevel,
-      onAudio: (audio) => stream.sendAudio(audio),
+      onAudio: (audio) => {
+        if (options.capture?.onAudio) {
+          options.capture.onAudio(audio, stream.sendAudio);
+          return;
+        }
+        stream.sendAudio(audio);
+      },
       sampleRateHz: options.capture?.sampleRateHz ?? preset.capture.sampleRateHz
     });
     return capture;

package/dist/svelte/index.js

@@ -2558,7 +2558,13 @@ var createVoiceController = (path, options = {}) => {
     capture = createMicrophoneCapture({
       channelCount: options.capture?.channelCount ?? preset.capture.channelCount,
       onLevel: options.capture?.onLevel,
-      onAudio: (audio) => stream.sendAudio(audio),
+      onAudio: (audio) => {
+        if (options.capture?.onAudio) {
+          options.capture.onAudio(audio, stream.sendAudio);
+          return;
+        }
+        stream.sendAudio(audio);
+      },
       sampleRateHz: options.capture?.sampleRateHz ?? preset.capture.sampleRateHz
     });
     return capture;

package/dist/testing/index.js

@@ -2948,7 +2948,13 @@ var createVoiceController = (path, options = {}) => {
     capture = createMicrophoneCapture({
       channelCount: options.capture?.channelCount ?? preset.capture.channelCount,
       onLevel: options.capture?.onLevel,
-      onAudio: (audio) => stream.sendAudio(audio),
+      onAudio: (audio) => {
+        if (options.capture?.onAudio) {
+          options.capture.onAudio(audio, stream.sendAudio);
+          return;
+        }
+        stream.sendAudio(audio);
+      },
       sampleRateHz: options.capture?.sampleRateHz ?? preset.capture.sampleRateHz
     });
     return capture;

package/dist/types.d.ts

@@ -734,6 +734,7 @@ export type VoiceConnectionOptions = {
 };
 export type VoiceCaptureOptions = {
     channelCount?: 1 | 2;
+    onAudio?: (audio: Uint8Array | ArrayBuffer, sendAudio: (audio: Uint8Array | ArrayBuffer) => void) => void;
     onLevel?: (level: number) => void;
     sampleRateHz?: number;
 };

package/dist/vue/index.js

@@ -2891,7 +2891,13 @@ var createVoiceController = (path, options = {}) => {
     capture = createMicrophoneCapture({
       channelCount: options.capture?.channelCount ?? preset.capture.channelCount,
       onLevel: options.capture?.onLevel,
-      onAudio: (audio) => stream.sendAudio(audio),
+      onAudio: (audio) => {
+        if (options.capture?.onAudio) {
+          options.capture.onAudio(audio, stream.sendAudio);
+          return;
+        }
+        stream.sendAudio(audio);
+      },
       sampleRateHz: options.capture?.sampleRateHz ?? preset.capture.sampleRateHz
     });
     return capture;

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@absolutejs/voice",
-	"version": "0.0.22-beta.92",
+	"version": "0.0.22-beta.93",
 	"description": "Voice primitives and Elysia plugin for AbsoluteJS",
 	"repository": {
 		"type": "git",