@absolutejs/voice 0.0.22-beta.673 → 0.0.22-beta.674

package/dist/client/htmxBootstrap.js

@@ -2135,6 +2135,49 @@ var createVoiceAudioPlayer = (source, options = {}) => {
   let maxGapMs = 0;
   let errorCount = 0;
   let hasScheduledAnyChunk = false;
+  let buffersScheduled = 0;
+  let lastScheduledTurnId;
+  const audioFormats = new Set;
+  let lastAudioFormat = null;
+  let formatChangeCount = 0;
+  let incompleteFrameChunkCount = 0;
+  let discardedPartialFrameBytes = 0;
+  let pendingFrameBytes = new Uint8Array;
+  let pendingFrameFormat = null;
+  let pendingFrameTurnId;
+  const formatKey = (format) => `${format.container}/${format.encoding}/${String(format.channels)}ch/${String(format.sampleRateHz)}hz`;
+  const frameAlignedChunk = (chunk) => {
+    const format = formatKey(chunk.format);
+    audioFormats.add(format);
+    if (lastAudioFormat !== null && lastAudioFormat !== format) {
+      formatChangeCount += 1;
+    }
+    lastAudioFormat = format;
+    if (pendingFrameBytes.byteLength > 0 && (pendingFrameFormat !== format || pendingFrameTurnId !== chunk.turnId)) {
+      discardedPartialFrameBytes += pendingFrameBytes.byteLength;
+      pendingFrameBytes = new Uint8Array;
+    }
+    const joined = pendingFrameBytes.byteLength === 0 ? chunk.chunk : (() => {
+      const bytes = new Uint8Array(pendingFrameBytes.byteLength + chunk.chunk.byteLength);
+      bytes.set(pendingFrameBytes);
+      bytes.set(chunk.chunk, pendingFrameBytes.byteLength);
+      return bytes;
+    })();
+    const frameBytes = Math.max(1, chunk.format.channels) * 2;
+    const completeBytes = joined.byteLength - joined.byteLength % frameBytes;
+    const remainderBytes = joined.byteLength - completeBytes;
+    if (remainderBytes > 0) {
+      incompleteFrameChunkCount += 1;
+      pendingFrameBytes = joined.slice(completeBytes);
+      pendingFrameFormat = format;
+      pendingFrameTurnId = chunk.turnId;
+    } else {
+      pendingFrameBytes = new Uint8Array;
+      pendingFrameFormat = null;
+      pendingFrameTurnId = undefined;
+    }
+    return completeBytes > 0 ? { ...chunk, chunk: joined.subarray(0, completeBytes) } : null;
+  };
   const sampleConcurrency = () => {
     if (activeAudioPlayers.size > observedPeakConcurrency) {
       observedPeakConcurrency = activeAudioPlayers.size;
@@ -2254,7 +2297,7 @@ var createVoiceAudioPlayer = (source, options = {}) => {
     queueEndTime = audioContext.currentTime;
     return audioContext;
   };
-  const scheduleBuffer = (context, buffer, rate) => {
+  const scheduleBuffer = (context, buffer, rate, turnId) => {
     const node = context.createBufferSource();
     node.buffer = buffer;
     if (node.playbackRate) {
@@ -2272,13 +2315,15 @@ var createVoiceAudioPlayer = (source, options = {}) => {
     };
     const earliestStart = context.currentTime + lookaheadSeconds;
     const startAt = Math.max(earliestStart, queueEndTime);
-    if (hasScheduledAnyChunk && earliestStart > queueEndTime + GAP_EPSILON_S) {
+    if (hasScheduledAnyChunk && turnId !== undefined && turnId === lastScheduledTurnId && earliestStart > queueEndTime + GAP_EPSILON_S) {
       const gapMs = (earliestStart - queueEndTime) * 1000;
       gapCount += 1;
       totalGapMs += gapMs;
       maxGapMs = Math.max(maxGapMs, gapMs);
     }
     hasScheduledAnyChunk = true;
+    lastScheduledTurnId = turnId;
+    buffersScheduled += 1;
     scheduledDurationMs += buffer.duration / rate * 1000;
     sampleConcurrency();
     queueEndTime = startAt + buffer.duration / rate;
@@ -2291,10 +2336,13 @@ var createVoiceAudioPlayer = (source, options = {}) => {
   };
   const scheduleChunk = async (chunk) => {
     const context = await ensureAudioContext();
-    const buffer = decodePCM16LEChunk(context, chunk);
+    const alignedChunk = frameAlignedChunk(chunk);
+    if (!alignedChunk)
+      return;
+    const buffer = decodePCM16LEChunk(context, alignedChunk);
     if (Math.abs(playbackRate - 1) <= STRETCH_BYPASS_EPSILON) {
       stretcher?.reset();
-      scheduleBuffer(context, buffer, playbackRate);
+      scheduleBuffer(context, buffer, playbackRate, chunk.turnId);
       return;
     }
     const channels = Math.max(1, chunk.format.channels);
@@ -2315,13 +2363,14 @@ var createVoiceAudioPlayer = (source, options = {}) => {
         continue;
       outBuffer.getChannelData(channelIndex).set(channelOut);
     }
-    scheduleBuffer(context, outBuffer, 1);
+    scheduleBuffer(context, outBuffer, 1, chunk.turnId);
   };
   const stopQueuedPlayback = (options2) => {
     for (const node of [...sourceNodes]) {
       node.stop?.();
     }
     queueEndTime = audioContext ? audioContext.currentTime : 0;
+    lastScheduledTurnId = undefined;
     if (options2?.forceClear) {
       for (const node of sourceNodes) {
         node.disconnect?.();
@@ -2416,13 +2465,19 @@ var createVoiceAudioPlayer = (source, options = {}) => {
       const chunksReceived = source.assistantAudio.length;
       const chunksScheduled = state.processedChunkCount;
       return {
+        audioFormatCount: audioFormats.size,
+        buffersScheduled,
         chunksReceived,
         chunksScheduled,
+        discardedPartialFrameBytes,
         errorCount,
+        formatChangeCount,
         gapCount,
+        incompleteFrameChunkCount,
         maxConcurrentPlayers: observedPeakConcurrency,
         maxGapMs: Math.round(maxGapMs),
-        ok: observedPeakConcurrency <= 1 && errorCount === 0 && chunksScheduled >= chunksReceived,
+        pendingPartialFrameBytes: pendingFrameBytes.byteLength,
+        ok: observedPeakConcurrency <= 1 && errorCount === 0 && chunksScheduled >= chunksReceived && discardedPartialFrameBytes === 0 && pendingFrameBytes.byteLength === 0,
         scheduledDurationMs: Math.round(scheduledDurationMs),
         totalGapMs: Math.round(totalGapMs)
       };

package/dist/client/index.js

@@ -886,6 +886,49 @@ var createVoiceAudioPlayer = (source, options = {}) => {
   let maxGapMs = 0;
   let errorCount = 0;
   let hasScheduledAnyChunk = false;
+  let buffersScheduled = 0;
+  let lastScheduledTurnId;
+  const audioFormats = new Set;
+  let lastAudioFormat = null;
+  let formatChangeCount = 0;
+  let incompleteFrameChunkCount = 0;
+  let discardedPartialFrameBytes = 0;
+  let pendingFrameBytes = new Uint8Array;
+  let pendingFrameFormat = null;
+  let pendingFrameTurnId;
+  const formatKey = (format) => `${format.container}/${format.encoding}/${String(format.channels)}ch/${String(format.sampleRateHz)}hz`;
+  const frameAlignedChunk = (chunk) => {
+    const format = formatKey(chunk.format);
+    audioFormats.add(format);
+    if (lastAudioFormat !== null && lastAudioFormat !== format) {
+      formatChangeCount += 1;
+    }
+    lastAudioFormat = format;
+    if (pendingFrameBytes.byteLength > 0 && (pendingFrameFormat !== format || pendingFrameTurnId !== chunk.turnId)) {
+      discardedPartialFrameBytes += pendingFrameBytes.byteLength;
+      pendingFrameBytes = new Uint8Array;
+    }
+    const joined = pendingFrameBytes.byteLength === 0 ? chunk.chunk : (() => {
+      const bytes = new Uint8Array(pendingFrameBytes.byteLength + chunk.chunk.byteLength);
+      bytes.set(pendingFrameBytes);
+      bytes.set(chunk.chunk, pendingFrameBytes.byteLength);
+      return bytes;
+    })();
+    const frameBytes = Math.max(1, chunk.format.channels) * 2;
+    const completeBytes = joined.byteLength - joined.byteLength % frameBytes;
+    const remainderBytes = joined.byteLength - completeBytes;
+    if (remainderBytes > 0) {
+      incompleteFrameChunkCount += 1;
+      pendingFrameBytes = joined.slice(completeBytes);
+      pendingFrameFormat = format;
+      pendingFrameTurnId = chunk.turnId;
+    } else {
+      pendingFrameBytes = new Uint8Array;
+      pendingFrameFormat = null;
+      pendingFrameTurnId = undefined;
+    }
+    return completeBytes > 0 ? { ...chunk, chunk: joined.subarray(0, completeBytes) } : null;
+  };
   const sampleConcurrency = () => {
     if (activeAudioPlayers.size > observedPeakConcurrency) {
       observedPeakConcurrency = activeAudioPlayers.size;
@@ -1005,7 +1048,7 @@ var createVoiceAudioPlayer = (source, options = {}) => {
     queueEndTime = audioContext.currentTime;
     return audioContext;
   };
-  const scheduleBuffer = (context, buffer, rate) => {
+  const scheduleBuffer = (context, buffer, rate, turnId) => {
     const node = context.createBufferSource();
     node.buffer = buffer;
     if (node.playbackRate) {
@@ -1023,13 +1066,15 @@ var createVoiceAudioPlayer = (source, options = {}) => {
     };
     const earliestStart = context.currentTime + lookaheadSeconds;
     const startAt = Math.max(earliestStart, queueEndTime);
-    if (hasScheduledAnyChunk && earliestStart > queueEndTime + GAP_EPSILON_S) {
+    if (hasScheduledAnyChunk && turnId !== undefined && turnId === lastScheduledTurnId && earliestStart > queueEndTime + GAP_EPSILON_S) {
       const gapMs = (earliestStart - queueEndTime) * 1000;
       gapCount += 1;
       totalGapMs += gapMs;
       maxGapMs = Math.max(maxGapMs, gapMs);
     }
     hasScheduledAnyChunk = true;
+    lastScheduledTurnId = turnId;
+    buffersScheduled += 1;
     scheduledDurationMs += buffer.duration / rate * 1000;
     sampleConcurrency();
     queueEndTime = startAt + buffer.duration / rate;
@@ -1042,10 +1087,13 @@ var createVoiceAudioPlayer = (source, options = {}) => {
   };
   const scheduleChunk = async (chunk) => {
     const context = await ensureAudioContext();
-    const buffer = decodePCM16LEChunk(context, chunk);
+    const alignedChunk = frameAlignedChunk(chunk);
+    if (!alignedChunk)
+      return;
+    const buffer = decodePCM16LEChunk(context, alignedChunk);
     if (Math.abs(playbackRate - 1) <= STRETCH_BYPASS_EPSILON) {
       stretcher?.reset();
-      scheduleBuffer(context, buffer, playbackRate);
+      scheduleBuffer(context, buffer, playbackRate, chunk.turnId);
       return;
     }
     const channels = Math.max(1, chunk.format.channels);
@@ -1066,13 +1114,14 @@ var createVoiceAudioPlayer = (source, options = {}) => {
         continue;
       outBuffer.getChannelData(channelIndex).set(channelOut);
     }
-    scheduleBuffer(context, outBuffer, 1);
+    scheduleBuffer(context, outBuffer, 1, chunk.turnId);
   };
   const stopQueuedPlayback = (options2) => {
     for (const node of [...sourceNodes]) {
       node.stop?.();
     }
     queueEndTime = audioContext ? audioContext.currentTime : 0;
+    lastScheduledTurnId = undefined;
     if (options2?.forceClear) {
       for (const node of sourceNodes) {
         node.disconnect?.();
@@ -1167,13 +1216,19 @@ var createVoiceAudioPlayer = (source, options = {}) => {
       const chunksReceived = source.assistantAudio.length;
       const chunksScheduled = state.processedChunkCount;
       return {
+        audioFormatCount: audioFormats.size,
+        buffersScheduled,
         chunksReceived,
         chunksScheduled,
+        discardedPartialFrameBytes,
         errorCount,
+        formatChangeCount,
         gapCount,
+        incompleteFrameChunkCount,
         maxConcurrentPlayers: observedPeakConcurrency,
         maxGapMs: Math.round(maxGapMs),
-        ok: observedPeakConcurrency <= 1 && errorCount === 0 && chunksScheduled >= chunksReceived,
+        pendingPartialFrameBytes: pendingFrameBytes.byteLength,
+        ok: observedPeakConcurrency <= 1 && errorCount === 0 && chunksScheduled >= chunksReceived && discardedPartialFrameBytes === 0 && pendingFrameBytes.byteLength === 0,
         scheduledDurationMs: Math.round(scheduledDurationMs),
         totalGapMs: Math.round(totalGapMs)
       };

package/dist/core/types.d.ts

@@ -1624,17 +1624,23 @@ export type VoiceAudioPlayerSource = {
     assistantAudio: VoiceStreamState["assistantAudio"];
     subscribe: (subscriber: () => void) => () => void;
 };
-/** Post-call playback integrity, for verifying the assistant audio actually
- *  played cleanly (no overlap, no dropped/stalled packets) — emit it as a trace
- *  so a garbled call is provable, not guessed. `ok` is the all-clear roll-up. */
+/** Post-call playback-pipeline integrity. This proves what the browser player
+ *  received and scheduled; it does not claim that the resulting waveform
+ *  sounded clean. A browser-side replay is required for that stronger claim. */
 export type VoiceAudioIntegrity = {
     ok: boolean;
     chunksReceived: number;
     chunksScheduled: number;
+    buffersScheduled: number;
     scheduledDurationMs: number;
     gapCount: number;
     totalGapMs: number;
     maxGapMs: number;
+    audioFormatCount: number;
+    formatChangeCount: number;
+    incompleteFrameChunkCount: number;
+    discardedPartialFrameBytes: number;
+    pendingPartialFrameBytes: number;
     maxConcurrentPlayers: number;
     errorCount: number;
 };

package/dist/testing/index.js

@@ -2228,6 +2228,49 @@ var createVoiceAudioPlayer = (source, options = {}) => {
   let maxGapMs = 0;
   let errorCount = 0;
   let hasScheduledAnyChunk = false;
+  let buffersScheduled = 0;
+  let lastScheduledTurnId;
+  const audioFormats = new Set;
+  let lastAudioFormat = null;
+  let formatChangeCount = 0;
+  let incompleteFrameChunkCount = 0;
+  let discardedPartialFrameBytes = 0;
+  let pendingFrameBytes = new Uint8Array;
+  let pendingFrameFormat = null;
+  let pendingFrameTurnId;
+  const formatKey = (format) => `${format.container}/${format.encoding}/${String(format.channels)}ch/${String(format.sampleRateHz)}hz`;
+  const frameAlignedChunk = (chunk) => {
+    const format = formatKey(chunk.format);
+    audioFormats.add(format);
+    if (lastAudioFormat !== null && lastAudioFormat !== format) {
+      formatChangeCount += 1;
+    }
+    lastAudioFormat = format;
+    if (pendingFrameBytes.byteLength > 0 && (pendingFrameFormat !== format || pendingFrameTurnId !== chunk.turnId)) {
+      discardedPartialFrameBytes += pendingFrameBytes.byteLength;
+      pendingFrameBytes = new Uint8Array;
+    }
+    const joined = pendingFrameBytes.byteLength === 0 ? chunk.chunk : (() => {
+      const bytes2 = new Uint8Array(pendingFrameBytes.byteLength + chunk.chunk.byteLength);
+      bytes2.set(pendingFrameBytes);
+      bytes2.set(chunk.chunk, pendingFrameBytes.byteLength);
+      return bytes2;
+    })();
+    const frameBytes = Math.max(1, chunk.format.channels) * 2;
+    const completeBytes = joined.byteLength - joined.byteLength % frameBytes;
+    const remainderBytes = joined.byteLength - completeBytes;
+    if (remainderBytes > 0) {
+      incompleteFrameChunkCount += 1;
+      pendingFrameBytes = joined.slice(completeBytes);
+      pendingFrameFormat = format;
+      pendingFrameTurnId = chunk.turnId;
+    } else {
+      pendingFrameBytes = new Uint8Array;
+      pendingFrameFormat = null;
+      pendingFrameTurnId = undefined;
+    }
+    return completeBytes > 0 ? { ...chunk, chunk: joined.subarray(0, completeBytes) } : null;
+  };
   const sampleConcurrency = () => {
     if (activeAudioPlayers.size > observedPeakConcurrency) {
       observedPeakConcurrency = activeAudioPlayers.size;
@@ -2347,7 +2390,7 @@ var createVoiceAudioPlayer = (source, options = {}) => {
     queueEndTime = audioContext.currentTime;
     return audioContext;
   };
-  const scheduleBuffer = (context, buffer, rate) => {
+  const scheduleBuffer = (context, buffer, rate, turnId) => {
     const node = context.createBufferSource();
     node.buffer = buffer;
     if (node.playbackRate) {
@@ -2365,13 +2408,15 @@ var createVoiceAudioPlayer = (source, options = {}) => {
     };
     const earliestStart = context.currentTime + lookaheadSeconds;
     const startAt = Math.max(earliestStart, queueEndTime);
-    if (hasScheduledAnyChunk && earliestStart > queueEndTime + GAP_EPSILON_S) {
+    if (hasScheduledAnyChunk && turnId !== undefined && turnId === lastScheduledTurnId && earliestStart > queueEndTime + GAP_EPSILON_S) {
       const gapMs = (earliestStart - queueEndTime) * 1000;
       gapCount += 1;
       totalGapMs += gapMs;
       maxGapMs = Math.max(maxGapMs, gapMs);
     }
     hasScheduledAnyChunk = true;
+    lastScheduledTurnId = turnId;
+    buffersScheduled += 1;
     scheduledDurationMs += buffer.duration / rate * 1000;
     sampleConcurrency();
     queueEndTime = startAt + buffer.duration / rate;
@@ -2384,10 +2429,13 @@ var createVoiceAudioPlayer = (source, options = {}) => {
   };
   const scheduleChunk = async (chunk) => {
     const context = await ensureAudioContext();
-    const buffer = decodePCM16LEChunk(context, chunk);
+    const alignedChunk = frameAlignedChunk(chunk);
+    if (!alignedChunk)
+      return;
+    const buffer = decodePCM16LEChunk(context, alignedChunk);
     if (Math.abs(playbackRate - 1) <= STRETCH_BYPASS_EPSILON) {
       stretcher?.reset();
-      scheduleBuffer(context, buffer, playbackRate);
+      scheduleBuffer(context, buffer, playbackRate, chunk.turnId);
       return;
     }
     const channels = Math.max(1, chunk.format.channels);
@@ -2408,13 +2456,14 @@ var createVoiceAudioPlayer = (source, options = {}) => {
         continue;
       outBuffer.getChannelData(channelIndex).set(channelOut);
     }
-    scheduleBuffer(context, outBuffer, 1);
+    scheduleBuffer(context, outBuffer, 1, chunk.turnId);
   };
   const stopQueuedPlayback = (options2) => {
     for (const node of [...sourceNodes]) {
       node.stop?.();
     }
     queueEndTime = audioContext ? audioContext.currentTime : 0;
+    lastScheduledTurnId = undefined;
     if (options2?.forceClear) {
       for (const node of sourceNodes) {
         node.disconnect?.();
@@ -2509,13 +2558,19 @@ var createVoiceAudioPlayer = (source, options = {}) => {
       const chunksReceived = source.assistantAudio.length;
       const chunksScheduled = state.processedChunkCount;
       return {
+        audioFormatCount: audioFormats.size,
+        buffersScheduled,
         chunksReceived,
         chunksScheduled,
+        discardedPartialFrameBytes,
         errorCount,
+        formatChangeCount,
         gapCount,
+        incompleteFrameChunkCount,
         maxConcurrentPlayers: observedPeakConcurrency,
         maxGapMs: Math.round(maxGapMs),
-        ok: observedPeakConcurrency <= 1 && errorCount === 0 && chunksScheduled >= chunksReceived,
+        pendingPartialFrameBytes: pendingFrameBytes.byteLength,
+        ok: observedPeakConcurrency <= 1 && errorCount === 0 && chunksScheduled >= chunksReceived && discardedPartialFrameBytes === 0 && pendingFrameBytes.byteLength === 0,
         scheduledDurationMs: Math.round(scheduledDurationMs),
         totalGapMs: Math.round(totalGapMs)
       };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@absolutejs/voice",
-  "version": "0.0.22-beta.673",
+  "version": "0.0.22-beta.674",
   "description": "Voice primitives and Elysia plugin for AbsoluteJS",
   "repository": {
     "type": "git",