sera-ai 1.0.10 → 1.0.12

package/dist/index.js

@@ -2670,7 +2670,6 @@ var createAudioProcessorWorker = () => {
         this._recordingStartTime = Date.now();
         this._initialSilenceThreshold = 44100 * 10;
         this._isInitialPhase = true;
-        this._audioActivityBuffer = []; // Track recent audio activity
         this._bufferSize = 0; // Track total samples in buffer
 
         this.port.onmessage = (event) => {
@@ -2765,12 +2764,6 @@ var createAudioProcessorWorker = () => {
             this._isInitialPhase = false;
             this._silentSampleCount = 0;
             this._lastAudioTime = Date.now();
-            
-            // Track audio activity for chunk validation
-            this._audioActivityBuffer.push(audioLevel);
-            if (this._audioActivityBuffer.length > 1000) {
-              this._audioActivityBuffer.shift(); // Keep only recent activity
-            }
           } else {
             this._silentSampleCount += samples.length;
             
@@ -2795,43 +2788,36 @@ var createAudioProcessorWorker = () => {
 
           if (this._uploadChunk && !this._uploadingChunk) {
             this._uploadingChunk = true;
-            
-            // Check if buffer has meaningful audio content before uploading
-            const recentActivity = this._audioActivityBuffer.slice(-100); // Last 100 activity measurements
-            const hasRecentAudio = recentActivity.some(level => level > this._audioThreshold * 2);
-            
+
             // Properly flatten the buffer by concatenating Float32Arrays
             let totalLength = 0;
             for (let i = 0; i < this._buffer.length; i++) {
               totalLength += this._buffer[i].length;
             }
-            
+
             const flat = new Float32Array(totalLength);
             let offset = 0;
             for (let i = 0; i < this._buffer.length; i++) {
               flat.set(this._buffer[i], offset);
               offset += this._buffer[i].length;
             }
-            
-            // Only upload if we have recent audio activity or if this is a significant buffer
-            if (hasRecentAudio || this._bufferSize > 44100 * 30) { // 30 seconds minimum
+
+            // Always send chunks to server - let server handle silence filtering
+            if (this._bufferSize > 0) {
               this.port.postMessage(
                 {
                   command: "chunk",
                   audioBuffer: flat.buffer,
-                  hasActivity: hasRecentAudio,
                   bufferDuration: this._bufferSize / 44100
                 },
                 [flat.buffer]
               );
-              // Clear buffer after successful upload
+              console.log('[UPLOAD] Sending chunk: ' + (this._bufferSize / 44100).toFixed(1) + 's');
+              // Clear buffer after upload
               this._buffer = [];
               this._bufferSize = 0;
-            } else {
-              console.log('[SKIP] Skipping silent chunk upload (' + (this._bufferSize / 44100).toFixed(1) + 's, no recent activity)');
-              // Don't clear buffer for skipped chunks, keep accumulating
             }
-            
+
             this._uploadChunk = false;
             this._uploadingChunk = false;
           }
@@ -3119,37 +3105,19 @@ var useAudioRecorder = ({
             console.log(
               `[PROCESSING] Processing audio chunk: ${audioData.length} samples (${(audioData.length / 44100).toFixed(2)}s)`
             );
-            let hasAudio = false;
             let maxAmplitude = 0;
             let nonZeroSamples = 0;
             for (let i = 0; i < audioData.length; i++) {
               const amplitude = Math.abs(audioData[i]);
               if (amplitude > maxAmplitude) maxAmplitude = amplitude;
               if (amplitude > 1e-3) {
-                hasAudio = true;
                 nonZeroSamples++;
               }
             }
             const audioPercentage = nonZeroSamples / audioData.length;
             console.log(
-              `[AUDIO] Audio validation: hasAudio=${hasAudio}, maxAmplitude=${maxAmplitude.toFixed(4)}, audioContent=${(audioPercentage * 100).toFixed(2)}%`
+              `[AUDIO] Audio stats: maxAmplitude=${maxAmplitude.toFixed(4)}, audioContent=${(audioPercentage * 100).toFixed(2)}%, sequence=${sequence}, isFinal=${isFinalChunk}`
             );
-            if (!hasAudio || maxAmplitude < 1e-3 || audioPercentage < 0.01) {
-              console.log(
-                `[SKIP] Skipping silent chunk (${audioPercentage < 0.01 ? "too little audio content" : "completely silent"})`
-              );
-              if (!isFinalChunk) {
-                console.log(`[OUT] Non-final silent chunk skipped entirely`);
-                return;
-              } else {
-                console.log(`[OUT] Final silent chunk - will send minimal audio to close session`);
-                const minimalAudio = new Float32Array(1e3);
-                for (let i = 0; i < minimalAudio.length; i += 100) {
-                  minimalAudio[i] = 1e-3;
-                }
-                audioData = minimalAudio;
-              }
-            }
             const sampleRate = audioContextRef.current?.sampleRate || 16e3;
             const timestamp = Date.now();
             const fileName = `audio-chunk-${timestamp}.wav`;
@@ -3186,11 +3154,8 @@ var useAudioRecorder = ({
               `[OUT] Final file for transcription: ${wavFile.size} bytes, ${wavFile.name}`
             );
             if (wavFile.size < 500) {
-              console.error(
-                `[ERROR] File too small (${wavFile.size} bytes) - likely contains no audio data`
-              );
-              throw new Error(
-                `Audio file too small (${wavFile.size} bytes) - no meaningful audio content`
+              console.warn(
+                `[WARN] Small audio file (${wavFile.size} bytes) - may contain minimal audio data, sending to server anyway`
               );
             }
             const requestData = {
@@ -5045,22 +5010,6 @@ var AudioDictation = ({
       startDictating();
     }
   };
-  React3__namespace.useEffect(() => {
-    const handleKeyDown = (e) => {
-      if (e.code === "Space" && !e.repeat && !isProcessing) {
-        e.preventDefault();
-        if (isDictating) {
-          stopDictating();
-        } else {
-          startDictating();
-        }
-      }
-    };
-    document.addEventListener("keydown", handleKeyDown);
-    return () => {
-      document.removeEventListener("keydown", handleKeyDown);
-    };
-  }, [isDictating, isProcessing, startDictating, stopDictating]);
   const getButtonContent = () => {
     if (isProcessing) {
       return /* @__PURE__ */ jsxRuntime.jsx(jsxRuntime.Fragment, { children: /* @__PURE__ */ jsxRuntime.jsx(lucideReact.Loader2, { className: "h-5 w-5 animate-spin" }) });