[googlestt] Fix drop bytes (openhab#14649)

* [googlestt] Fix drop bytes * fix unhandled cancelation error when using single utterance mode Signed-off-by: Miguel Álvarez <[email protected]>
renescherer · Mar 23, 2023 · 4d0e924 · 4d0e924
1 parent c38e42d
commit 4d0e924
Showing 1 changed file with 22 additions and 12 deletions.
diff --git a/....voice.googlestt/src/main/java/org/openhab/voice/googlestt/internal/GoogleSTTService.java b/....voice.googlestt/src/main/java/org/openhab/voice/googlestt/internal/GoogleSTTService.java
@@ -255,14 +255,20 @@ private void streamAudio(ClientStream<StreamingRecognizeRequest> clientStream, A
         long startTime = System.currentTimeMillis();
         long maxTranscriptionMillis = (config.maxTranscriptionSeconds * 1000L);
         long maxSilenceMillis = (config.maxSilenceSeconds * 1000L);
-        int readBytes = 6400;
-        while (!aborted.get()) {
-            byte[] data = new byte[readBytes];
-            int dataN = audioStream.read(data);
+        final int bufferSize = 6400;
+        int numBytesRead;
+        int remaining = bufferSize;
+        byte[] audioBuffer = new byte[bufferSize];
+        while (!aborted.get() && !responseObserver.isDone()) {
+            numBytesRead = audioStream.read(audioBuffer, bufferSize - remaining, remaining);
             if (aborted.get()) {
                 logger.debug("Stops listening, aborted");
                 break;
             }
+            if (numBytesRead == -1) {
+                logger.debug("End of stream");
+                break;
+            }
             if (isExpiredInterval(maxTranscriptionMillis, startTime)) {
                 logger.debug("Stops listening, max transcription time reached");
                 break;
@@ -272,18 +278,17 @@ && isExpiredInterval(maxSilenceMillis, responseObserver.getLastInputTime())) {
                 logger.debug("Stops listening, max silence time reached");
                 break;
             }
-            if (dataN != readBytes) {
-                try {
-                    Thread.sleep(100);
-                } catch (InterruptedException e) {
-                }
+            if (numBytesRead != remaining) {
+                remaining = remaining - numBytesRead;
                 continue;
             }
+            remaining = bufferSize;
             StreamingRecognizeRequest dataRequest = StreamingRecognizeRequest.newBuilder()
-                    .setAudioContent(ByteString.copyFrom(data)).build();
-            logger.debug("Sending audio data {}", dataN);
+                    .setAudioContent(ByteString.copyFrom(audioBuffer)).build();
+            logger.debug("Sending audio data {}", bufferSize);
             clientStream.send(dataRequest);
         }
+        audioStream.close();
     }
 
     private void sendStreamConfig(ClientStream<StreamingRecognizeRequest> clientStream,
@@ -335,6 +340,7 @@ private static class TranscriptionListener implements ResponseObserver<Streaming
         private float confidenceSum = 0;
         private int responseCount = 0;
         private long lastInputTime = 0;
+        private boolean done = false;
 
         public TranscriptionListener(STTListener sttListener, GoogleSTTConfiguration config, AtomicBoolean aborted) {
             this.sttListener = sttListener;
@@ -374,7 +380,7 @@ public void onResponse(StreamingRecognizeResponse response) {
                     responseCount++;
                     // when in single utterance mode we can just get one final result so complete
                     if (config.singleUtteranceMode) {
-                        onComplete();
+                        done = true;
                     }
                 }
             });
@@ -411,6 +417,10 @@ public void onError(@Nullable Throwable t) {
             }
         }
 
+        public boolean isDone() {
+            return done;
+        }
+
         public long getLastInputTime() {
             return lastInputTime;
         }