feat: Adds metrics for the amount of audio sent to each transcriber (#558)

Author: rpurdel

src/main/java/org/jitsi/jigasi/stats/Statistics.java

@@ -129,6 +129,26 @@ public class Statistics
      */
     public static final String TOTAL_TRANSCRIBER_G_MINUTES = "total_transcriber_g_minutes";
 
+    /**
+     * The total number of milliseconds submitted to Google API for transcription.
+     */
+    public static final String TOTAL_TRANSCRIBER_GGL_MILLIS = "total_transcriber_ggl_millis";
+
+    /**
+     * The total number of milliseconds submitted to Oracle API for transcription.
+     */
+    public static final String TOTAL_TRANSCRIBER_OCI_MILLIS = "total_transcriber_oci_millis";
+
+    /**
+     * The total number of milliseconds submitted to Skynet/Whisper for transcription.
+     */
+    public static final String TOTAL_TRANSCRIBER_WSP_MILLIS = "total_transcriber_wsp_millis";
+
+    /**
+     * The total number of milliseconds submitted to Vosk for transcription.
+     */
+    public static final String TOTAL_TRANSCRIBER_VSK_MILLIS = "total_transcriber_vsk_millis";
+
     /**
      * The total number of requests submitted to the Google Cloud Speech API.
      */
@@ -317,6 +337,34 @@ public class Statistics
             TOTAL_TRANSCRIBER_G_REQUESTS,
             "Total number of transcriber requests.");
 
+    /**
+     * The total number of milliseconds submitted to Google API for transcription.
+     */
+    private static LongGaugeMetric totalTranscriberGoogleMillis = JigasiMetricsContainer.INSTANCE.registerLongGauge(
+            TOTAL_TRANSCRIBER_GGL_MILLIS,
+            "Total number of milliseconds sent to Google's API.");
+
+    /**
+     * The total number of milliseconds submitted to Oracle Cloud API for transcription.
+     */
+    private static LongGaugeMetric totalTranscriberOracleMillis = JigasiMetricsContainer.INSTANCE.registerLongGauge(
+            TOTAL_TRANSCRIBER_OCI_MILLIS,
+            "Total number of milliseconds sent to OCI API.");
+
+    /**
+     * The total number of milliseconds submitted to Skynet Whisper for transcription.
+     */
+    private static LongGaugeMetric totalTranscriberWhisperMillis = JigasiMetricsContainer.INSTANCE.registerLongGauge(
+            TOTAL_TRANSCRIBER_WSP_MILLIS,
+            "Total number of milliseconds sent to Skynet Whisper.");
+
+    /**
+     * The total number of milliseconds submitted to Vosk for transcription.
+     */
+    private static LongGaugeMetric totalTranscriberVoskMillis = JigasiMetricsContainer.INSTANCE.registerLongGauge(
+            TOTAL_TRANSCRIBER_VSK_MILLIS,
+            "Total number of milliseconds sent to Vosk.");
+
     /**
      * Cumulative number of seconds of all conferences.
      */
@@ -407,6 +455,10 @@ public static synchronized void sendJSON(
 
         stats.put(TOTAL_TRANSCRIBER_G_REQUESTS, totalTrasnscriberRequests.get());
         stats.put(TOTAL_TRANSCRIBER_G_MINUTES, totalTranscriberMinutes.get());
+        stats.put(TOTAL_TRANSCRIBER_GGL_MILLIS, totalTranscriberGoogleMillis.get());
+        stats.put(TOTAL_TRANSCRIBER_OCI_MILLIS, totalTranscriberOracleMillis.get());
+        stats.put(TOTAL_TRANSCRIBER_WSP_MILLIS, totalTranscriberWhisperMillis.get());
+        stats.put(TOTAL_TRANSCRIBER_VSK_MILLIS, totalTranscriberVoskMillis.get());
 
         stats.put(TOTAL_TRANSCRIBER_STARTED, totalTrasnscriberStarted.get());
         stats.put(TOTAL_TRANSCRIBER_STOPPED, totalTrasnscriberStopped.get());
@@ -636,6 +688,38 @@ public static void incrementTotalTranscriberMinutes(long value)
         totalTranscriberMinutes.addAndGet(value);
     }
 
+    /**
+     * Increment the value of total number of milliseconds sent to Google API for transcription.
+     */
+    public static void incrementTotalTranscriberGoogleMillis(long value)
+    {
+        totalTranscriberGoogleMillis.addAndGet(value);
+    }
+
+    /**
+     * Increment the value of total number of milliseconds sent to Oracle API for transcription.
+     */
+    public static void incrementTotalTranscriberOracleMillis(long value)
+    {
+        totalTranscriberOracleMillis.addAndGet(value);
+    }
+
+    /**
+     * Increment the value of total number of milliseconds sent to Skynet Whisper for transcription.
+     */
+    public static void incrementTotalTranscriberWhisperMillis(long value)
+    {
+        totalTranscriberWhisperMillis.addAndGet(value);
+    }
+
+    /**
+     * Increment the value of total number of milliseconds sent to Vosk for transcription.
+     */
+    public static void incrementTotalTranscriberVoskMillis(long value)
+    {
+        totalTranscriberVoskMillis.addAndGet(value);
+    }
+
     /**
      * Increment the value of total number of transcriber request.
      */

src/main/java/org/jitsi/jigasi/transcription/Participant.java

@@ -23,6 +23,7 @@
 import org.jitsi.xmpp.extensions.jitsimeet.*;
 import org.jitsi.utils.logging.*;
 import org.jivesoftware.smack.packet.*;
+import org.jitsi.jigasi.stats.*;
 
 import javax.media.format.*;
 import java.nio.*;
@@ -152,6 +153,8 @@ public class Participant
      */
     private SilenceFilter silenceFilter = null;
 
+    private String transcriptionServiceName;
+
     /**
      * Create a participant with a given name and audio stream
      *
@@ -173,6 +176,7 @@ public class Participant
     {
         this.transcriber = transcriber;
         this.identifier = identifier;
+        this.transcriptionServiceName = transcriber.getTranscriptionService().getClass().getSimpleName();
 
         if (filterAudio)
         {
@@ -661,6 +665,35 @@ else if (silenceFilter.newSpeech())
            });
     }
 
+    private void incrementSentStats(int byteCount)
+    {
+        int divider = EXPECTED_AUDIO_LENGTH;
+
+        if (transcriptionServiceName.equals("WhisperTranscriptionService")
+                || transcriptionServiceName.equals("OracleTranscriptionService"))
+        {
+            // the byte count for each 20ms packet if the audio format is 16kHz mono
+            divider = 640;
+        }
+
+        long millis = byteCount / divider * 20;
+
+        switch (transcriptionServiceName) {
+            case "WhisperTranscriptionService":
+                Statistics.incrementTotalTranscriberWhisperMillis(millis);
+                break;
+            case "OracleTranscriptionService":
+                Statistics.incrementTotalTranscriberOracleMillis(millis);
+                break;
+            case "VoskTranscriptionService":
+                Statistics.incrementTotalTranscriberVoskMillis(millis);
+                break;
+            case "GoogleCloudTranscriptionService":
+                Statistics.incrementTotalTranscriberGoogleMillis(millis);
+                break;
+        }
+    }
+
     /**
      * Send the specified audio to the TranscriptionService.
      * <p>
@@ -681,6 +714,7 @@ private void sendRequest(byte[] audio)
             if (session != null && !session.ended())
             {
                 session.sendRequest(request);
+                incrementSentStats(audio.length);
             }
             else if (transcriber.getTranscriptionService().supportsStreamRecognition())
             // re-establish prematurely ended streaming session
@@ -704,6 +738,7 @@ else if (transcriber.getTranscriptionService().supportsStreamRecognition())
                 transcriber.getTranscriptionService().sendSingleRequest(
                         request,
                         this::notify);
+                incrementSentStats(audio.length);
             }
         });
     }