feat(android): natives PcmStreamRecorder-Modul — 16 kHz mono s16le → JS-Events

Neues Native-Modul fuer die Streaming-STT-Pipeline: PcmStreamRecorder.start() — oeffnet AudioRecord 16 kHz mono PCM, VOICE_COMMUNICATION-Source mit AEC/NS, PARTIAL_WAKE_LOCK gegen Doze PcmStreamRecorder.stop() — sauber schliessen Event "PcmStreamChunk" — {pcm: base64-s16le, seq, ts} alle 200ms Event "PcmStreamError" — bei Capture-Crash 200ms-Chunks: gross genug fuer geringen RVS-Overhead, klein genug fuer granulares Endpointing in der Whisper-Bridge. Mic-Ownership: darf NICHT parallel zu OpenWakeWord laufen — beide wollen AudioRecord. Coordination liegt bei audio.ts (stop OWW vor start, start OWW nach stop), genau wie's bisher mit react-native- audio-recorder-player gemacht wurde.
2026-05-30 21:33:18 +02:00
parent e99bf0b032
commit 199297a3a1
3 changed files with 263 additions and 0 deletions
@@ -21,6 +21,7 @@ class MainApplication : Application(), ReactApplication {
              add(ApkInstallerPackage())
              add(AudioFocusPackage())
              add(PcmStreamPlayerPackage())
+              add(PcmStreamRecorderPackage())
              add(OpenWakeWordPackage())
              add(PhoneCallPackage())
              add(BackgroundAudioPackage())
@@ -0,0 +1,246 @@
+package com.ariacockpit
+
+import android.Manifest
+import android.content.Context
+import android.content.pm.PackageManager
+import android.media.AudioFormat
+import android.media.AudioRecord
+import android.media.MediaRecorder
+import android.media.audiofx.AcousticEchoCanceler
+import android.media.audiofx.AutomaticGainControl
+import android.media.audiofx.NoiseSuppressor
+import android.os.PowerManager
+import android.util.Base64
+import android.util.Log
+import androidx.core.content.ContextCompat
+import com.facebook.react.bridge.Arguments
+import com.facebook.react.bridge.Promise
+import com.facebook.react.bridge.ReactApplicationContext
+import com.facebook.react.bridge.ReactContextBaseJavaModule
+import com.facebook.react.bridge.ReactMethod
+import com.facebook.react.modules.core.DeviceEventManagerModule
+import java.util.concurrent.atomic.AtomicBoolean
+
+/**
+ * PCM-Streaming-Recorder fuer die Streaming-Whisper-Bridge.
+ *
+ * Oeffnet AudioRecord (16 kHz mono s16le, VOICE_COMMUNICATION-Source mit
+ * automatischer AEC + NS) und feuert ~200ms-Chunks als base64-Event
+ * "PcmStreamChunk" an die JS-Bridge.
+ *
+ * audio.ts schickt die Chunks via RVS direkt an die whisper-bridge die
+ * dort einen ML-Endpointer laufen laesst — kein dB-VAD-Tuning mehr.
+ *
+ * Mic-Ownership: dieser Recorder DARF nicht gleichzeitig mit
+ * OpenWakeWord laufen — beide wollen AudioRecord vom MIC. Caller
+ * muss OpenWakeWord.stop() vor start() hier aufrufen und nach stop()
+ * hier wieder OpenWakeWord.start() — genau wie's audio.ts ohnehin
+ * macht.
+ *
+ * Events:
+ *   "PcmStreamChunk"   { pcm: base64-s16le, seq: N, ts: epochMs }
+ *   "PcmStreamError"   { error: string }
+ */
+class PcmStreamRecorderModule(reactContext: ReactApplicationContext) :
+    ReactContextBaseJavaModule(reactContext) {
+
+    override fun getName() = "PcmStreamRecorder"
+
+    companion object {
+        private const val TAG = "PcmStreamRecorder"
+        private const val SAMPLE_RATE = 16000
+        // 200ms-Chunks: gross genug fuer wenig RVS-Overhead, klein genug damit
+        // der Endpointer im Whisper-Bridge granular sieht. 200ms ist auch das
+        // Whisper-VAD-Frame-Hop — passt also zu downstream.
+        private const val CHUNK_SAMPLES = 3200       // 200ms @ 16 kHz
+        private const val BYTES_PER_SAMPLE = 2       // s16
+        private const val CHUNK_BYTES = CHUNK_SAMPLES * BYTES_PER_SAMPLE
+    }
+
+    private var audioRecord: AudioRecord? = null
+    private val running = AtomicBoolean(false)
+    private var captureThread: Thread? = null
+
+    private var aec: AcousticEchoCanceler? = null
+    private var ns: NoiseSuppressor? = null
+    private var agc: AutomaticGainControl? = null
+
+    // PARTIAL_WAKE_LOCK damit der JS-Bridge-Loop weiterlaeuft auch wenn das
+    // Display aus ist — sonst sammeln sich zwar Chunks in der nativen Queue
+    // an, aber emit() landet nicht zeitnah in JS und der Whisper-Bridge
+    // bekommt die Audio-Chunks erst beim App-Foreground-Resume.
+    private var wakeLock: PowerManager.WakeLock? = null
+
+    private var seq: Long = 0L
+
+    @ReactMethod
+    fun start(promise: Promise) {
+        if (running.get()) {
+            promise.resolve(true)
+            return
+        }
+        val perm = ContextCompat.checkSelfPermission(
+            reactApplicationContext, Manifest.permission.RECORD_AUDIO
+        )
+        if (perm != PackageManager.PERMISSION_GRANTED) {
+            promise.reject("NO_MIC_PERMISSION", "RECORD_AUDIO Permission fehlt")
+            return
+        }
+
+        try {
+            val minBuf = AudioRecord.getMinBufferSize(
+                SAMPLE_RATE,
+                AudioFormat.CHANNEL_IN_MONO,
+                AudioFormat.ENCODING_PCM_16BIT,
+            ).coerceAtLeast(CHUNK_BYTES * 4)  // 4x Chunk-Size als Sicherheit
+
+            val record = AudioRecord(
+                MediaRecorder.AudioSource.VOICE_COMMUNICATION,
+                SAMPLE_RATE,
+                AudioFormat.CHANNEL_IN_MONO,
+                AudioFormat.ENCODING_PCM_16BIT,
+                minBuf,
+            )
+            if (record.state != AudioRecord.STATE_INITIALIZED) {
+                record.release()
+                promise.reject("AUDIO_INIT", "AudioRecord nicht initialisiert (Mikro belegt? OpenWakeWord noch aktiv?)")
+                return
+            }
+            audioRecord = record
+
+            // AEC/NS/AGC explizit anschalten — manche Geraete liefern's via
+            // VOICE_COMMUNICATION zwar mit, aber Belt-and-Suspenders.
+            try {
+                if (AcousticEchoCanceler.isAvailable()) {
+                    aec = AcousticEchoCanceler.create(record.audioSessionId)?.apply { enabled = true }
+                }
+            } catch (e: Exception) { Log.w(TAG, "AEC failed: ${e.message}") }
+            try {
+                if (NoiseSuppressor.isAvailable()) {
+                    ns = NoiseSuppressor.create(record.audioSessionId)?.apply { enabled = true }
+                }
+            } catch (e: Exception) { Log.w(TAG, "NS failed: ${e.message}") }
+            try {
+                if (AutomaticGainControl.isAvailable()) {
+                    agc = AutomaticGainControl.create(record.audioSessionId)?.apply { enabled = true }
+                }
+            } catch (e: Exception) { Log.w(TAG, "AGC failed: ${e.message}") }
+
+            seq = 0L
+            running.set(true)
+            record.startRecording()
+
+            try {
+                val pm = reactApplicationContext.getSystemService(Context.POWER_SERVICE) as PowerManager
+                wakeLock = pm.newWakeLock(PowerManager.PARTIAL_WAKE_LOCK,
+                                          "AriaCockpit:PcmStreamRecord").apply {
+                    setReferenceCounted(false)
+                    acquire(8 * 60 * 60 * 1000L)  // 8h Cap
+                }
+            } catch (e: Exception) {
+                Log.w(TAG, "WakeLock acquire fehlgeschlagen: ${e.message}")
+            }
+
+            captureThread = Thread({ captureLoop() }, "PcmStreamRecorderCapture").apply {
+                isDaemon = true
+                start()
+            }
+
+            Log.i(TAG, "Recording gestartet (16kHz mono s16le, ${CHUNK_SAMPLES} samples/chunk)")
+            promise.resolve(true)
+        } catch (e: Exception) {
+            Log.e(TAG, "start fehlgeschlagen", e)
+            running.set(false)
+            audioRecord?.release()
+            audioRecord = null
+            releaseAudioEffects()
+            releaseWakeLock()
+            promise.reject("START_FAILED", e.message ?: "Unbekannter Fehler", e)
+        }
+    }
+
+    @ReactMethod
+    fun stop(promise: Promise) {
+        running.set(false)
+        try {
+            captureThread?.join(1500)
+        } catch (_: InterruptedException) {}
+        captureThread = null
+        try { audioRecord?.stop() } catch (_: Exception) {}
+        try { audioRecord?.release() } catch (_: Exception) {}
+        audioRecord = null
+        releaseAudioEffects()
+        releaseWakeLock()
+        Log.i(TAG, "Recording gestoppt (seq=$seq Chunks gesendet)")
+        promise.resolve(true)
+    }
+
+    @ReactMethod
+    fun isRecording(promise: Promise) {
+        promise.resolve(running.get())
+    }
+
+    private fun captureLoop() {
+        val buffer = ByteArray(CHUNK_BYTES)
+        val rec = audioRecord ?: return
+        try {
+            while (running.get()) {
+                var offset = 0
+                // Solange lesen bis ein voller 200ms-Chunk zusammen ist.
+                // AudioRecord.read kann weniger als angefordert liefern.
+                while (offset < CHUNK_BYTES && running.get()) {
+                    val n = rec.read(buffer, offset, CHUNK_BYTES - offset)
+                    if (n <= 0) {
+                        if (!running.get()) break
+                        // Fehlerzustand — kurze Pause, dann weiter probieren
+                        Thread.sleep(5)
+                        continue
+                    }
+                    offset += n
+                }
+                if (offset < CHUNK_BYTES) break
+
+                val b64 = Base64.encodeToString(buffer, Base64.NO_WRAP)
+                val ts = System.currentTimeMillis()
+                val params = Arguments.createMap().apply {
+                    putString("pcm", b64)
+                    // putLong existiert nicht in WritableMap — putDouble fuer ts/seq.
+                    putDouble("seq", seq.toDouble())
+                    putDouble("ts", ts.toDouble())
+                }
+                reactApplicationContext
+                    .getJSModule(DeviceEventManagerModule.RCTDeviceEventEmitter::class.java)
+                    .emit("PcmStreamChunk", params)
+                seq++
+            }
+        } catch (e: Exception) {
+            Log.e(TAG, "captureLoop crashed", e)
+            try {
+                val err = Arguments.createMap().apply {
+                    putString("error", e.message ?: "unknown")
+                }
+                reactApplicationContext
+                    .getJSModule(DeviceEventManagerModule.RCTDeviceEventEmitter::class.java)
+                    .emit("PcmStreamError", err)
+            } catch (_: Exception) {}
+        }
+    }
+
+    private fun releaseAudioEffects() {
+        try { aec?.release() } catch (_: Exception) {}
+        try { ns?.release() } catch (_: Exception) {}
+        try { agc?.release() } catch (_: Exception) {}
+        aec = null; ns = null; agc = null
+    }
+
+    private fun releaseWakeLock() {
+        try {
+            if (wakeLock?.isHeld == true) wakeLock?.release()
+        } catch (_: Exception) {}
+        wakeLock = null
+    }
+
+    // Damit RCTEventEmitter den Listener-Lifecycle nicht crasht
+    @ReactMethod fun addListener(eventName: String) {}
+    @ReactMethod fun removeListeners(count: Int) {}
+}
@@ -0,0 +1,16 @@
+package com.ariacockpit
+
+import com.facebook.react.ReactPackage
+import com.facebook.react.bridge.NativeModule
+import com.facebook.react.bridge.ReactApplicationContext
+import com.facebook.react.uimanager.ViewManager
+
+class PcmStreamRecorderPackage : ReactPackage {
+    override fun createNativeModules(reactContext: ReactApplicationContext): List<NativeModule> {
+        return listOf(PcmStreamRecorderModule(reactContext))
+    }
+
+    override fun createViewManagers(reactContext: ReactApplicationContext): List<ViewManager<*, *>> {
+        return emptyList()
+    }
+}