Make audio controller an actor

Author: daymxn

firebaseai/LiveAudioExample/Audio/AudioController.swift

@@ -15,14 +15,15 @@
 import AVFoundation
 
 /// Controls audio playback and recording.
-class AudioController {
+actor AudioController {
   /// Data processed from the microphone.
   private let microphoneData: AsyncStream<AVAudioPCMBuffer>
   private let microphoneDataQueue: AsyncStream<AVAudioPCMBuffer>.Continuation
   private var audioPlayer: AudioPlayer?
   private var audioEngine: AVAudioEngine?
   private var microphone: Microphone?
   private var listenTask: Task<Void, Never>?
+  private var routeTask: Task<Void, Never>?
 
   /// Port types that are considered "headphones" for our use-case.
   ///
@@ -40,7 +41,7 @@ class AudioController {
 
   private var stopped = false
 
-  public init() throws {
+  public init() async throws {
     let session = AVAudioSession.sharedInstance()
     try session.setCategory(
       .playAndRecord,
@@ -80,7 +81,25 @@ class AudioController {
   }
 
   deinit {
-    stop()
+    stopped = true
+    listenTask?.cancel()
+    // audio engine needs to be stopped before disconnecting nodes
+    audioEngine?.pause()
+    audioEngine?.stop()
+    if let audioEngine {
+      do {
+        // the VP IO leaves behind artifacts, so we need to disable it to properly clean up
+        if audioEngine.inputNode.isVoiceProcessingEnabled {
+          try audioEngine.inputNode.setVoiceProcessingEnabled(false)
+        }
+      } catch {
+        print("Failed to disable voice processing: \(error.localizedDescription)")
+      }
+    }
+    microphone?.stop()
+    audioPlayer?.stop()
+    microphoneDataQueue.finish()
+    routeTask?.cancel()
   }
 
   /// Kicks off audio processing, and returns a stream of recorded microphone audio data.
@@ -96,11 +115,7 @@ class AudioController {
     stopped = true
     stopListeningAndPlayback()
     microphoneDataQueue.finish()
-    NotificationCenter.default.removeObserver(
-      self,
-      name: AVAudioSession.routeChangeNotification,
-      object: nil
-    )
+    routeTask?.cancel()
   }
 
   /// Queues audio for playback.
@@ -206,15 +221,17 @@ class AudioController {
 
   /// When the output device changes, ensure the audio playback and recording classes are properly restarted.
   private func listenForRouteChange() {
-    NotificationCenter.default.addObserver(
-      self,
-      selector: #selector(handleRouteChange),
-      name: AVAudioSession.routeChangeNotification,
-      object: nil
-    )
+    routeTask?.cancel()
+    routeTask = Task { [weak self] in
+      for await notification in NotificationCenter.default.notifications(
+        named: AVAudioSession.routeChangeNotification
+      ) {
+        await self?.handleRouteChange(notification: notification)
+      }
+    }
   }
 
-  @objc private func handleRouteChange(notification: Notification) {
+  private func handleRouteChange(notification: Notification) {
     guard let userInfo = notification.userInfo,
       let reasonValue = userInfo[AVAudioSessionRouteChangeReasonKey] as? UInt,
       let reason = AVAudioSession.RouteChangeReason(rawValue: reasonValue) else {

firebaseai/LiveAudioExample/ViewModels/LiveViewModel.swift

@@ -94,9 +94,9 @@ class LiveViewModel: ObservableObject {
 
     do {
       liveSession = try await model.connect()
-      audioController = try AudioController()
+      audioController = try await AudioController()
 
-      try startRecording()
+      try await startRecording()
 
       state = .connected
       try await startProcessingResponses()
@@ -111,7 +111,7 @@ class LiveViewModel: ObservableObject {
   ///
   /// Will stop any pending playback, and the recording of the mic.
   func disconnect() async {
-    audioController?.stop()
+    await audioController?.stop()
     await liveSession?.close()
     microphoneTask.cancel()
     state = .idle
@@ -124,10 +124,10 @@ class LiveViewModel: ObservableObject {
   }
 
   /// Starts recording data from the user's microphone, and sends it to the model.
-  private func startRecording() throws {
+  private func startRecording() async throws {
     guard let audioController, let liveSession else { return }
 
-    let stream = try audioController.listenToMic()
+    let stream = try await audioController.listenToMic()
     microphoneTask = Task {
       for await audioBuffer in stream {
         await liveSession.sendAudioRealtime(audioBuffer.int16Data())
@@ -190,7 +190,7 @@ class LiveViewModel: ObservableObject {
 
     if content.wasInterrupted {
       logger.warning("Model was interrupted")
-      audioController?.interrupt()
+      await audioController?.interrupt()
       transcriptViewModel.clearPending()
       // adds an em dash to indiciate that the model was cutoff
       transcriptViewModel.appendTranscript("— ")
@@ -203,7 +203,7 @@ class LiveViewModel: ObservableObject {
     for part in content.parts {
       if let part = part as? InlineDataPart {
         if part.mimeType.starts(with: "audio/pcm") {
-          audioController?.playAudio(audio: part.data)
+          await audioController?.playAudio(audio: part.data)
         } else {
           logger.warning("Received non audio inline data part: \(part.mimeType)")
         }