@livekit/react-native@3.0.0

Major Changes

• Move to the LiveKit prefixed WebRTC builds: io.github.webrtc-sdk:android-prefixed on Android and the LiveKitWebRTC pod on iOS. This prevents collisions against any other WebRTC implementations. - #455 (@davidliu)

  Apps that only use the JavaScript API need no changes.

  Native integrations must migrate to the prefixed symbols:

  • Android: org.webrtc.* becomes livekit.org.webrtc.*.
    Custom -keep class org.webrtc.** ProGuard rules must be updated to livekit.org.webrtc.**.
  • iOS: types gain an LK prefix (RTCAudioRenderer becomes LKRTCAudioRenderer, RTCAudioBuffer becomes LKRTCAudioBuffer, and so on).

Minor Changes

• iOS: align the default playAndRecord presets with the LiveKit Swift SDK. - #455 (@davidliu)

  The duplex (recording) presets now request mixWithOthers, allowBluetooth,
  allowBluetoothA2DP and allowAirPlay, matching playAndRecordOptions in the
  Swift SDK. Previously only allowBluetooth and mixWithOthers were requested,
  so A2DP output devices and AirPlay routes were unavailable during a call on this
  platform but available on others.

  When preferSpeakerOutput is set, defaultToSpeaker is now requested as well.
  The videoChat mode implies a speaker route, but iOS may rewrite the mode when
  Voice Processing I/O is instantiated - the Swift SDK observed it switching to
  voiceChat, which routes to the receiver. Requesting the option explicitly
  keeps the speaker route across that rewrite.

  This affects the defaults used by setupIOSAudioManagement and the deprecated
  getDefaultAppleAudioConfigurationForMode. A custom IOSAudioSessionPolicy or
  AppleAudioConfiguration is unaffected - those supply their own options.

• iOS: pick the audio session mode based on Apple Voice Processing I/O state. - #455 (@davidliu)

  iOS applies a reduced, call-tuned speaker gain while capture is active under the
  voiceChat/videoChat session modes. Apple Voice Processing I/O compensates
  with its own loudness stage, so with VPIO off remote audio played back
  noticeably quieter.

  LiveKitWebRTC 144.7559.15 reports the resolved VPIO state to the audio device
  module's willEnableEngine hook. The default recording configuration now uses
  that: with VPIO running it keeps voiceChat/videoChat as before, and with VPIO
  off (for example after AudioDeviceModule.setVoiceProcessingEnabled(false)) it
  switches to the default mode, which keeps media gain. That mode routes to the
  receiver, so preferSpeakerOutput is expressed through the defaultToSpeaker
  category option instead of the implicit routing the chat modes provide. This
  matches what the LiveKit Swift SDK already does.

  IOSAudioSessionPolicy gains an optional recordingWithoutVoiceProcessing
  configuration for the VPIO-off case. Policies that omit it keep using their
  recording configuration for both, so existing setups are unaffected.
  AudioEngineConfigurationState, passed to the deprecated
  setupIOSAudioManagement callback form, gains a matching
  isVoiceProcessingEnabled field.

  Requires @livekit/react-native-webrtc with the five-argument willEnableEngine
  delegate method. Older versions crash with an unrecognized selector when the
  audio engine first enables.