@siteed/expo-audio-stream 1.4.0 → 1.5.0

package/CHANGELOG.md

@@ -8,13 +8,19 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ## [Unreleased]
 
 
+## [1.5.0] - 2024-12-10
+- feat(expo-audio-stream): add comprehensive ios audio session configuration support (#54) ([ba296ac](https://github.com/deeeed/expo-audio-stream/commit/ba296ac9be0d61e7a46cc64953e6f2c66881fdfd))
+- fix(expo-audio-stream): prevent invalid WAV files when stopping recording too quickly (#53) ([80f4898](https://github.com/deeeed/expo-audio-stream/commit/80f4898625cea52da8f3e34e425e61d7641353f7))
+
 ## [1.4.0] - 2024-12-05
 - chore: remove unusded dependencies ([ad81dd5](https://github.com/deeeed/expo-audio-stream/commit/ad81dd560c93dd1d04995a323a4ae72d4de20f3e))
 
+
 ## [1.3.1] - 2024-12-05
 - feat(web): implement throttling and optimize event processing (#49) ([da28765](https://github.com/deeeed/expo-audio-stream/commit/da2876524c2c9d6e0a980fde40a0197b929d8a7f))
 
 
+
 ## [1.3.0] - 2024-11-28
 ### Added
 - refactor(permissions): standardize permission status response structure across platforms (#44) ([7c9c800](https://github.com/deeeed/expo-audio-stream/commit/7c9c800d83b7cea3516643371484d5e1f3b99e4c))
@@ -25,6 +31,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 
 
+
 ## [1.2.5] - 2024-11-12
 ### Added
 - docs(license): add MIT license to all packages (6 files changed)
@@ -32,6 +39,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 
 
+
 ## [1.2.4] - 2024-11-05
 ### Changed
 - Android minimum audio interval set to 10ms. 
@@ -42,6 +50,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 
 
+
 ## [1.2.0] - 2024-10-24
 ### Added
 - Feature: Keep device awake during recording with `keepAwake` option
@@ -52,6 +61,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 
 
+
 ## [1.1.17] - 2024-10-21
 ### Added
 - Support bluetooth headset on ios
@@ -59,6 +69,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 
 
+
 ## [1.0.0] - 2024-04-01
 
 ### Added
@@ -70,7 +81,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Feature: Audio features extraction during recording.
 - Feature: Consistent WAV PCM recording format across all platforms.
 
-[unreleased]: https://github.com/deeeed/expo-audio-stream/compare/@siteed/expo-audio-stream@1.4.0...HEAD
+[unreleased]: https://github.com/deeeed/expo-audio-stream/compare/@siteed/expo-audio-stream@1.5.0...HEAD
+[1.5.0]: https://github.com/deeeed/expo-audio-stream/compare/@siteed/expo-audio-stream@1.4.0...@siteed/expo-audio-stream@1.5.0
 [1.4.0]: https://github.com/deeeed/expo-audio-stream/compare/@siteed/expo-audio-stream@1.3.1...@siteed/expo-audio-stream@1.4.0
 [1.3.1]: https://github.com/deeeed/expo-audio-stream/compare/@siteed/expo-audio-stream@1.3.0...@siteed/expo-audio-stream@1.3.1
 [Unreleased]: https://github.com/deeeed/expo-audio-stream/compare/v1.0.0...HEAD

package/android/src/main/java/net/siteed/audiostream/AudioNotificationsManager.kt

@@ -66,7 +66,7 @@ class AudioNotificationManager private constructor(context: Context) {
             val channel = NotificationChannel(
                 recordingConfig.notification.channelId,
                 recordingConfig.notification.channelName,
-                NotificationManager.IMPORTANCE_HIGH
+                NotificationManager.IMPORTANCE_LOW // Lower importance means no sound by default
             ).apply {
                 description = recordingConfig.notification.channelDescription
                 enableLights(true)
@@ -74,6 +74,13 @@ class AudioNotificationManager private constructor(context: Context) {
                 enableVibration(true)
                 setShowBadge(true)
             }
+
+            // Set description, disable lights, vibration, and sound.
+            channel.enableLights(false);
+            channel.enableVibration(false);
+            channel.setSound(null, null);
+            channel.setShowBadge(false);
+
             notificationManager.createNotificationChannel(channel)
         }
     }

package/ios/AudioStreamManager.swift

@@ -65,7 +65,7 @@ class AudioStreamManager: NSObject {
     private var remoteCommandCenter: MPRemoteCommandCenter?
 
     weak var delegate: AudioStreamManagerDelegate?  // Define the delegate here
-    
+        
     /// Initializes the AudioStreamManager
     override init() {
         super.init()
@@ -415,18 +415,19 @@ class AudioStreamManager: NSObject {
         }
         
         var newSettings = settings  // Make settings mutable
+        let session = AVAudioSession.sharedInstance()
         
         // Determine the commonFormat based on bitDepth
         let commonFormat: AVAudioCommonFormat
         switch newSettings.bitDepth {
-        case 16:
-            commonFormat = .pcmFormatInt16
-        case 32:
-            commonFormat = .pcmFormatInt32
-        default:
-            Logger.debug("Unsupported bit depth. Defaulting to 16-bit PCM")
-            commonFormat = .pcmFormatInt16
-            newSettings.bitDepth = 16
+            case 16:
+                commonFormat = .pcmFormatInt16
+            case 32:
+                commonFormat = .pcmFormatInt32
+            default:
+                Logger.debug("Unsupported bit depth. Defaulting to 16-bit PCM")
+                commonFormat = .pcmFormatInt16
+                newSettings.bitDepth = 16
         }
         
         emissionInterval = max(100.0, Double(intervalMilliseconds)) / 1000.0
@@ -436,7 +437,6 @@ class AudioStreamManager: NSObject {
         pausedDuration = 0
         isPaused = false
         
-        let session = AVAudioSession.sharedInstance()
         do {
             Logger.debug("Debug: Configuring audio session with sample rate: \(settings.sampleRate) Hz")
             
@@ -448,15 +448,41 @@ class AudioStreamManager: NSObject {
                 newSettings.sampleRate = session.sampleRate
             }
             
-            try session.setCategory(.playAndRecord,
-                mode: .default,
-                options: [.allowBluetooth, .defaultToSpeaker, .mixWithOthers]
-            )
-
+            // Configure audio session based on iOS settings if provided
+            if let audioSessionConfig = settings.ios?.audioSession {
+                try session.setCategory(audioSessionConfig.category, options: audioSessionConfig.categoryOptions)
+                try session.setMode(audioSessionConfig.mode)
+                Logger.debug("""
+                        Audio Session Configuration (Custom):
+                        - Category: \(audioSessionConfig.category)
+                        - Mode: \(audioSessionConfig.mode)
+                        - Options: \(audioSessionConfig.categoryOptions)
+                        """)
+            } else {
+                let defaultCategory = AVAudioSession.Category.playAndRecord
+                let defaultMode = AVAudioSession.Mode.default
+                let defaultOptions: AVAudioSession.CategoryOptions = [
+                    .allowBluetooth,
+                    .defaultToSpeaker,
+                    .mixWithOthers,
+                    .allowBluetoothA2DP,
+                    .allowAirPlay
+                ]
+                
+                try session.setCategory(defaultCategory, mode: defaultMode, options: defaultOptions)
+                Logger.debug("""
+                    Audio Session Configuration (Default):
+                    - Category: \(defaultCategory)
+                    - Mode: \(defaultMode)
+                    - Options: \(defaultOptions)
+                    """)
+            }
+            
             try session.setPreferredSampleRate(settings.sampleRate)
             try session.setPreferredIOBufferDuration(1024 / settings.sampleRate)
             try session.setActive(true)
             Logger.debug("Debug: Audio session activated successfully.")
+
             
             let actualSampleRate = session.sampleRate
             if actualSampleRate != newSettings.sampleRate {
@@ -697,8 +723,14 @@ class AudioStreamManager: NSObject {
             let fileAttributes = try FileManager.default.attributesOfItem(atPath: filePath)
             let fileSize = fileAttributes[FileAttributeKey.size] as? Int64 ?? 0
             
+            // Return nil if the file is too small (less than WAV header size)
+            if fileSize <= 44 {
+                Logger.debug("Recording file is too small (≤ 44 bytes), likely no audio data was recorded")
+                return nil
+            }
+            
             // Update the WAV header with the correct file size
-            updateWavHeader(fileURL: fileURL, totalDataSize: fileSize - 44) // Subtract the header size to get audio data size
+            updateWavHeader(fileURL: fileURL, totalDataSize: fileSize - 44)
             
             let result = RecordingResult(
                 fileUri: fileURL.absoluteString,
@@ -803,24 +835,30 @@ class AudioStreamManager: NSObject {
     ///   - fileURL: The URL of the WAV file.
     ///   - totalDataSize: The total size of the audio data.
     private func updateWavHeader(fileURL: URL, totalDataSize: Int64) {
+        // Prevent negative values - minimum WAV file size should be at least the header size (44 bytes)
+        guard totalDataSize >= 0 else {
+            Logger.debug("Invalid file size: total data size is negative")
+            return
+        }
+
         do {
             let fileHandle = try FileHandle(forUpdating: fileURL)
             defer { fileHandle.closeFile() }
-            
+
             // Calculate sizes
             let fileSize = totalDataSize + 44 - 8 // Total file size minus 8 bytes for 'RIFF' and size field itself
             let dataSize = totalDataSize // Size of the 'data' sub-chunk
-            
+
             // Update RIFF chunk size at offset 4
             fileHandle.seek(toFileOffset: 4)
             let fileSizeBytes = UInt32(fileSize).littleEndianBytes
             fileHandle.write(Data(fileSizeBytes))
-            
+
             // Update data chunk size at offset 40
             fileHandle.seek(toFileOffset: 40)
             let dataSizeBytes = UInt32(dataSize).littleEndianBytes
             fileHandle.write(Data(dataSizeBytes))
-            
+
         } catch let error {
             Logger.debug("Error updating WAV header: \(error)")
         }

package/ios/RecordingSettings.swift

@@ -1,13 +1,20 @@
 // RecordingSettings.swift
 
+import AVFoundation
+
 struct NotificationAction {
     var title: String
     var identifier: String
 }
 
+struct IOSAudioSessionConfig {
+    var category: AVAudioSession.Category
+    var mode: AVAudioSession.Mode
+    var categoryOptions: AVAudioSession.CategoryOptions
+}
+
 struct IOSNotificationConfig {
     var categoryIdentifier: String?
-    var actions: [NotificationAction]?
 }
 
 struct NotificationConfig {
@@ -17,6 +24,10 @@ struct NotificationConfig {
     var ios: IOSNotificationConfig?
 }
 
+struct IOSConfig {
+    var audioSession: IOSAudioSessionConfig?
+}
+
 struct RecordingSettings {
     // Core recording settings
     var sampleRate: Double
@@ -35,6 +46,9 @@ struct RecordingSettings {
     var algorithm: String? = "rms"
     var featureOptions: [String: Bool]? = ["rms": true, "zcr": true]
     
+    // iOS-specific configuration
+    var ios: IOSConfig?
+    
     // Notification configuration
     var notification: NotificationConfig?
     
@@ -59,6 +73,68 @@ struct RecordingSettings {
         settings.algorithm = dict["algorithm"] as? String
         settings.featureOptions = dict["features"] as? [String: Bool]
         
+        // Parse iOS-specific config
+        if let iosDict = dict["ios"] as? [String: Any],
+           let audioSessionDict = iosDict["audioSession"] as? [String: Any] {
+            
+            // Map category
+            let category: AVAudioSession.Category
+            if let categoryStr = audioSessionDict["category"] as? String {
+                switch categoryStr {
+                    case "Ambient": category = .ambient
+                    case "SoloAmbient": category = .soloAmbient
+                    case "Playback": category = .playback
+                    case "Record": category = .record
+                    case "PlayAndRecord": category = .playAndRecord
+                    case "MultiRoute": category = .multiRoute
+                    default: category = .record
+                }
+            } else {
+                category = .record
+            }
+            
+            // Map mode
+            let mode: AVAudioSession.Mode
+            if let modeStr = audioSessionDict["mode"] as? String {
+                switch modeStr {
+                    case "Default": mode = .default
+                    case "VoiceChat": mode = .voiceChat
+                    case "VideoChat": mode = .videoChat
+                    case "GameChat": mode = .gameChat
+                    case "VideoRecording": mode = .videoRecording
+                    case "Measurement": mode = .measurement
+                    case "MoviePlayback": mode = .moviePlayback
+                    case "SpokenAudio": mode = .spokenAudio
+                    default: mode = .default
+                }
+            } else {
+                mode = .default
+            }
+            
+            // Map category options
+            var categoryOptions: AVAudioSession.CategoryOptions = []
+            if let optionsArray = audioSessionDict["categoryOptions"] as? [String] {
+                for option in optionsArray {
+                    switch option {
+                        case "MixWithOthers": categoryOptions.insert(.mixWithOthers)
+                        case "DuckOthers": categoryOptions.insert(.duckOthers)
+                        case "InterruptSpokenAudioAndMixWithOthers": categoryOptions.insert(.interruptSpokenAudioAndMixWithOthers)
+                        case "AllowBluetooth": categoryOptions.insert(.allowBluetooth)
+                        case "AllowBluetoothA2DP": categoryOptions.insert(.allowBluetoothA2DP)
+                        case "AllowAirPlay": categoryOptions.insert(.allowAirPlay)
+                        case "DefaultToSpeaker": categoryOptions.insert(.defaultToSpeaker)
+                        default: break
+                    }
+                }
+            }
+            
+            settings.ios = IOSConfig(audioSession: IOSAudioSessionConfig(
+                category: category,
+                mode: mode,
+                categoryOptions: categoryOptions
+            ))
+        }
+        
         // Parse notification config
         if let notificationDict = dict["notification"] as? [String: Any] {
             var notificationConfig = NotificationConfig()
@@ -67,20 +143,10 @@ struct RecordingSettings {
             notificationConfig.icon = notificationDict["icon"] as? String
             
             // Parse iOS-specific notification config
-            if let iosDict = notificationDict["ios"] as? [String: Any] {
-                var iosConfig = IOSNotificationConfig()
-                iosConfig.categoryIdentifier = iosDict["categoryIdentifier"] as? String
-                
-                if let actionsArray = iosDict["actions"] as? [[String: Any]] {
-                    iosConfig.actions = actionsArray.map { actionDict in
-                        NotificationAction(
-                            title: actionDict["title"] as? String ?? "",
-                            identifier: actionDict["identifier"] as? String ?? ""
-                        )
-                    }
-                }
-                
-                notificationConfig.ios = iosConfig
+            if let iosNotificationDict = notificationDict["ios"] as? [String: Any] {
+                notificationConfig.ios = IOSNotificationConfig(
+                    categoryIdentifier: iosNotificationDict["categoryIdentifier"] as? String
+                )
             }
             
             settings.notification = notificationConfig

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@siteed/expo-audio-stream",
-  "version": "1.4.0",
+  "version": "1.5.0",
   "description": "stream audio crossplatform",
   "license": "MIT",
   "main": "build/index.js",
@@ -21,6 +21,26 @@
     "expo-audio-stream",
     "ExpoAudioStream"
   ],
+  "files": [
+    "src",
+    "android",
+    "ios",
+    "cpp",
+    "generated",
+    "README.md",
+    "package.json",
+    "*.podspec",
+    "!ios/build",
+    "!android/build",
+    "!android/gradle",
+    "!android/gradlew",
+    "!android/gradlew.bat",
+    "!android/local.properties",
+    "!**/__tests__",
+    "!**/__fixtures__",
+    "!**/__mocks__",
+    "!**/.*"
+  ],
   "scripts": {
     "build": "expo-module build",
     "clean": "expo-module clean",
@@ -53,9 +73,9 @@
     "eslint-plugin-prettier": "^5.1.3",
     "eslint-plugin-promise": "^6.1.1",
     "eslint-plugin-react": "^7.34.1",
-    "expo": "^52.0.8",
+    "expo": "^52.0.17",
     "expo-module-scripts": "^4.0.2",
-    "expo-modules-core": "^2.0.6",
+    "expo-modules-core": "^2.1.1",
     "jest": "^29.7.0",
     "prettier": "^3.2.5",
     "react-native": "0.74.5",

package/src/ExpoAudioStream.types.ts

@@ -70,6 +70,38 @@ export interface StartRecordingResult {
     sampleRate?: SampleRate
 }
 
+export interface AudioSessionConfig {
+    category?:
+        | 'Ambient'
+        | 'SoloAmbient'
+        | 'Playback'
+        | 'Record'
+        | 'PlayAndRecord'
+        | 'MultiRoute'
+    mode?:
+        | 'Default'
+        | 'VoiceChat'
+        | 'VideoChat'
+        | 'GameChat'
+        | 'VideoRecording'
+        | 'Measurement'
+        | 'MoviePlayback'
+        | 'SpokenAudio'
+    categoryOptions?: (
+        | 'MixWithOthers'
+        | 'DuckOthers'
+        | 'InterruptSpokenAudioAndMixWithOthers'
+        | 'AllowBluetooth'
+        | 'AllowBluetoothA2DP'
+        | 'AllowAirPlay'
+        | 'DefaultToSpeaker'
+    )[]
+}
+
+export interface IOSConfig {
+    audioSession?: AudioSessionConfig
+}
+
 export interface RecordingConfig {
     // Sample rate for recording (16000, 44100, or 48000 Hz)
     sampleRate?: SampleRate
@@ -98,6 +130,9 @@ export interface RecordingConfig {
     // Enable audio processing (default is false)
     enableProcessing?: boolean
 
+    // iOS-specific configuration
+    ios?: IOSConfig
+
     // Number of data points to extract per second of audio (default is 1000)
     pointsPerSecond?: number
 

package/src/ExpoAudioStream.web.ts

@@ -30,7 +30,7 @@ export interface ExpoAudioStreamWebProps {
 
 export class ExpoAudioStreamWeb extends LegacyEventEmitter {
     customRecorder: WebRecorder | null
-    audioChunks: ArrayBuffer[]
+    audioChunks: Float32Array[]
     isRecording: boolean
     isPaused: boolean
     recordingStartTime: number
@@ -118,7 +118,7 @@ export class ExpoAudioStreamWeb extends LegacyEventEmitter {
                 data,
                 position,
             }: EmitAudioEventProps) => {
-                this.audioChunks.push(data)
+                this.audioChunks.push(new Float32Array(data))
                 this.currentSize += data.byteLength
                 this.emitAudioEvent({ data, position })
                 this.lastEmittedTime = Date.now()
@@ -190,7 +190,7 @@ export class ExpoAudioStreamWeb extends LegacyEventEmitter {
 
         // Rewrite wav header with correct data size
         const wavConfig: WavHeaderOptions = {
-            buffer: fullPcmBufferArray.buffer,
+            buffer: new ArrayBuffer(fullPcmBufferArray.buffer.byteLength),
             sampleRate: this.recordingConfig?.sampleRate ?? 44100,
             numChannels: this.recordingConfig?.channels ?? 1,
             bitDepth: this.bitDepth,