@edkimmel/expo-audio-stream 0.4.1 → 0.5.0

package/ios/Microphone.swift

@@ -28,6 +28,14 @@ class Microphone {
     private var isSilent: Bool = false
     private var frequencyBandAnalyzer: FrequencyBandAnalyzer?
     private var frequencyBandConfig: (lowCrossoverHz: Float, highCrossoverHz: Float)?
+
+    /// Interval (in ms) the consumer last requested. Used to rebuild the tap
+    /// with the same cadence when resuming after an interruption.
+    private var lastIntervalMs: Int = 100
+    /// Set when an audio session interruption begins while recording is active.
+    /// Cleared either by `stopRecording` (consumer chose disconnect) or by the
+    /// `.ended` handler after a successful auto-resume.
+    private var pendingInterruptionResume: Bool = false
     
     init() {
         NotificationCenter.default.addObserver(
@@ -36,8 +44,18 @@ class Microphone {
             name: AVAudioSession.routeChangeNotification,
             object: nil
         )
+        NotificationCenter.default.addObserver(
+            self,
+            selector: #selector(handleInterruption),
+            name: AVAudioSession.interruptionNotification,
+            object: nil
+        )
     }
-    
+
+    deinit {
+        NotificationCenter.default.removeObserver(self)
+    }
+
     /// Handles audio route changes (e.g. headphones connected/disconnected)
     /// - Parameter notification: The notification object containing route change information
     @objc private func handleRouteChange(notification: Notification) {
@@ -46,7 +64,7 @@ class Microphone {
               let reason = AVAudioSession.RouteChangeReason(rawValue: reasonValue) else {
             return
         }
-        
+
         Logger.debug("[Microphone] Route is changed \(reason)")
 
         switch reason {
@@ -55,7 +73,7 @@ class Microphone {
                 stopRecording(resolver: nil)
                 DispatchQueue.main.asyncAfter(deadline: .now() + 0.5) { [weak self] in
                     guard let self = self, let settings = self.recordingSettings else { return }
-                    
+
                     _ = startRecording(settings: self.recordingSettings!, intervalMilliseconds: 100, frequencyBandConfig: self.frequencyBandConfig)
                 }
             }
@@ -65,6 +83,96 @@ class Microphone {
             break
         }
     }
+
+    /// Handles audio session interruptions (e.g. Siri, phone call, alarm).
+    ///
+    /// On `.began` we stop the engine (iOS deactivates the session regardless)
+    /// but keep `isRecording = true` and emit an `INTERRUPTED` error so the
+    /// consumer can decide:
+    ///   - Call `stopRecording` to disconnect explicitly (clears the resume flag).
+    ///   - Do nothing to let the library auto-resume when the system gives the
+    ///     mic back on `.ended`.
+    @objc private func handleInterruption(notification: Notification) {
+        guard let info = notification.userInfo,
+              let typeValue = info[AVAudioSessionInterruptionTypeKey] as? UInt,
+              let type = AVAudioSession.InterruptionType(rawValue: typeValue) else {
+            return
+        }
+
+        switch type {
+        case .began:
+            Logger.debug("[Microphone] Audio session interruption began")
+            guard isRecording else { return }
+            // Tear down the engine but leave isRecording true so the consumer's
+            // intent is preserved. The .ended branch will decide whether to
+            // resume based on whether stopRecording was called in between.
+            pendingInterruptionResume = true
+            if let engine = audioEngine {
+                engine.inputNode.removeTap(onBus: 0)
+                engine.stop()
+            }
+            delegate?.onMicrophoneError("INTERRUPTED", "Audio session interrupted by system")
+        case .ended:
+            Logger.debug("[Microphone] Audio session interruption ended")
+            guard pendingInterruptionResume else { return }
+            pendingInterruptionResume = false
+            // iOS uses AVAudioSessionInterruptionOptionKey.shouldResume to hint
+            // whether the app can safely reactivate. Absent means another app
+            // took over (e.g. an ongoing call) — in that case we surface a
+            // terminal error instead of attempting a doomed restart.
+            let optionsValue = info[AVAudioSessionInterruptionOptionKey] as? UInt ?? 0
+            let options = AVAudioSession.InterruptionOptions(rawValue: optionsValue)
+            guard options.contains(.shouldResume) else {
+                Logger.debug("[Microphone] System did not signal shouldResume — treating as terminal stop")
+                isRecording = false
+                delegate?.onMicrophoneError(
+                    "RESUME_DENIED",
+                    "System did not permit microphone resume after interruption"
+                )
+                return
+            }
+            do {
+                try AVAudioSession.sharedInstance().setActive(true)
+                try installTapAndStartEngine(intervalMilliseconds: lastIntervalMs)
+                Logger.debug("[Microphone] Auto-resumed after interruption")
+            } catch {
+                Logger.debug("[Microphone] Auto-resume failed: \(error.localizedDescription)")
+                // Engine couldn't restart — surface as terminal stop.
+                isRecording = false
+                delegate?.onMicrophoneError(
+                    "RESUME_FAILED",
+                    "Could not restart microphone after interruption: \(error.localizedDescription)"
+                )
+            }
+        @unknown default:
+            break
+        }
+    }
+
+    /// Installs the audio tap on the input node and starts the engine.
+    /// Shared between fresh `startRecording` and post-interruption resume.
+    private func installTapAndStartEngine(intervalMilliseconds: Int) throws {
+        let hardwareFormat = audioEngine.inputNode.inputFormat(forBus: 0)
+        let intervalSamples = AVAudioFrameCount(
+            Double(intervalMilliseconds) / 1000.0 * hardwareFormat.sampleRate
+        )
+        let tapBufferSize = max(intervalSamples, 256)
+
+        audioEngine.inputNode.installTap(onBus: 0, bufferSize: tapBufferSize, format: nil) { [weak self] (buffer, time) in
+            guard let self = self else { return }
+
+            guard buffer.frameLength > 0 else {
+                Logger.debug("Error: received empty buffer in tap callback")
+                self.delegate?.onMicrophoneError("READ_ERROR", "Received empty audio buffer")
+                return
+            }
+
+            self.processAudioBuffer(buffer)
+            self.lastBufferTime = time
+        }
+
+        try audioEngine.start()
+    }
     
     func toggleSilence(isSilent: Bool) {
         Logger.debug("[Microphone] toggleSilence")
@@ -101,6 +209,7 @@ class Microphone {
         recordingSettings = newSettings  // Update the class property with the new settings
 
         self.frequencyBandConfig = frequencyBandConfig
+        self.lastIntervalMs = intervalMilliseconds
         // Analyzer uses the desired (target) sample rate, not hardware rate
         let targetRate = Int(settings.desiredSampleRate ?? settings.sampleRate)
         let fbConfig = frequencyBandConfig ?? (lowCrossoverHz: Float(300), highCrossoverHz: Float(2000))
@@ -110,30 +219,9 @@ class Microphone {
             highCrossoverHz: fbConfig.highCrossoverHz
         )
 
-        // Compute tap buffer size from interval so Core Audio delivers at the right cadence
-        let intervalSamples = AVAudioFrameCount(
-            Double(intervalMilliseconds) / 1000.0 * hardwareFormat.sampleRate
-        )
-        let tapBufferSize = max(intervalSamples, 256) // floor at 256 frames (~5ms at 48kHz)
-
-        // Pass nil for format to use the hardware's native format, avoiding format mismatch crashes.
-        // Core Audio does not support format conversion (e.g. Float32 -> Int16) on the tap itself.
-        audioEngine.inputNode.installTap(onBus: 0, bufferSize: tapBufferSize, format: nil) { [weak self] (buffer, time) in
-            guard let self = self else { return }
-
-            guard buffer.frameLength > 0 else {
-                Logger.debug("Error: received empty buffer in tap callback")
-                self.delegate?.onMicrophoneError("READ_ERROR", "Received empty audio buffer")
-                return
-            }
-
-            self.processAudioBuffer(buffer)
-            self.lastBufferTime = time
-        }
-        
         do {
             startTime = Date()
-            try audioEngine.start()
+            try installTapAndStartEngine(intervalMilliseconds: intervalMilliseconds)
             isRecording = true
             Logger.debug("Debug: Recording started successfully.")
             return StartRecordingResult(
@@ -151,6 +239,9 @@ class Microphone {
     }
     
     public func stopRecording(resolver promise: Promise?) {
+        // Clear resume intent up-front so a stopRecording call during an
+        // active interruption window cancels the auto-resume on .ended.
+        pendingInterruptionResume = false
         guard self.isRecording else {
             if let promiseResolver = promise {
                 promiseResolver.resolve(nil)

package/ios/PipelineIntegration.swift

@@ -20,6 +20,7 @@ class PipelineIntegration: PipelineListener {
     static let EVENT_ZOMBIE_DETECTED     = "PipelineZombieDetected"
     static let EVENT_UNDERRUN            = "PipelineUnderrun"
     static let EVENT_DRAINED             = "PipelineDrained"
+    static let EVENT_PLAYBACK_STOPPED    = "PipelinePlaybackStopped"
     static let EVENT_AUDIO_FOCUS_LOST    = "PipelineAudioFocusLost"
     static let EVENT_AUDIO_FOCUS_RESUMED = "PipelineAudioFocusResumed"
     static let EVENT_FREQUENCY_BANDS    = "PipelineFrequencyBands"
@@ -72,7 +73,7 @@ class PipelineIntegration: PipelineListener {
             sharedEngine: sharedEngine,
             listener: self
         )
-        p.connect()
+        try p.connect()
         pipeline = p
 
         return [
@@ -152,6 +153,11 @@ class PipelineIntegration: PipelineListener {
         return pipeline?.getState().rawValue ?? PipelineState.idle.rawValue
     }
 
+    /// Current platform output latency in milliseconds. Returns 0 if not connected.
+    func outputLatencyMs() -> Double {
+        return pipeline?.outputLatencyMs() ?? 0
+    }
+
     /// Register the pipeline as a delegate on the shared engine.
     /// Called by the module after connect() so route changes and interruptions
     /// are forwarded to the AudioPipeline instance.
@@ -206,6 +212,10 @@ class PipelineIntegration: PipelineListener {
         sendEvent(PipelineIntegration.EVENT_DRAINED, ["turnId": turnId])
     }
 
+    func onPlaybackStopped(turnId: String) {
+        sendEvent(PipelineIntegration.EVENT_PLAYBACK_STOPPED, ["turnId": turnId])
+    }
+
     func onAudioFocusLost() {
         sendEvent(PipelineIntegration.EVENT_AUDIO_FOCUS_LOST, [:])
     }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@edkimmel/expo-audio-stream",
-  "version": "0.4.1",
+  "version": "0.5.0",
   "description": "Expo Play Audio Stream module",
   "main": "build/index.js",
   "types": "build/index.d.ts",
@@ -9,7 +9,6 @@
     "build": "expo-module build",
     "clean": "expo-module clean",
     "lint": "expo-module lint",
-    "test": "expo-module test",
     "prepare": "expo-module prepare && husky || true",
     "prepublishOnly": "expo-module prepublishOnly",
     "expo-module": "expo-module",

package/plugin/build/index.js

@@ -2,12 +2,17 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 const config_plugins_1 = require("@expo/config-plugins");
 const MICROPHONE_USAGE = 'Allow $(PRODUCT_NAME) to access your microphone';
+// AVFoundation enumerates AirPlay/Continuity audio devices on the local network
+// even though we don't use them — without this description, iOS shows a generic
+// prompt and a denial leaves the audio session unable to activate.
+const LOCAL_NETWORK_USAGE = 'Allow $(PRODUCT_NAME) to discover audio devices on your local network';
 const withRecordingPermission = (config, existingPerms) => {
     if (!existingPerms) {
         console.warn('No previous permissions provided');
     }
     config = (0, config_plugins_1.withInfoPlist)(config, (config) => {
         config.modResults['NSMicrophoneUsageDescription'] = config.modResults['NSMicrophoneUsageDescription'] || MICROPHONE_USAGE;
+        config.modResults['NSLocalNetworkUsageDescription'] = config.modResults['NSLocalNetworkUsageDescription'] || LOCAL_NETWORK_USAGE;
         // Add audio to UIBackgroundModes to allow background audio recording
         const existingBackgroundModes = config.modResults.UIBackgroundModes || [];
         if (!existingBackgroundModes.includes('audio')) {

package/plugin/src/index.ts

@@ -6,6 +6,10 @@ import {
 } from '@expo/config-plugins'
 
 const MICROPHONE_USAGE = 'Allow $(PRODUCT_NAME) to access your microphone'
+// AVFoundation enumerates AirPlay/Continuity audio devices on the local network
+// even though we don't use them — without this description, iOS shows a generic
+// prompt and a denial leaves the audio session unable to activate.
+const LOCAL_NETWORK_USAGE = 'Allow $(PRODUCT_NAME) to discover audio devices on your local network'
 
 const withRecordingPermission: ConfigPlugin<{
     microphonePermission: string
@@ -15,6 +19,7 @@ const withRecordingPermission: ConfigPlugin<{
     }
     config = withInfoPlist(config, (config) => {
         config.modResults['NSMicrophoneUsageDescription'] = config.modResults['NSMicrophoneUsageDescription'] || MICROPHONE_USAGE
+        config.modResults['NSLocalNetworkUsageDescription'] = config.modResults['NSLocalNetworkUsageDescription'] || LOCAL_NETWORK_USAGE
 
         // Add audio to UIBackgroundModes to allow background audio recording
         const existingBackgroundModes =

package/src/events.ts

@@ -21,6 +21,10 @@ export interface AudioEventPayload {
   streamUuid: string;
   soundLevel?: number;
   frequencyBands?: { low: number; mid: number; high: number };
+  /** Set by native when a mid-recording error occurs (interruption, read failure).
+   * When present, `encoded` is absent and the recording is no longer active. */
+  error?: string;
+  errorMessage?: string;
 }
 
 export const DeviceReconnectedReasons = {

package/src/index.ts

@@ -59,9 +59,12 @@ export class ExpoPlayAudioStream {
   }> {
     let subscription: Subscription | undefined;
     try {
-      const { onAudioStream, ...options } = recordingConfig;
+      const { onAudioStream, onError, ...options } = recordingConfig;
 
-      if (onAudioStream && typeof onAudioStream == "function") {
+      if (
+        (onAudioStream && typeof onAudioStream == "function") ||
+        (onError && typeof onError == "function")
+      ) {
         subscription = addAudioEventListener(
           async (event: AudioEventPayload) => {
             const {
@@ -72,7 +75,13 @@ export class ExpoPlayAudioStream {
               encoded,
               soundLevel,
               frequencyBands,
+              error,
+              errorMessage,
             } = event;
+            if (error) {
+              onError?.({ code: error, message: errorMessage ?? "" });
+              return;
+            }
             if (!encoded) {
               console.error(
                 `[ExpoPlayAudioStream] Encoded audio data is missing`
@@ -259,6 +268,7 @@ export type {
   PipelineZombieDetectedEvent,
   PipelineUnderrunEvent,
   PipelineDrainedEvent,
+  PipelinePlaybackStoppedEvent,
   PipelineAudioFocusLostEvent,
   PipelineAudioFocusResumedEvent,
 } from "./pipeline";

package/src/pipeline/index.ts

@@ -103,6 +103,22 @@ export class Pipeline {
     return ExpoPlayAudioStreamModule.getPipelineTelemetry() as PipelineTelemetry;
   }
 
+  /**
+   * Query the platform's current output latency — i.e., how long after a
+   * sample is written to the native buffer before it actually leaves the
+   * speaker.
+   *
+   * Value can change mid-session, notably on audio route changes such as
+   * switching from built-in speaker to Bluetooth (Bluetooth typically adds
+   * 100+ ms). **Always query at the moment you care; do not cache.**
+   *
+   * Returns 0 if the pipeline is not connected or the platform cannot
+   * report a value.
+   */
+  static getOutputLatencyMs(): number {
+    return ExpoPlayAudioStreamModule.getPipelineOutputLatencyMs() as number;
+  }
+
   // ════════════════════════════════════════════════════════════════════════
   // Event subscriptions
   // ════════════════════════════════════════════════════════════════════════
@@ -211,6 +227,7 @@ export type {
   PipelineZombieDetectedEvent,
   PipelineUnderrunEvent,
   PipelineDrainedEvent,
+  PipelinePlaybackStoppedEvent,
   PipelineAudioFocusLostEvent,
   PipelineAudioFocusResumedEvent,
 } from './types';

package/src/pipeline/types.ts

@@ -6,6 +6,18 @@ import { PlaybackMode, FrequencyBandConfig, FrequencyBands } from "../types";
 
 // ── Connect ─────────────────────────────────────────────────────────────────
 
+/**
+ * How the pipeline's playback should coexist with other audio on the device.
+ *
+ * - `'mixWithOthers'` (default): plays alongside other apps without
+ *   interrupting them. On Android no audio focus is requested. Best for
+ *   sound effects and short clips.
+ * - `'duckOthers'`: requests audio focus with ducking. Other apps lower
+ *   their volume but keep playing.
+ * - `'doNotMix'`: requests exclusive audio focus. Other apps pause.
+ */
+export type PipelineAudioMode = 'mixWithOthers' | 'duckOthers' | 'doNotMix';
+
 /** Options passed to `connectPipeline()`. */
 export interface ConnectPipelineOptions {
   /** Sample rate in Hz (default 24000). */
@@ -25,6 +37,15 @@ export interface ConnectPipelineOptions {
   frequencyBandIntervalMs?: number;
   /** Optional frequency band crossover configuration. */
   frequencyBandConfig?: FrequencyBandConfig;
+  /**
+   * How pipeline playback should coexist with other apps' audio.
+   * Default is `'mixWithOthers'` (matches expo-audio).
+   *
+   * Note: this is a **behavior change** vs. prior versions of this library,
+   * which effectively used `'doNotMix'`. Pass `'doNotMix'` explicitly to
+   * preserve that old behavior.
+   */
+  audioMode?: PipelineAudioMode;
 }
 
 /** Result returned from a successful `connectPipeline()` call. */
@@ -114,6 +135,20 @@ export interface PipelineDrainedEvent {
   turnId: string;
 }
 
+/**
+ * Payload for `PipelinePlaybackStopped`.
+ *
+ * Fired when the last sample physically leaves the speaker, approximately
+ * `outputLatencyMs` after `PipelineDrained` for the same turn. Pairs with
+ * `PipelinePlaybackStarted` (start-of-emission ↔ end-of-emission).
+ *
+ * Note: this is a physical-world milestone, distinct from `state: 'idle'`
+ * (the pipeline-state-machine value reported via `PipelineStateChanged`).
+ */
+export interface PipelinePlaybackStoppedEvent {
+  turnId: string;
+}
+
 /** Payload for `PipelineAudioFocusLost` (empty — presence is the signal). */
 export type PipelineAudioFocusLostEvent = Record<string, never>;
 
@@ -134,6 +169,7 @@ export interface PipelineEventMap {
   PipelineZombieDetected: PipelineZombieDetectedEvent;
   PipelineUnderrun: PipelineUnderrunEvent;
   PipelineDrained: PipelineDrainedEvent;
+  PipelinePlaybackStopped: PipelinePlaybackStoppedEvent;
   PipelineAudioFocusLost: PipelineAudioFocusLostEvent;
   PipelineAudioFocusResumed: PipelineAudioFocusResumedEvent;
   PipelineFrequencyBands: PipelineFrequencyBandsEvent;

package/src/types.ts

@@ -129,10 +129,19 @@ export interface RecordingConfig {
   enableProcessing?: boolean; // Boolean to enable/disable audio processing (default is false)
   pointsPerSecond?: number; // Number of data points to extract per second of audio (default is 1000)
   onAudioStream?: (event: AudioDataEvent) => Promise<void>; // Callback function to handle audio stream
+  /** Fired when the native layer reports a mid-recording error (e.g. system
+   * interruption like Siri or a phone call). The consumer should treat the
+   * recording session as terminated and clean up. */
+  onError?: (event: MicrophoneErrorEvent) => void;
   /** Optional frequency band crossover configuration. */
   frequencyBandConfig?: FrequencyBandConfig;
 }
 
+export interface MicrophoneErrorEvent {
+  code: string;
+  message: string;
+}
+
 export interface Chunk {
   text: string;
   timestamp: [number, number | null];