Merge pull request #405 from RunanywhereAI/mac-changes

Mac changes (Changes needed to enable MacOS for distribution)

Author: shubhammalhotra28

Package.swift

@@ -37,7 +37,7 @@ import Foundation
 //   ./scripts/build-swift.sh --set-remote  (sets useLocalBinaries = false)
 //
 // =============================================================================
-let useLocalBinaries = false //  Toggle: true for local dev, false for release
+let useLocalBinaries = true //  Toggle: true for local dev, false for release
 
 // Version for remote XCFrameworks (used when testLocal = false)
 // Updated automatically by CI/CD during releases
@@ -131,6 +131,7 @@ let package = Package(
                 .product(name: "Sentry", package: "sentry-cocoa"),
                 .product(name: "StableDiffusion", package: "ml-stable-diffusion"),
                 "CRACommons",
+                "RACommonsBinary",
             ],
             path: "sdk/runanywhere-swift/Sources/RunAnywhere",
             exclude: ["CRACommons"],
@@ -150,6 +151,8 @@ let package = Package(
             dependencies: [
                 "RunAnywhere",
                 "ONNXBackend",
+                "RABackendONNXBinary",
+                "ONNXRuntimeBinary",
             ],
             path: "sdk/runanywhere-swift/Sources/ONNXRuntime",
             exclude: ["include"],
@@ -170,6 +173,7 @@ let package = Package(
             dependencies: [
                 "RunAnywhere",
                 "LlamaCPPBackend",
+                "RABackendLlamaCPPBinary",
             ],
             path: "sdk/runanywhere-swift/Sources/LlamaCPPRuntime",
             exclude: ["include"],

examples/ios/RunAnywhereAI/RunAnywhereAI.xcodeproj/project.pbxproj

@@ -412,6 +412,7 @@
 			isa = XCBuildConfiguration;
 			buildSettings = {
 				ALWAYS_SEARCH_USER_PATHS = NO;
+				ARCHS = arm64;
 				ASSETCATALOG_COMPILER_GENERATE_SWIFT_ASSET_SYMBOL_EXTENSIONS = YES;
 				CLANG_ANALYZER_NONNULL = YES;
 				CLANG_ANALYZER_NUMBER_OBJECT_CONVERSION = YES_AGGRESSIVE;
@@ -465,6 +466,7 @@
 		5480A2132E2F250400337F2F /* Debug */ = {
 			isa = XCBuildConfiguration;
 			buildSettings = {
+				ARCHS = arm64;
 				ASSETCATALOG_COMPILER_APPICON_NAME = AppIcon;
 				ASSETCATALOG_COMPILER_GLOBAL_ACCENT_COLOR_NAME = AccentColor;
 				CODE_SIGN_ENTITLEMENTS = RunAnywhereAI/RunAnywhereAI.entitlements;
@@ -475,6 +477,7 @@
 				ENABLE_PREVIEWS = YES;
 				GENERATE_INFOPLIST_FILE = YES;
 				INFOPLIST_KEY_CFBundleDisplayName = RunAnywhere;
+				INFOPLIST_KEY_LSApplicationCategoryType = "public.app-category.utilities";
 				INFOPLIST_KEY_NSCameraUsageDescription = "RunAnywhere AI needs access to your camera for vision language model features to analyze images.";
 				INFOPLIST_KEY_NSMicrophoneUsageDescription = "RunAnywhere AI needs access to your microphone to transcribe your voice input and provide voice-based interactions.";
 				INFOPLIST_KEY_NSSpeechRecognitionUsageDescription = "RunAnywhere AI uses speech recognition to convert your voice to text for processing.";
@@ -508,6 +511,7 @@
 		5480A2142E2F250400337F2F /* Release */ = {
 			isa = XCBuildConfiguration;
 			buildSettings = {
+				ARCHS = arm64;
 				ASSETCATALOG_COMPILER_APPICON_NAME = AppIcon;
 				ASSETCATALOG_COMPILER_GLOBAL_ACCENT_COLOR_NAME = AccentColor;
 				CODE_SIGN_ENTITLEMENTS = RunAnywhereAI/RunAnywhereAI.entitlements;
@@ -518,6 +522,7 @@
 				ENABLE_PREVIEWS = YES;
 				GENERATE_INFOPLIST_FILE = YES;
 				INFOPLIST_KEY_CFBundleDisplayName = RunAnywhere;
+				INFOPLIST_KEY_LSApplicationCategoryType = "public.app-category.utilities";
 				INFOPLIST_KEY_NSCameraUsageDescription = "RunAnywhere AI needs access to your camera for vision language model features to analyze images.";
 				INFOPLIST_KEY_NSMicrophoneUsageDescription = "RunAnywhere AI needs access to your microphone to transcribe your voice input and provide voice-based interactions.";
 				INFOPLIST_KEY_NSSpeechRecognitionUsageDescription = "RunAnywhere AI uses speech recognition to convert your voice to text for processing.";
@@ -536,6 +541,7 @@
 				"LD_RUNPATH_SEARCH_PATHS[sdk=macosx*]" = "@executable_path/../Frameworks";
 				MACOSX_DEPLOYMENT_TARGET = 15.5;
 				MARKETING_VERSION = 0.17.2;
+				OTHER_LDFLAGS = "-all_load";
 				PRODUCT_BUNDLE_IDENTIFIER = com.runanywhere.RunAnywhere;
 				PRODUCT_NAME = "$(TARGET_NAME)";
 				REGISTER_APP_GROUPS = YES;

sdk/runanywhere-swift/Sources/RunAnywhere/Features/STT/Services/AudioCaptureManager.swift

@@ -84,22 +84,30 @@ public class AudioCaptureManager: ObservableObject {
         }
 
         #if os(iOS) || os(tvOS)
-        // Configure audio session (iOS/tvOS only)
-        // watchOS is NOT supported - AVAudioEngine inputNode tap does not work on watchOS
         let audioSession = AVAudioSession.sharedInstance()
         try audioSession.setCategory(.record, mode: .measurement)
         try audioSession.setActive(true)
         #endif
 
-        // Create audio engine (works on all platforms)
         let engine = AVAudioEngine()
         let inputNode = engine.inputNode
 
-        // Get input format
+        #if os(macOS)
+        // On macOS there is no AVAudioSession. Preparing the engine before
+        // reading the input node format establishes the audio-unit graph
+        // connections and avoids kAudioUnitErr_NoConnection (-10877).
+        engine.prepare()
+        #endif
+
         let inputFormat = inputNode.outputFormat(forBus: 0)
+
+        guard inputFormat.sampleRate > 0, inputFormat.channelCount > 0 else {
+            logger.error("No valid audio input device (sampleRate=\(inputFormat.sampleRate), channels=\(inputFormat.channelCount))")
+            throw AudioCaptureError.noInputDevice
+        }
+
         logger.info("Input format: \(inputFormat.sampleRate) Hz, \(inputFormat.channelCount) channels")
 
-        // Create converter format (16kHz, mono, int16)
         guard let outputFormat = AVAudioFormat(
             commonFormat: .pcmFormatInt16,
             sampleRate: targetSampleRate,
@@ -109,32 +117,26 @@ public class AudioCaptureManager: ObservableObject {
             throw AudioCaptureError.formatConversionFailed
         }
 
-        // Create audio converter
         guard let converter = AVAudioConverter(from: inputFormat, to: outputFormat) else {
             throw AudioCaptureError.formatConversionFailed
         }
 
-        // Install tap on input node
         inputNode.installTap(onBus: 0, bufferSize: 4096, format: inputFormat) { [weak self] buffer, _ in
             guard let self = self else { return }
 
-            // Update audio level for visualization
             self.updateAudioLevel(buffer: buffer)
 
-            // Convert to target format
             guard let convertedBuffer = self.convert(buffer: buffer, using: converter, to: outputFormat) else {
                 return
             }
 
-            // Convert to Data (int16 PCM)
             if let audioData = self.bufferToData(buffer: convertedBuffer) {
                 DispatchQueue.main.async {
                     onAudioData(audioData)
                 }
             }
         }
 
-        // Start engine (remove tap on failure to avoid resource leak)
         do {
             try engine.start()
         } catch {
@@ -177,12 +179,18 @@ public class AudioCaptureManager: ObservableObject {
 
     // MARK: - Private Helpers
 
-    /// Converts a PCM buffer to the target format. Internal for unit testing (converter input block single-use behavior).
+    /// Converts a PCM buffer to the target format. Internal for unit testing.
     internal func convert(
         buffer: AVAudioPCMBuffer,
         using converter: AVAudioConverter,
         to format: AVAudioFormat
     ) -> AVAudioPCMBuffer? {
+        // The input block returns .endOfStream after providing one buffer.
+        // On macOS the converter stays in that "finished" state across calls,
+        // producing empty output for every subsequent buffer. Resetting before
+        // each conversion clears the state so the next buffer is processed.
+        converter.reset()
+
         let capacity = AVAudioFrameCount(Double(buffer.frameLength) * (format.sampleRate / buffer.format.sampleRate))
 
         guard let convertedBuffer = AVAudioPCMBuffer(
@@ -260,6 +268,7 @@ public enum AudioCaptureError: LocalizedError {
     case permissionDenied
     case formatConversionFailed
     case engineStartFailed
+    case noInputDevice
 
     public var errorDescription: String? {
         switch self {
@@ -269,6 +278,8 @@ public enum AudioCaptureError: LocalizedError {
             return "Failed to convert audio format"
         case .engineStartFailed:
             return "Failed to start audio engine"
+        case .noInputDevice:
+            return "No audio input device available"
         }
     }
 }