Swift Package IndexBuild Information
Successful build of Qwen3Speech, reference v0.0.29 (7f382b), with Swift 6.4 for macOS (SPM) on 28 Oct 2026 01:47:16 UTC.
Swift 6 data race errors: 0
Build Command
env DEVELOPER_DIR=/Applications/Xcode-27.0.0.app xcrun swift build --arch arm64Build Log
[2434 / 3341] Cmlx
[2437 / 3341] Cmlx
[2440 / 3341] Cmlx
[2442 / 3341] Cmlx
[2443 / 3341] Cmlx
[2444 / 3341] Cmlx
[2445 / 3341] Cmlx
[2451 / 3341] Cmlx
[2453 / 3341] Cmlx
[2454 / 3341] Cmlx
[2455 / 3341] Cmlx
[2457 / 3341] Cmlx
[2459 / 3341] Cmlx
[2462 / 3341] Cmlx
[2463 / 3341] Cmlx
[2466 / 3341] Cmlx
[2470 / 3341] Cmlx
[2471 / 3341] Cmlx
[2474 / 3341] MLX
[2475 / 3341] MLXRandom
[2476 / 3341] MLX-product
[2479 / 3341] MLX
[2481 / 3341] MLX
[2485 / 3341] Jinja
[2487 / 3341] Jinja
[2491 / 3341] DequeModule
[2492 / 3341] DequeModule-product
[2498 / 3341] mlx-swift_Cmlx
[2503 / 3371]
[2503 / 3372] Hub
[2511 / 3372]
[2513 / 3372]
[2521 / 3374]
[2524 / 3376]
[2526 / 3377] DequeModule
[2529 / 3390] Collections
[2533 / 3418] MLXNN
[2546 / 3421] MLXRandom
[2570 / 3423] Tokenizers
[2581 / 3423]
[2585 / 3423]
[2595 / 3423]
[2601 / 3423] NIOCore
[2604 / 3423]
[2605 / 3423] NIOCore
[2608 / 3423] Cmlx
[2611 / 3423] NIOCore
[2612 / 3423] NIOCore
[2612 / 3423] MLXRandom
[2615 / 3423] MLXRandom-product
[2617 / 3423] MLXRandom-product
[2622 / 3423] MLXNN
[2624 / 3423] MLXFast
[2628 / 3423] Collections-product
[2634 / 3426] MLXFFT
[2639 / 3427] AsyncAlgorithms
[2643 / 3438] AsyncAlgorithms
[2644 / 3441] MLXOptimizers
[2647 / 3471]
[2657 / 3471] Generation
[2667 / 3471] MLXLMCommon
[2679 / 3471] Models
[2692 / 3471]
[2701 / 3472]
[2706 / 3472]
[2709 / 3484] ServiceLifecycle
[2713 / 3484] ServiceLifecycle
[2717 / 3484] ServiceLifecycle
[2731 / 3501] NIOHTTPTypes
[2744 / 3501] AsyncAlgorithms
[2752 / 3501] NIOPosix
[2753 / 3501] NIOPosix
[2755 / 3501] ServiceLifecycle
[2759 / 3501] NIOPosix
[2760 / 3501] NIOHTTPTypes
[2761 / 3501] AudioCommon
[2769 / 3501] AudioCommon-product
[2772 / 3501] NIOPosix
[2776 / 3502]
[2778 / 3504]
[2780 / 3509] SpeechCore
[2782 / 3521] SupertonicTTS
[2784 / 3533] PocketTTS
[2785 / 3533] SpeechRestoration
[2801 / 3545] WhisperASR
[2806 / 3559] Qwen3TTSCoreML
[2811 / 3565] SpeechCore
[2838 / 3565] SpeechCore
[2845 / 3565] SpeechRestoration
[2850 / 3565] SpeechWakeWord
[2866 / 3565] SpeechWakeWord-product
[2871 / 3565] SpeechRestoration
[2873 / 3565] SpeechRestoration-product
[2881 / 3565] Qwen3TTSCoreML
[2886 / 3565] PocketTTS-product
[2891 / 3565] NIOFoundationEssentialsCompat
[2896 / 3567] AudioCommon
[2900 / 3571] NIOPosix
[2904 / 3575] FunctionGemma
[2904 / 3579] ParakeetASR
[2906 / 3588] CanaryASR
[2912 / 3602] NIOFoundationEssentialsCompat
[2917 / 3629] FunctionGemma
[2927 / 3646] Models
[2956 / 3646] MLXCommon
[2971 / 3653] CanaryASR
[2976 / 3653] MLXCommon
[2990 / 3653] ParakeetASR
[3001 / 3653] MossTTSNano-product
[3005 / 3653] Models
[3012 / 3653] MLXLMCommon
[3017 / 3653] MLXLMCommon-product
[3023 / 3653] Models
[3027 / 3653] VoxtralTTS
[3030 / 3653] MLXCommon
[3036 / 3653] KokoroTTS
[3038 / 3653] KokoroTTS
[3041 / 3653] KokoroTTS
[3045 / 3653] IndicMioTTS
[3050 / 3653] FlashSR
[3055 / 3653] CanaryASR-product
[3061 / 3655]
[3064 / 3659] VibeVoiceTTS
[3065 / 3659] SpeechEnhancement
[3065 / 3659]
[3068 / 3665] SpeechVAD
[3071 / 3676] pocket-tts-bench-product
[3073 / 3690] VoxCPM2TTS
[3075 / 3713] StableAudio3MusicGen
[3079 / 3724] PersonaPlex
[3080 / 3729] VoxtralTTS
[3096 / 3740] SpeechEnhancement
[3096 / 3740] pocket-tts-bench-product
[3107 / 3751] SpeechLanguageID
[3132 / 3762] VoxtralTTS-product
[3135 / 3773] PersonaPlex
[3151 / 3773] VoxtralASR
[3156 / 3773] VoxtralASR
[3159 / 3773] VoxtralTTS
[3174 / 3773] VoxCPM2TTS-product
[3179 / 3773] VibeVoiceTTS
[3182 / 3773] VoxtralASR
[3185 / 3773] VibeVoiceTTS
[3188 / 3773] StableAudio3MusicGen
[3190 / 3773] StableAudio3MusicGen-product
[3196 / 3773] StableAudio3MusicGen
[3198 / 3773] SpeechVAD
[3202 / 3773] SpeechVAD
[3205 / 3773] VibeVoiceTTS
[3209 / 3773] VoxCPM2TTS
[3211 / 3773] SpeechLanguageID-product
[3216 / 3773] SpeechEnhancement
[3219 / 3773] SpeechEnhancement
[3223 / 3773] SpeechEnhancement
[3227 / 3773] SourceSeparation
[3229 / 3773] SourceSeparation
[3233 / 3773] SourceSeparation
[3240 / 3773] Qwen3TTS
[3244 / 3773] Qwen3TTS
[3250 / 3773] SourceSeparation
[3254 / 3773] PersonaPlex-product
[3257 / 3773] PersonaPlex
[3260 / 3774]
[3265 / 3779] MossTranscribe
[3265 / 3779] NIOHTTPTypes-product
[3267 / 3792] Models
[3269 / 3794] MLXLLM
[3269 / 3794] PersonaPlex
[3272 / 3794] NIO
[3279 / 3805] Qwen3ASR
[3284 / 3859] MossTranscribe
[3285 / 3867] MagpieTTS
[3286 / 3878] MLXLLM
[3298 / 3885] OmniVoiceTTS
[3315 / 3907] NemotronStreamingASR
[3331 / 3907] CosyVoiceTTS
[3331 / 3907] MADLADTranslation
[3345 / 3907] ChatterboxTTS
[3360 / 3907] Qwen3ASR
[3361 / 3912] GLiNER
[3362 / 3921] FlashSR
[3369 / 3959] GLiNER
[3390 / 3970] ChatterboxTTS
[3391 / 3970] Qwen3ASRRuntime
[3409 / 3970] FishAudioTTS
[3417 / 3970] OmnilingualASR
[3427 / 3970] OmnilingualASR-product
[3430 / 3970] MLXLLM
[3443 / 3970] OmniVoiceTTS
[3444 / 3970] OmniVoiceTTS
[3454 / 3970] NemotronStreamingASR
[3456 / 3970] NIO
[3463 / 3970] NIO
[3466 / 3970] MossTranscribe-product
[3468 / 3970] Models
[3474 / 3970] MagpieTTSCoreML
[3479 / 3970] MagpieTTS
[3483 / 3970] MLXLLM
[3485 / 3970] MLXLLM-product
[3490 / 3970] MLXLLM
[3491 / 3970] MagpieTTS
[3493 / 3970] MLX
[3496 / 3970] MAGNeTMusicGen
[3498 / 3970] MAGNeTMusicGen-product
[3502 / 3970] MADLADTranslation
[3508 / 3970] IndicMioTTS
[3511 / 3970] IndicMioTTS
[3514 / 3970] IndicMioTTS-product
[3519 / 3970] HiggsTTS
[3525 / 3970] HiggsTTS
[3528 / 3970] HiggsTTS
[3533 / 3970] gliner-bench-product
[3537 / 3970] FlashSR-product
[3543 / 3970] HiggsTTS
[3548 / 3970] GLiNER
[3553 / 3970] FishAudioTTS
[3555 / 3970] F5TTS-product
[3559 / 3970] F5TTS
[3564 / 3970] CosyVoiceTTS
[3567 / 3970] CosyVoiceTTS
[3572 / 3970] CosyVoiceTTS
[3574 / 3970] CohereTranscribeASR
[3576 / 3970] CohereTranscribeASR-product
[3581 / 3970] ChatterboxTTS
[3584 / 3970] ChatterboxTTS
[3588 / 3970] ChatterboxTTS
[3592 / 3970]
[3597 / 3972]
[3599 / 3974] NIOEmbedded
[3603 / 3993] VoiceChat
[3604 / 3996] gliner-bench-product
[3608 / 4008] Audio2Face3D
[3618 / 4008] gliner-bench-product
[3621 / 4008] VoiceChat
[3625 / 4018] VoiceChat
[3641 / 4018] gliner-bench-product
[3647 / 4018] Qwen3Chat
[3658 / 4018] Qwen3Chat-product
[3661 / 4018] Qwen3Chat
[3665 / 4018] NIO
[3669 / 4018] asr-bench-product
[3673 / 4018] KyutaiSTT
[3677 / 4018] Audio2Face3D-product
[3684 / 4019]
[3689 / 4028] NIOFoundationCompat
[3691 / 4028] HibikiTranslate
[3693 / 4046] CSM
[3697 / 4064] Qwen3Chat
[3700 / 4064] NIOTLS
[3707 / 4075] voicechat-bench-product
[3710 / 4082] CSM
[3721 / 4088] asr-bench-product
[3732 / 4098] CSM
[3758 / 4098] asr-bench-product
[3761 / 4098] BreezeTTS
[3776 / 4098] Qwen3TTSCoreML
[3783 / 4098] ParakeetASR
[3788 / 4098] NIOWebSocket
[3792 / 4098] NIOHTTP1
[3797 / 4098] NIOFoundationCompat-product
[3801 / 4098] MagpieTTSCoreML
[3809 / 4098] NIOHTTP1
[3810 / 4098] NIOHTTP1
[3817 / 4098] FunctionGemma
[3822 / 4098] CanaryASR
[3828 / 4098] BreezeTTS-product
[3835 / 4098] BenchmarkSupport
[3845 / 4101]
[3848 / 4101] SpeechRestoration-product
[3857 / 4102] MLXFast
[3859 / 4102] MLXFFT
[3862 / 4105] NIOExtras
[3865 / 4116] IndexTTS2TTS
[3869 / 4132]
[3876 / 4158] diarization-bench-product
[3892 / 4158] NIOWebSocket
[3903 / 4158] NIOTransportServices
[3919 / 4158] vad-bench-product
[3925 / 4158] NIOTransportServices
[3928 / 4158] NIOTransportServices-product
[3935 / 4158] NIOExtras
[3942 / 4158] NIOTransportServices
[3951 / 4158] IndexTTS2TTS
[3952 / 4158] HummingbirdCore
[3956 / 4160]
[3962 / 4160] NIOHTTP1-product
[3969 / 4161] KokoroTTS-product
[3972 / 4161] VoxtralASR
[3982 / 4172] Qwen3TTS
[3996 / 4172] CosyVoiceTTS
[4007 / 4172] WSCore
[4012 / 4189] WSCore
[4021 / 4189] WSCore-product
[4025 / 4189] Clef
[4035 / 4189] HummingbirdCore
[4045 / 4189] Hummingbird
[4054 / 4189]
[4058 / 4190] IndicMioTTS
[4066 / 4191] Hummingbird
[4071 / 4191] HummingbirdCore
[4081 / 4192] OmniVoiceTTS-product
[4087 / 4192] Qwen3ASR-product
[4091 / 4192] HummingbirdCore
[4101 / 4192] GLiNER-product
[4103 / 4192] MagpieTTSCoreML-product
[4108 / 4192] ChatterboxTTS-product
[4114 / 4194] clef-decide-product
[4119 / 4194] diarization-bench-product
[4123 / 4205] clef-decide-product
[4127 / 4216] Hummingbird
[4131 / 4216] clef-decide-product
[4133 / 4216] Hummingbird
[4137 / 4216] AudioCLILib
[4139 / 4216] AudioCLILib
[4152 / 4216] clef-decide-product
[4154 / 4216] AudioCLILib
[4156 / 4216] AudioCLILib
[4158 / 4216] speech-product
[4165 / 4217]
[4169 / 4218]
[4175 / 4219]
[4177 / 4219] HummingbirdWebSocket
[4181 / 4226] HummingbirdWebSocket
[4186 / 4228] HummingbirdWebSocket
[4195 / 4231] audio-product
[4196 / 4231] speech-product
[4197 / 4231] AudioCLILib
[4201 / 4231] AudioCLILib
[4201 / 4231] AudioServer
[4217 / 4247] AudioServer
[4229 / 4249] AudioServer
[4229 / 4249] speech-server-product
[4233 / 4253] audio-server-product
[4235 / 4253] AudioCLILib
[4237 / 4253] audio-server-product
[4241 / 4253] speech-product
[4243 / 4253] speech-product
[4245 / 4253] AudioServer
[4249 / 4253] speech-server-product
[4251 / 4253] speech-server-product
Build complete! (108.36 sec)
warning: missing creator for mutated node: ('/Users/admin/builder/spi-builder-workspace/.build/out/Products/Debug/mlx-swift_Cmlx.bundle/Contents/MacOS')
In file included from /Users/admin/builder/spi-builder-workspace/.build/checkouts/mlx-swift/Source/Cmlx/mlx-generated/metal/steel/attn/kernels/steel_attention.metal:6:
In file included from /Users/admin/builder/spi-builder-workspace/.build/checkouts/mlx-swift/Source/Cmlx/mlx-generated/metal/steel/attn/kernels/../../../steel/attn/kernels/steel_attention.h:3:
In file included from /Users/admin/builder/spi-builder-workspace/.build/checkouts/mlx-swift/Source/Cmlx/mlx-generated/metal/steel/attn/kernels/../../../steel/attn/attn.h:6:
In file included from /Users/admin/builder/spi-builder-workspace/.build/checkouts/mlx-swift/Source/Cmlx/mlx-generated/metal/steel/attn/kernels/../../../steel/attn/../../steel/attn/mma.h:11:
/Users/admin/builder/spi-builder-workspace/.build/checkouts/mlx-swift/Source/Cmlx/mlx-generated/metal/steel/attn/kernels/../../../steel/attn/../../steel/utils/integral_constant.h:108:6: warning: constexpr if is a C++17 extension [-Wc++17-extensions]
if constexpr (start < stop) {
^
In file included from /Users/admin/builder/spi-builder-workspace/.build/checkouts/mlx-swift/Source/Cmlx/mlx-generated/metal/steel/attn/kernels/steel_attention.metal:6:
/Users/admin/builder/spi-builder-workspace/.build/checkouts/mlx-swift/Source/Cmlx/mlx-generated/metal/steel/attn/kernels/../../../steel/attn/kernels/steel_attention.h:356:16: warning: constexpr if is a C++17 extension [-Wc++17-extensions]
if constexpr (is_bool) {
^
/Users/admin/builder/spi-builder-workspace/.build/checkouts/mlx-swift/Source/Cmlx/mlx-generated/metal/steel/attn/kernels/../../../steel/attn/kernels/steel_attention.h:426:14: warning: constexpr if is a C++17 extension [-Wc++17-extensions]
if constexpr (BD == 128) {
^
/Users/admin/builder/spi-builder-workspace/.build/checkouts/mlx-swift/Source/Cmlx/mlx-generated/metal/steel/attn/kernels/../../../steel/attn/kernels/steel_attention.h:436:14: warning: constexpr if is a C++17 extension [-Wc++17-extensions]
if constexpr (BD == 128) {
^
4 warnings generated.
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/ModelLoader.swift:6:16: [1;33mwarning: [1;39mstored property 'vad' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any StreamingVADProvider'; this is an error in the Swift 6 language mode[0;0m
[0;36m4 |[0;0m /// Loaded model set — holds references to all loaded models.
[0;36m5 |[0;0m public struct ModelSet: Sendable {
[0;36m6 |[0;0m public let vad: (any StreamingVADProvider)?
[0;36m|[0;0m `- [1;33mwarning: [1;39mstored property 'vad' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any StreamingVADProvider'; this is an error in the Swift 6 language mode[0;0m
[0;36m7 |[0;0m public let stt: (any SpeechRecognitionModel)?
[0;36m8 |[0;0m public let tts: (any SpeechGenerationModel)?
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/Protocols.swift:377:17: [1;39mnote: [1;39mprotocol 'StreamingVADProvider' does not conform to the 'Sendable' protocol[0;0m
[0;36m375 |[0;0m ///
[0;36m376 |[0;0m /// Maps directly to speech-core's `sc_vad_vtable_t` for pipeline integration.
[0;36m377 |[0;0m public protocol StreamingVADProvider: AnyObject {
[0;36m|[0;0m `- [1;39mnote: [1;39mprotocol 'StreamingVADProvider' does not conform to the 'Sendable' protocol[0;0m
[0;36m378 |[0;0m /// Expected input sample rate in Hz
[0;36m379 |[0;0m var inputSampleRate: Int { get }
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/ModelLoader.swift:7:16: [1;33mwarning: [1;39mstored property 'stt' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any SpeechRecognitionModel'; this is an error in the Swift 6 language mode[0;0m
[0;36m5 |[0;0m public struct ModelSet: Sendable {
[0;36m6 |[0;0m public let vad: (any StreamingVADProvider)?
[0;36m7 |[0;0m public let stt: (any SpeechRecognitionModel)?
[0;36m|[0;0m `- [1;33mwarning: [1;39mstored property 'stt' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any SpeechRecognitionModel'; this is an error in the Swift 6 language mode[0;0m
[0;36m8 |[0;0m public let tts: (any SpeechGenerationModel)?
[0;36m9 |[0;0m
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/Protocols.swift:222:17: [1;39mnote: [1;39mprotocol 'SpeechRecognitionModel' does not conform to the 'Sendable' protocol[0;0m
[0;36m220 |[0;0m
[0;36m221 |[0;0m /// A speech-to-text model that transcribes audio.
[0;36m222 |[0;0m public protocol SpeechRecognitionModel: AnyObject {
[0;36m|[0;0m `- [1;39mnote: [1;39mprotocol 'SpeechRecognitionModel' does not conform to the 'Sendable' protocol[0;0m
[0;36m223 |[0;0m /// Expected input sample rate in Hz
[0;36m224 |[0;0m var inputSampleRate: Int { get }
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/ModelLoader.swift:8:16: [1;33mwarning: [1;39mstored property 'tts' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any SpeechGenerationModel'; this is an error in the Swift 6 language mode[0;0m
[0;36m6 |[0;0m public let vad: (any StreamingVADProvider)?
[0;36m7 |[0;0m public let stt: (any SpeechRecognitionModel)?
[0;36m8 |[0;0m public let tts: (any SpeechGenerationModel)?
[0;36m|[0;0m `- [1;33mwarning: [1;39mstored property 'tts' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any SpeechGenerationModel'; this is an error in the Swift 6 language mode[0;0m
[0;36m9 |[0;0m
[0;36m10 |[0;0m public init(
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/Protocols.swift:127:17: [1;39mnote: [1;39mprotocol 'SpeechGenerationModel' does not conform to the 'Sendable' protocol[0;0m
[0;36m125 |[0;0m
[0;36m126 |[0;0m /// A text-to-speech model that generates audio from text.
[0;36m127 |[0;0m public protocol SpeechGenerationModel: AnyObject {
[0;36m|[0;0m `- [1;39mnote: [1;39mprotocol 'SpeechGenerationModel' does not conform to the 'Sendable' protocol[0;0m
[0;36m128 |[0;0m /// Output sample rate in Hz
[0;36m129 |[0;0m var sampleRate: Int { get }
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/ModelLoader.swift:6:16: [1;33mwarning: [1;39mstored property 'vad' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any StreamingVADProvider'; this is an error in the Swift 6 language mode[0;0m
[0;36m4 |[0;0m /// Loaded model set — holds references to all loaded models.
[0;36m5 |[0;0m public struct ModelSet: Sendable {
[0;36m6 |[0;0m public let vad: (any StreamingVADProvider)?
[0;36m|[0;0m `- [1;33mwarning: [1;39mstored property 'vad' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any StreamingVADProvider'; this is an error in the Swift 6 language mode[0;0m
[0;36m7 |[0;0m public let stt: (any SpeechRecognitionModel)?
[0;36m8 |[0;0m public let tts: (any SpeechGenerationModel)?
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/Protocols.swift:377:17: [1;39mnote: [1;39mprotocol 'StreamingVADProvider' does not conform to the 'Sendable' protocol[0;0m
[0;36m375 |[0;0m ///
[0;36m376 |[0;0m /// Maps directly to speech-core's `sc_vad_vtable_t` for pipeline integration.
[0;36m377 |[0;0m public protocol StreamingVADProvider: AnyObject {
[0;36m|[0;0m `- [1;39mnote: [1;39mprotocol 'StreamingVADProvider' does not conform to the 'Sendable' protocol[0;0m
[0;36m378 |[0;0m /// Expected input sample rate in Hz
[0;36m379 |[0;0m var inputSampleRate: Int { get }
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/ModelLoader.swift:7:16: [1;33mwarning: [1;39mstored property 'stt' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any SpeechRecognitionModel'; this is an error in the Swift 6 language mode[0;0m
[0;36m5 |[0;0m public struct ModelSet: Sendable {
[0;36m6 |[0;0m public let vad: (any StreamingVADProvider)?
[0;36m7 |[0;0m public let stt: (any SpeechRecognitionModel)?
[0;36m|[0;0m `- [1;33mwarning: [1;39mstored property 'stt' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any SpeechRecognitionModel'; this is an error in the Swift 6 language mode[0;0m
[0;36m8 |[0;0m public let tts: (any SpeechGenerationModel)?
[0;36m9 |[0;0m
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/Protocols.swift:222:17: [1;39mnote: [1;39mprotocol 'SpeechRecognitionModel' does not conform to the 'Sendable' protocol[0;0m
[0;36m220 |[0;0m
[0;36m221 |[0;0m /// A speech-to-text model that transcribes audio.
[0;36m222 |[0;0m public protocol SpeechRecognitionModel: AnyObject {
[0;36m|[0;0m `- [1;39mnote: [1;39mprotocol 'SpeechRecognitionModel' does not conform to the 'Sendable' protocol[0;0m
[0;36m223 |[0;0m /// Expected input sample rate in Hz
[0;36m224 |[0;0m var inputSampleRate: Int { get }
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/ModelLoader.swift:8:16: [1;33mwarning: [1;39mstored property 'tts' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any SpeechGenerationModel'; this is an error in the Swift 6 language mode[0;0m
[0;36m6 |[0;0m public let vad: (any StreamingVADProvider)?
[0;36m7 |[0;0m public let stt: (any SpeechRecognitionModel)?
[0;36m8 |[0;0m public let tts: (any SpeechGenerationModel)?
[0;36m|[0;0m `- [1;33mwarning: [1;39mstored property 'tts' of 'Sendable'-conforming struct 'ModelSet' contains non-Sendable type 'any SpeechGenerationModel'; this is an error in the Swift 6 language mode[0;0m
[0;36m9 |[0;0m
[0;36m10 |[0;0m public init(
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/Protocols.swift:127:17: [1;39mnote: [1;39mprotocol 'SpeechGenerationModel' does not conform to the 'Sendable' protocol[0;0m
[0;36m125 |[0;0m
[0;36m126 |[0;0m /// A text-to-speech model that generates audio from text.
[0;36m127 |[0;0m public protocol SpeechGenerationModel: AnyObject {
[0;36m|[0;0m `- [1;39mnote: [1;39mprotocol 'SpeechGenerationModel' does not conform to the 'Sendable' protocol[0;0m
[0;36m128 |[0;0m /// Output sample rate in Hz
[0;36m129 |[0;0m var sampleRate: Int { get }
/Users/admin/builder/spi-builder-workspace/Sources/AudioCommon/ModelLoader.swift:123:14: [1;33mwarning: [1;39mimmutable value 'name' was never used; consider replacing with '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m121 |[0;0m var tts: (any SpeechGenerationModel)?
[0;36m122 |[0;0m
[0;36m123 |[0;0m for (name, model) in results {
[0;36m|[0;0m `- [1;33mwarning: [1;39mimmutable value 'name' was never used; consider replacing with '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m124 |[0;0m if let m = model as? any StreamingVADProvider { vad = m }
[0;36m125 |[0;0m if let m = model as? any SpeechRecognitionModel { stt = m }
[#NoUsage]: <https://docs.swift.org/compiler/documentation/diagnostics/no-usage>
/Users/admin/builder/spi-builder-workspace/Sources/SupertonicTTS/SupertonicGraphs.swift:28:9: [1;33mwarning: [1;39mswitch must be exhaustive[0;0m
[0;36m26 |[0;0m let n = arr.count
[0;36m27 |[0;0m var out = [Float](repeating: 0, count: n)
[0;36m28 |[0;0m switch arr.dataType {
[0;36m|[0;0m |- [1;33mwarning: [1;39mswitch must be exhaustive[0;0m
[0;36m|[0;0m `- [1;39mnote: [1;39madd missing case: '.int8'[0;0m
[0;36m29 |[0;0m case .float32:
[0;36m30 |[0;0m let p = arr.dataPointer.bindMemory(to: Float.self, capacity: n)
/Users/admin/builder/spi-builder-workspace/Sources/Qwen3TTSCoreML/EmbedderModels.swift:100:9: [1;33mwarning: [1;39minitialization of immutable value 'isContiguous' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m98 |[0;0m // Check if strides indicate non-contiguous layout
[0;36m99 |[0;0m let strides = array.strides.map { $0.intValue }
[0;36m100 |[0;0m let isContiguous = strides.last == 1 && (strides.count < 2 || strides[strides.count - 2] <= 1
[0;36m|[0;0m `- [1;33mwarning: [1;39minitialization of immutable value 'isContiguous' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m101 |[0;0m || (strides.count >= 1 && strides[0] == 1))
[0;36m102 |[0;0m
[#NoUsage]: <https://docs.swift.org/compiler/documentation/diagnostics/no-usage>
/Users/admin/builder/spi-builder-workspace/Sources/SpeechCore/VoicePipeline.swift:503:83: [1;33mwarning: [1;39mcapture of 'chunkCtx' with non-Sendable type 'UnsafeMutableRawPointer?' in an isolated closure[0;0m
[0;36m501 |[0;0m let isFinal = chunk.isFinal
[0;36m502 |[0;0m chunk.samples.withUnsafeBufferPointer { buf in
[0;36m503 |[0;0m onChunk?(buf.baseAddress, buf.count, isFinal, chunkCtx)
[0;36m|[0;0m `- [1;33mwarning: [1;39mcapture of 'chunkCtx' with non-Sendable type 'UnsafeMutableRawPointer?' in an isolated closure[0;0m
[0;36m504 |[0;0m }
[0;36m505 |[0;0m if isFinal { sentFinal = true; break }
Swift.Optional:1:21: [1;39mnote: [1;39mgeneric enum 'Optional' does not conform to the 'Sendable' protocol[0;0m
[0;36m1 |[0;0m @frozen public enum Optional<Wrapped> : ~Copyable, ~Escapable where Wrapped : ~Copyable, Wrapped : ~Escapable {
[0;36m|[0;0m `- [1;39mnote: [1;39mgeneric enum 'Optional' does not conform to the 'Sendable' protocol[0;0m
[0;36m2 |[0;0m case none
[0;36m3 |[0;0m case some(Wrapped)
/Users/admin/builder/spi-builder-workspace/Sources/SpeechCore/VoicePipeline.swift:507:47: [1;33mwarning: [1;39mimplicit capture of 'bridge' requires that 'TTSBridge' conforms to 'Sendable'[0;0m
[0;36m102 |[0;0m
[0;36m103 |[0;0m /// Bridges a SpeechGenerationModel to the C vtable.
[0;36m104 |[0;0m private final class TTSBridge {
[0;36m|[0;0m `- [1;39mnote: [1;39mclass 'TTSBridge' does not conform to the 'Sendable' protocol[0;0m
[0;36m105 |[0;0m let model: SpeechGenerationModel
[0;36m106 |[0;0m private let _cancelled = OSAllocatedUnfairLock(initialState: false)
[0;36m:[0;0m
[0;36m505 |[0;0m if isFinal { sentFinal = true; break }
[0;36m506 |[0;0m }
[0;36m507 |[0;0m if !sentFinal && !bridge.cancelled {
[0;36m|[0;0m `- [1;33mwarning: [1;39mimplicit capture of 'bridge' requires that 'TTSBridge' conforms to 'Sendable'[0;0m
[0;36m508 |[0;0m onChunk?(nil, 0, true, chunkCtx)
[0;36m509 |[0;0m }
/Users/admin/builder/spi-builder-workspace/Sources/Qwen3TTSCoreML/MultiCodeDecoderCoreML.swift:51:17: [1;33mwarning: [1;39mvariable 'kc0' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m49 |[0;0m
[0;36m50 |[0;0m // Position 0: feed hidden_states from CodeDecoder
[0;36m51 |[0;0m var (_, kc0, vc0) = try step(
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'kc0' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m52 |[0;0m embed: ensureNCHW(hiddenState, channels: inputWidth),
[0;36m53 |[0;0m position: 0, keyCache: keyCache, valueCache: valueCache, state: mlState)
/Users/admin/builder/spi-builder-workspace/Sources/Qwen3TTSCoreML/MultiCodeDecoderCoreML.swift:51:22: [1;33mwarning: [1;39mvariable 'vc0' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m49 |[0;0m
[0;36m50 |[0;0m // Position 0: feed hidden_states from CodeDecoder
[0;36m51 |[0;0m var (_, kc0, vc0) = try step(
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'vc0' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m52 |[0;0m embed: ensureNCHW(hiddenState, channels: inputWidth),
[0;36m53 |[0;0m position: 0, keyCache: keyCache, valueCache: valueCache, state: mlState)
[#VariableNeverMutated]: <https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated>
/Users/admin/builder/spi-builder-workspace/Sources/MLXCommon/PreQuantizedEmbedding.swift:47:9: [1;33mwarning: [1;39m'quantizedMatmul(_:_:scales:biases:transpose:groupSize:bits:mode:stream:)' is deprecated: renamed to 'quantizedMM(_:_:scales:biases:transpose:groupSize:bits:mode:stream:)'[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration\DeprecatedDeclaration]8;;\]
[0;36m45 |[0;0m /// For use as LM head (matmul with transposed weight)
[0;36m46 |[0;0m public func asLinear(_ x: MLXArray) -> MLXArray {
[0;36m47 |[0;0m quantizedMatmul(
[0;36m|[0;0m |- [1;33mwarning: [1;39m'quantizedMatmul(_:_:scales:biases:transpose:groupSize:bits:mode:stream:)' is deprecated: renamed to 'quantizedMM(_:_:scales:biases:transpose:groupSize:bits:mode:stream:)'[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration\DeprecatedDeclaration]8;;\]
[0;36m|[0;0m `- [1;39mnote: [1;39muse 'quantizedMM(_:_:scales:biases:transpose:groupSize:bits:mode:stream:)' instead[0;0m
[0;36m48 |[0;0m x, weight, scales: scales, biases: biases, transpose: true,
[0;36m49 |[0;0m groupSize: groupSize, bits: bits)
[#DeprecatedDeclaration]: <https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration>
/Users/admin/builder/spi-builder-workspace/Sources/VibeVoiceTTS/Inference/WeightLoader.swift:229:13: [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m227 |[0;0m if let v = weights["\(prefix).downsample_layers.\(idx).conv.biases"] { w["biases"] = v }
[0;36m228 |[0;0m if !w.isEmpty {
[0;36m229 |[0;0m [4;39mtry? sconv.conv.update(parameters: ModuleParameters.unflattened(w), verify: .none)[0;0m
[0;36m|[0;0m `- [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m230 |[0;0m }
[0;36m231 |[0;0m }
/Users/admin/builder/spi-builder-workspace/Sources/VibeVoiceTTS/Inference/WeightLoader.swift:239:9: [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m237 |[0;0m if let v = weights["\(prefix).head.conv.biases"] { hw["biases"] = v }
[0;36m238 |[0;0m if !hw.isEmpty {
[0;36m239 |[0;0m [4;39mtry? encoder.head.conv.update(parameters: ModuleParameters.unflattened(hw), verify: .none)[0;0m
[0;36m|[0;0m `- [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m240 |[0;0m }
[0;36m241 |[0;0m // 3) normFinal — only if it's a ConvRMSNorm (not Identity).
/Users/admin/builder/spi-builder-workspace/Sources/VibeVoiceTTS/Inference/WeightLoader.swift:245:9: [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m243 |[0;0m let nw = weights["\(prefix).normFinal.weight"] {
[0;36m244 |[0;0m let params = ModuleParameters.unflattened(["weight": nw])
[0;36m245 |[0;0m [4;39mtry? norm.update(parameters: params, verify: .none)[0;0m
[0;36m|[0;0m `- [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m246 |[0;0m }
[0;36m247 |[0;0m // 4) stages — nested [[Block1D]].
/Users/admin/builder/spi-builder-workspace/Sources/VibeVoiceTTS/Inference/WeightLoader.swift:263:13: [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m261 |[0;0m if let normWeight = weights["\(prefix).norm.weight"] {
[0;36m262 |[0;0m let params = ModuleParameters.unflattened(["weight": normWeight])
[0;36m263 |[0;0m [4;39mtry? block.norm.update(parameters: params, verify: .none)[0;0m
[0;36m|[0;0m `- [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m264 |[0;0m }
[0;36m265 |[0;0m
/Users/admin/builder/spi-builder-workspace/Sources/VibeVoiceTTS/Inference/WeightLoader.swift:268:13: [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m266 |[0;0m if let ffnNormWeight = weights["\(prefix).ffnNorm.weight"] {
[0;36m267 |[0;0m let params = ModuleParameters.unflattened(["weight": ffnNormWeight])
[0;36m268 |[0;0m [4;39mtry? block.ffnNorm.update(parameters: params, verify: .none)[0;0m
[0;36m|[0;0m `- [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m269 |[0;0m }
[0;36m270 |[0;0m
/Users/admin/builder/spi-builder-workspace/Sources/VibeVoiceTTS/Inference/WeightLoader.swift:276:13: [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m274 |[0;0m if !mixerWeights.isEmpty {
[0;36m275 |[0;0m let params = ModuleParameters.unflattened(mixerWeights)
[0;36m276 |[0;0m [4;39mtry? block.mixer.conv.update(parameters: params, verify: .none)[0;0m
[0;36m|[0;0m `- [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m277 |[0;0m }
[0;36m278 |[0;0m
/Users/admin/builder/spi-builder-workspace/Sources/VibeVoiceTTS/Inference/WeightLoader.swift:324:9: [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m322 |[0;0m if !headWeights.isEmpty {
[0;36m323 |[0;0m let params = ModuleParameters.unflattened(headWeights)
[0;36m324 |[0;0m [4;39mtry? decoder.head.conv.update(parameters: params, verify: .none)[0;0m
[0;36m|[0;0m `- [1;33mwarning: [1;39mresult of 'try?' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m325 |[0;0m }
[0;36m326 |[0;0m }
[#NoUsage]: <https://docs.swift.org/compiler/documentation/diagnostics/no-usage>
/Users/admin/builder/spi-builder-workspace/Sources/SpeechVAD/SortformerDiarizer.swift:271:26: [1;33mwarning: [1;39mresult of call to 'withUnsafeMutableBufferPointer' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m269 |[0;0m let elements = framesToCopy * nMels
[0;36m270 |[0;0m let srcStart = chunkStart * nMels
[0;36m271 |[0;0m chunkMel.withUnsafeMutableBufferPointer [4;39m{ dst in[0;0m
[0;36m|[0;0m `- [1;33mwarning: [1;39mresult of call to 'withUnsafeMutableBufferPointer' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m272 |[0;0m melSpec.withUnsafeBufferPointer { src in
[0;36m273 |[0;0m memcpy(dst.baseAddress!, src.baseAddress! + srcStart,
[#NoUsage]: <https://docs.swift.org/compiler/documentation/diagnostics/no-usage>
/Users/admin/builder/spi-builder-workspace/Sources/SpeechVAD/SortformerStreamingSession.swift:232:18: [1;33mwarning: [1;39mresult of call to 'withUnsafeMutableBufferPointer' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m230 |[0;0m var chunkMel = [Float](repeating: 0, count: coreMLInputFrames * nMels)
[0;36m231 |[0;0m let framesToCopy = min(actualLen, coreMLInputFrames)
[0;36m232 |[0;0m chunkMel.withUnsafeMutableBufferPointer [4;39m{ dst in[0;0m
[0;36m|[0;0m `- [1;33mwarning: [1;39mresult of call to 'withUnsafeMutableBufferPointer' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m233 |[0;0m melSpec.withUnsafeBufferPointer { src in
[0;36m234 |[0;0m memcpy(
[#NoUsage]: <https://docs.swift.org/compiler/documentation/diagnostics/no-usage>
/Users/admin/builder/spi-builder-workspace/Sources/SpeechVAD/MelFeatureExtractor.swift:222:40: [1;33mwarning: [1;39mcannot use inout expression here; argument #1 must be a pointer that outlives the call to '+'[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/temporary-pointers\TemporaryPointers]8;;\]
[0;36m220 |[0;0m for frame in 0..<nFrames {
[0;36m221 |[0;0m let base = frame * nMels
[0;36m222 |[0;0m vDSP_vsub(binMeans, 1, [4;39m&melSpec[0;0m + base, 1, &melSpec + base, 1, vDSP_Length(nMels))
[0;36m|[0;0m |- [1;33mwarning: [1;39mcannot use inout expression here; argument #1 must be a pointer that outlives the call to '+'[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/temporary-pointers\TemporaryPointers]8;;\]
[0;36m|[0;0m |- [1;39mnote: [1;39mimplicit argument conversion from '[Float]' to 'UnsafePointer<Float>' produces a pointer valid only for the duration of the call to '+'[0;0m
[0;36m|[0;0m `- [1;39mnote: [1;39muse the 'withUnsafeBufferPointer' method on Array in order to explicitly convert argument to buffer pointer valid for a defined scope[0;0m
[0;36m223 |[0;0m }
[0;36m224 |[0;0m }
/Users/admin/builder/spi-builder-workspace/Sources/SpeechVAD/MelFeatureExtractor.swift:222:60: [1;33mwarning: [1;39mcannot use inout expression here; argument #1 must be a pointer that outlives the call to '+'[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/temporary-pointers\TemporaryPointers]8;;\]
[0;36m220 |[0;0m for frame in 0..<nFrames {
[0;36m221 |[0;0m let base = frame * nMels
[0;36m222 |[0;0m vDSP_vsub(binMeans, 1, &melSpec + base, 1, [4;39m&melSpec[0;0m + base, 1, vDSP_Length(nMels))
[0;36m|[0;0m |- [1;33mwarning: [1;39mcannot use inout expression here; argument #1 must be a pointer that outlives the call to '+'[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/temporary-pointers\TemporaryPointers]8;;\]
[0;36m|[0;0m |- [1;39mnote: [1;39mimplicit argument conversion from '[Float]' to 'UnsafeMutablePointer<Float>' produces a pointer valid only for the duration of the call to '+'[0;0m
[0;36m|[0;0m `- [1;39mnote: [1;39muse the 'withUnsafeMutableBufferPointer' method on Array in order to explicitly convert argument to buffer pointer valid for a defined scope[0;0m
[0;36m223 |[0;0m }
[0;36m224 |[0;0m }
[#TemporaryPointers]: <https://docs.swift.org/compiler/documentation/diagnostics/temporary-pointers>
/Users/admin/builder/spi-builder-workspace/Sources/SourceSeparation/SourceSeparation.swift:243:13: [1;33mwarning: [1;39mno calls to throwing functions occur within 'try' expression[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/unnecessary-effect-marker\UnnecessaryEffectMarker]8;;\]
[0;36m241 |[0;0m let weights = try MLX.loadArrays(url: weightsURL)
[0;36m242 |[0;0m let params = ModuleParameters.unflattened(weights)
[0;36m243 |[0;0m try model.update(parameters: params)
[0;36m|[0;0m `- [1;33mwarning: [1;39mno calls to throwing functions occur within 'try' expression[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/unnecessary-effect-marker\UnnecessaryEffectMarker]8;;\]
[0;36m244 |[0;0m model.train(false) // Use pretrained running_mean/running_var in BatchNorm
[0;36m245 |[0;0m // Concatenate LSTM input/hidden weight matrices so each timestep
[#UnnecessaryEffectMarker]: <https://docs.swift.org/compiler/documentation/diagnostics/unnecessary-effect-marker>
/Users/admin/builder/spi-builder-workspace/Sources/SourceSeparation/HTDemucs/HTDemucs.swift:60:17: [1;33mwarning: [1;39mvariable 'stri' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m58 |[0;0m let freq = freqs > 1
[0;36m59 |[0;0m var ker = freq ? a.kernelSize : a.timeStride * 2
[0;36m60 |[0;0m var stri = freq ? a.stride : a.timeStride
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'stri' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m61 |[0;0m var padBool = true
[0;36m62 |[0;0m var lastFreq = false
/Users/admin/builder/spi-builder-workspace/Sources/SourceSeparation/HTDemucs/HTDemucs.swift:141:13: [1;33mwarning: [1;39minitialization of immutable value 'length' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m139 |[0;0m func callAsFunction(_ mixIn: MLXArray) -> MLXArray {
[0;36m140 |[0;0m var mix = mixIn
[0;36m141 |[0;0m let length = mix.dim(-1)
[0;36m|[0;0m `- [1;33mwarning: [1;39minitialization of immutable value 'length' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m142 |[0;0m var lengthPrePad: Int? = nil
[0;36m143 |[0;0m if mix.dim(-1) < trainingLength {
/Users/admin/builder/spi-builder-workspace/Sources/SourceSeparation/HTDemucs/HTDemucs.swift:221:13: [1;33mwarning: [1;39mvariable 'xtOut' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m219 |[0;0m
[0;36m220 |[0;0m let actualLen = xt.dim(-1)
[0;36m221 |[0;0m var xtOut = (xt * stdt + meant).reshaped([B, S, cfg.audioChannels, actualLen])
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'xtOut' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m222 |[0;0m out = centerTrim(out, length: actualLen)
[0;36m223 |[0;0m out = xtOut + out
[#NoUsage]: <https://docs.swift.org/compiler/documentation/diagnostics/no-usage>
[#VariableNeverMutated]: <https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated>
/Users/admin/builder/spi-builder-workspace/Sources/Qwen3TTS/Qwen3TTS+ICL.swift:353:13: [1;33mwarning: [1;39minitialization of immutable value 'codecSuffixEmbed' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m351 |[0;0m codecPrefixEmbed = concatenated([part0, spkEmbedReshaped, part1], axis: 1)
[0;36m352 |[0;0m
[0;36m353 |[0;0m let codecSuffixEmbed = talker.embedCodec(
[0;36m|[0;0m `- [1;33mwarning: [1;39minitialization of immutable value 'codecSuffixEmbed' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m354 |[0;0m MLXArray([Int32(CodecTokens.codecPad), Int32(CodecTokens.codecBos)]).expandedDimensions(axis: 0))
[0;36m355 |[0;0m
[#NoUsage]: <https://docs.swift.org/compiler/documentation/diagnostics/no-usage>
/Users/admin/builder/spi-builder-workspace/Sources/PersonaPlex/MimiTransformer.swift:51:13: [1;33mwarning: [1;39minitialization of immutable value 'hd' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m49 |[0;0m let b = xs.shape[0]
[0;36m50 |[0;0m let t = xs.shape[1]
[0;36m51 |[0;0m let hd = xs.shape[2]
[0;36m|[0;0m `- [1;33mwarning: [1;39minitialization of immutable value 'hd' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m52 |[0;0m
[0;36m53 |[0;0m let qkv = in_proj(xs).reshaped([b, t, 3, cfg.numHeads, cfg.headDim])
[#NoUsage]: <https://docs.swift.org/compiler/documentation/diagnostics/no-usage>
/Users/admin/builder/spi-builder-workspace/Sources/Qwen3TTS/Qwen3TTS.swift:676:13: [1;33mwarning: [1;39minitialization of immutable value 'actualContext' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m674 |[0;0m ) -> [Float] {
[0;36m675 |[0;0m let contextStart = max(chunkStart - decoderLeftContext, 0)
[0;36m676 |[0;0m let actualContext = chunkStart - contextStart
[0;36m|[0;0m `- [1;33mwarning: [1;39minitialization of immutable value 'actualContext' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m677 |[0;0m
[0;36m678 |[0;0m // Build [1, 16, contextFrames + chunkFrames] from accumulated codebooks
[#NoUsage]: <https://docs.swift.org/compiler/documentation/diagnostics/no-usage>
warning: 'HiggsTTS' is missing a dependency on 'MLXRandom' because dependency scan of Swift module 'HiggsTTS' discovered a dependency on 'MLXRandom'
/Users/admin/builder/spi-builder-workspace/Sources/PersonaPlex/PersonaPlex.swift:987:19: [1;33mwarning: [1;39m'catch' block is unreachable because no errors are thrown in 'do' block[0;0m
[0;36m985 |[0;0m
[0;36m986 |[0;0m continuation.finish()
[0;36m987 |[0;0m } catch {
[0;36m|[0;0m `- [1;33mwarning: [1;39m'catch' block is unreachable because no errors are thrown in 'do' block[0;0m
[0;36m988 |[0;0m continuation.finish(throwing: error)
[0;36m989 |[0;0m }
/Users/admin/builder/spi-builder-workspace/Sources/CosyVoiceTTS/VoiceCloning.swift:30:16: [1;33mwarning: [1;39mstored property 'promptToken' of 'Sendable'-conforming struct 'CosyVoiceVoiceProfile' contains non-Sendable type 'MLXArray'; this is an error in the Swift 6 language mode[0;0m
[0;36m28 |[0;0m public struct CosyVoiceVoiceProfile: Sendable {
[0;36m29 |[0;0m public let speakerEmbedding: [Float]?
[0;36m30 |[0;0m public let promptToken: MLXArray?
[0;36m|[0;0m `- [1;33mwarning: [1;39mstored property 'promptToken' of 'Sendable'-conforming struct 'CosyVoiceVoiceProfile' contains non-Sendable type 'MLXArray'; this is an error in the Swift 6 language mode[0;0m
[0;36m31 |[0;0m public let promptFeat: MLXArray?
[0;36m32 |[0;0m public let promptText: String?
MLX.MLXArray:1:20: [1;39mnote: [1;39mclass 'MLXArray' does not conform to the 'Sendable' protocol[0;0m
[0;36m1 |[0;0m final public class MLXArray {
[0;36m|[0;0m `- [1;39mnote: [1;39mclass 'MLXArray' does not conform to the 'Sendable' protocol[0;0m
[0;36m2 |[0;0m public internal(set) var ctx: mlx_array { get }
[0;36m3 |[0;0m public init(_ ctx: consuming mlx_array)
/Users/admin/builder/spi-builder-workspace/Sources/CosyVoiceTTS/VoiceCloning.swift:31:16: [1;33mwarning: [1;39mstored property 'promptFeat' of 'Sendable'-conforming struct 'CosyVoiceVoiceProfile' contains non-Sendable type 'MLXArray'; this is an error in the Swift 6 language mode[0;0m
[0;36m29 |[0;0m public let speakerEmbedding: [Float]?
[0;36m30 |[0;0m public let promptToken: MLXArray?
[0;36m31 |[0;0m public let promptFeat: MLXArray?
[0;36m|[0;0m `- [1;33mwarning: [1;39mstored property 'promptFeat' of 'Sendable'-conforming struct 'CosyVoiceVoiceProfile' contains non-Sendable type 'MLXArray'; this is an error in the Swift 6 language mode[0;0m
[0;36m32 |[0;0m public let promptText: String?
[0;36m33 |[0;0m
MLX.MLXArray:1:20: [1;39mnote: [1;39mclass 'MLXArray' does not conform to the 'Sendable' protocol[0;0m
[0;36m1 |[0;0m final public class MLXArray {
[0;36m|[0;0m `- [1;39mnote: [1;39mclass 'MLXArray' does not conform to the 'Sendable' protocol[0;0m
[0;36m2 |[0;0m public internal(set) var ctx: mlx_array { get }
[0;36m3 |[0;0m public init(_ ctx: consuming mlx_array)
/Users/admin/builder/spi-builder-workspace/Sources/OmniVoiceTTS/OmniVoiceBackbone.swift:55:13: [1;33mwarning: [1;39mvariable 'v' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m53 |[0;0m var q = qNorm(qProj(x).reshaped([b, l, nHeads, headDim])).transposed(0, 2, 1, 3)
[0;36m54 |[0;0m var k = kNorm(kProj(x).reshaped([b, l, nKVHeads, headDim])).transposed(0, 2, 1, 3)
[0;36m55 |[0;0m var v = vProj(x).reshaped([b, l, nKVHeads, headDim]).transposed(0, 2, 1, 3)
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'v' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m56 |[0;0m q = freqs.rope(q)
[0;36m57 |[0;0m k = freqs.rope(k)
[#VariableNeverMutated]: <https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated>
/Users/admin/builder/spi-builder-workspace/Sources/Qwen3ASR/AudioPreprocessing.swift:128:21: [1;33mwarning: [1;39minitialization of immutable value 'centerFreq' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m126 |[0;0m // Filter edges: filterFreqs[mel], filterFreqs[mel+1], filterFreqs[mel+2]
[0;36m127 |[0;0m let lowFreq = filterFreqs[mel]
[0;36m128 |[0;0m let centerFreq = filterFreqs[mel + 1]
[0;36m|[0;0m `- [1;33mwarning: [1;39minitialization of immutable value 'centerFreq' was never used; consider replacing with assignment to '_' or removing it[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m129 |[0;0m let highFreq = filterFreqs[mel + 2]
[0;36m130 |[0;0m
[#NoUsage]: <https://docs.swift.org/compiler/documentation/diagnostics/no-usage>
/Users/admin/builder/spi-builder-workspace/Sources/OmniVoiceTTS/OmniVoiceDiffusion.swift:29:13: [1;33mwarning: [1;39mvariable 'cond' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m27 |[0;0m
[0;36m28 |[0;0m var tokens = MLXArray.full([1, C, targetLen], values: MLXArray(Int32(maskId)))
[0;36m29 |[0;0m var cond = condInputIds
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'cond' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m30 |[0;0m // The target region must start fully masked regardless of what the caller
[0;36m31 |[0;0m // left in `condInputIds` (the loop fills it in from here).
/Users/admin/builder/spi-builder-workspace/Sources/OmniVoiceTTS/OmniVoiceDiffusion.swift:64:17: [1;33mwarning: [1;39mvariable 'maskCol' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m62 |[0;0m var logProbs = logSoftmax(cLog + guidance * (cLog - uLog))
[0;36m63 |[0;0m // Never emit the mask token.
[0;36m64 |[0;0m var maskCol = MLXArray.zeros([cfg.audioVocabSize])
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'maskCol' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m65 |[0;0m maskCol[maskId] = negInf
[0;36m66 |[0;0m logProbs = logProbs + maskCol.reshaped([1, 1, 1, cfg.audioVocabSize])
/Users/admin/builder/spi-builder-workspace/Sources/OmniVoiceTTS/OmniVoiceDiffusion.swift:78:17: [1;33mwarning: [1;39mvariable 'pick' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m76 |[0;0m let order = MLX.argSort(-flat) // descending
[0;36m77 |[0;0m let topk = order[0 ..< k]
[0;36m78 |[0;0m var pick = MLXArray.zeros([C * targetLen]).asType(.int32)
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'pick' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m79 |[0;0m pick[topk] = MLXArray.ones([k]).asType(.int32)
[0;36m80 |[0;0m let pickMask = (pick.reshaped([1, C, targetLen]) .!= MLXArray(Int32(0)))
[#VariableNeverMutated]: <https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated>
/Users/admin/builder/spi-builder-workspace/Sources/FlashSR/Mel.swift:158:13: [1;33mwarning: [1;39mvariable 'hzPoints' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m156 |[0;0m melPoints[i] = melMin + frac * (melMax - melMin)
[0;36m157 |[0;0m }
[0;36m158 |[0;0m var hzPoints = melPoints.map { melToHzSlaney($0) }
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'hzPoints' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m159 |[0;0m // Build triangular filters
[0;36m160 |[0;0m var fb = [Float](repeating: 0, count: nMels * nBins)
[#VariableNeverMutated]: <https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated>
/Users/admin/builder/spi-builder-workspace/Sources/CosyVoiceTTS/VoiceCloning.swift:30:16: [1;33mwarning: [1;39mstored property 'promptToken' of 'Sendable'-conforming struct 'CosyVoiceVoiceProfile' contains non-Sendable type 'MLXArray'; this is an error in the Swift 6 language mode[0;0m
[0;36m28 |[0;0m public struct CosyVoiceVoiceProfile: Sendable {
[0;36m29 |[0;0m public let speakerEmbedding: [Float]?
[0;36m30 |[0;0m public let promptToken: MLXArray?
[0;36m|[0;0m `- [1;33mwarning: [1;39mstored property 'promptToken' of 'Sendable'-conforming struct 'CosyVoiceVoiceProfile' contains non-Sendable type 'MLXArray'; this is an error in the Swift 6 language mode[0;0m
[0;36m31 |[0;0m public let promptFeat: MLXArray?
[0;36m32 |[0;0m public let promptText: String?
MLX.MLXArray:1:20: [1;39mnote: [1;39mclass 'MLXArray' does not conform to the 'Sendable' protocol[0;0m
[0;36m1 |[0;0m final public class MLXArray {
[0;36m|[0;0m `- [1;39mnote: [1;39mclass 'MLXArray' does not conform to the 'Sendable' protocol[0;0m
[0;36m2 |[0;0m public internal(set) var ctx: mlx_array { get }
[0;36m3 |[0;0m public init(_ ctx: consuming mlx_array)
/Users/admin/builder/spi-builder-workspace/Sources/CosyVoiceTTS/VoiceCloning.swift:31:16: [1;33mwarning: [1;39mstored property 'promptFeat' of 'Sendable'-conforming struct 'CosyVoiceVoiceProfile' contains non-Sendable type 'MLXArray'; this is an error in the Swift 6 language mode[0;0m
[0;36m29 |[0;0m public let speakerEmbedding: [Float]?
[0;36m30 |[0;0m public let promptToken: MLXArray?
[0;36m31 |[0;0m public let promptFeat: MLXArray?
[0;36m|[0;0m `- [1;33mwarning: [1;39mstored property 'promptFeat' of 'Sendable'-conforming struct 'CosyVoiceVoiceProfile' contains non-Sendable type 'MLXArray'; this is an error in the Swift 6 language mode[0;0m
[0;36m32 |[0;0m public let promptText: String?
[0;36m33 |[0;0m
MLX.MLXArray:1:20: [1;39mnote: [1;39mclass 'MLXArray' does not conform to the 'Sendable' protocol[0;0m
[0;36m1 |[0;0m final public class MLXArray {
[0;36m|[0;0m `- [1;39mnote: [1;39mclass 'MLXArray' does not conform to the 'Sendable' protocol[0;0m
[0;36m2 |[0;0m public internal(set) var ctx: mlx_array { get }
[0;36m3 |[0;0m public init(_ ctx: consuming mlx_array)
/Users/admin/builder/spi-builder-workspace/Sources/CosyVoiceTTS/CamPlusPlusMelExtractor.swift:212:40: [1;33mwarning: [1;39mcannot use inout expression here; argument #1 must be a pointer that outlives the call to '+'[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/temporary-pointers\TemporaryPointers]8;;\]
[0;36m210 |[0;0m for frame in 0..<nFrames {
[0;36m211 |[0;0m let base = frame * nMels
[0;36m212 |[0;0m vDSP_vsub(binMeans, 1, [4;39m&melSpec[0;0m + base, 1, &melSpec + base, 1, vDSP_Length(nMels))
[0;36m|[0;0m |- [1;33mwarning: [1;39mcannot use inout expression here; argument #1 must be a pointer that outlives the call to '+'[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/temporary-pointers\TemporaryPointers]8;;\]
[0;36m|[0;0m |- [1;39mnote: [1;39mimplicit argument conversion from '[Float]' to 'UnsafePointer<Float>' produces a pointer valid only for the duration of the call to '+'[0;0m
[0;36m|[0;0m `- [1;39mnote: [1;39muse the 'withUnsafeBufferPointer' method on Array in order to explicitly convert argument to buffer pointer valid for a defined scope[0;0m
[0;36m213 |[0;0m }
[0;36m214 |[0;0m }
/Users/admin/builder/spi-builder-workspace/Sources/CosyVoiceTTS/CamPlusPlusMelExtractor.swift:212:60: [1;33mwarning: [1;39mcannot use inout expression here; argument #1 must be a pointer that outlives the call to '+'[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/temporary-pointers\TemporaryPointers]8;;\]
[0;36m210 |[0;0m for frame in 0..<nFrames {
[0;36m211 |[0;0m let base = frame * nMels
[0;36m212 |[0;0m vDSP_vsub(binMeans, 1, &melSpec + base, 1, [4;39m&melSpec[0;0m + base, 1, vDSP_Length(nMels))
[0;36m|[0;0m |- [1;33mwarning: [1;39mcannot use inout expression here; argument #1 must be a pointer that outlives the call to '+'[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/temporary-pointers\TemporaryPointers]8;;\]
[0;36m|[0;0m |- [1;39mnote: [1;39mimplicit argument conversion from '[Float]' to 'UnsafeMutablePointer<Float>' produces a pointer valid only for the duration of the call to '+'[0;0m
[0;36m|[0;0m `- [1;39mnote: [1;39muse the 'withUnsafeMutableBufferPointer' method on Array in order to explicitly convert argument to buffer pointer valid for a defined scope[0;0m
[0;36m213 |[0;0m }
[0;36m214 |[0;0m }
[#TemporaryPointers]: <https://docs.swift.org/compiler/documentation/diagnostics/temporary-pointers>
/Users/admin/builder/spi-builder-workspace/Sources/CosyVoiceTTS/CosyVoiceTTS.swift:456:19: [1;33mwarning: [1;39m'catch' block is unreachable because no errors are thrown in 'do' block[0;0m
[0;36m454 |[0;0m continuation.yield(chunk)
[0;36m455 |[0;0m continuation.finish()
[0;36m456 |[0;0m } catch {
[0;36m|[0;0m `- [1;33mwarning: [1;39m'catch' block is unreachable because no errors are thrown in 'do' block[0;0m
[0;36m457 |[0;0m continuation.finish(throwing: error)
[0;36m458 |[0;0m }
/Users/admin/builder/spi-builder-workspace/Sources/Qwen3Chat/JSONSchema/XGrammarConstraint.swift:212:17: [1;33mwarning: [1;39mvariable 'masked' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m210 |[0;0m let floor = MLXArray(-Float.greatestFiniteMagnitude)
[0;36m211 |[0;0m if flags.dim(0) < vocab {
[0;36m212 |[0;0m var masked = full([vocab], values: floor)
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'masked' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m213 |[0;0m let n = flags.dim(0)
[0;36m214 |[0;0m masked[0 ..< n] = which(flags .!= MLXArray(UInt32(0)), logits[0 ..< n], floor)
[#VariableNeverMutated]: <https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated>
/Users/admin/builder/spi-builder-workspace/Sources/Qwen3Chat/ChatTokenizer.swift:143:13: [1;33mwarning: [1;39mvariable 'words' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m141 |[0;0m
[0;36m142 |[0;0m // Simple BPE encoding
[0;36m143 |[0;0m var words = tokenizeToWords(text)
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'words' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m144 |[0;0m var allTokens: [Int] = []
[0;36m145 |[0;0m
[#VariableNeverMutated]: <https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated>
/Users/admin/builder/spi-builder-workspace/Sources/Qwen3Chat/MLXGenerator.swift:331:9: [1;33mwarning: [1;39mresult of call to 'withUnsafeMutablePointer(to:_:)' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m329 |[0;0m var memInfo = mach_task_basic_info()
[0;36m330 |[0;0m var memCount = mach_msg_type_number_t(MemoryLayout<mach_task_basic_info>.size) / 4
[0;36m331 |[0;0m withUnsafeMutablePointer(to: &memInfo) {
[0;36m|[0;0m `- [1;33mwarning: [1;39mresult of call to 'withUnsafeMutablePointer(to:_:)' is unused[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/no-usage\NoUsage]8;;\]
[0;36m332 |[0;0m $0.withMemoryRebound(to: integer_t.self, capacity: Int(memCount)) {
[0;36m333 |[0;0m task_info(mach_task_self_, task_flavor_t(MACH_TASK_BASIC_INFO), $0, &memCount)
[#NoUsage]: <https://docs.swift.org/compiler/documentation/diagnostics/no-usage>
/Users/admin/builder/spi-builder-workspace/Sources/MagpieTTSCoreML/CoreMLHelpers.swift:111:9: [1;33mwarning: [1;39mswitch must be exhaustive[0;0m
[0;36m109 |[0;0m let count = array.count
[0;36m110 |[0;0m var out = [Float](repeating: 0, count: count)
[0;36m111 |[0;0m switch array.dataType {
[0;36m|[0;0m |- [1;33mwarning: [1;39mswitch must be exhaustive[0;0m
[0;36m|[0;0m `- [1;39mnote: [1;39madd missing case: '.int8'[0;0m
[0;36m112 |[0;0m case .float32:
[0;36m113 |[0;0m let p = array.dataPointer.bindMemory(to: Float.self, capacity: count)
/Users/admin/builder/spi-builder-workspace/Sources/MagpieTTSCoreML/StatefulDecoder.swift:169:9: [1;33mwarning: [1;39mvariable 'exp' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m167 |[0;0m let bits = value.bitPattern
[0;36m168 |[0;0m let sign = UInt16((bits >> 16) & 0x8000)
[0;36m169 |[0;0m var exp = Int32((bits >> 23) & 0xFF) - 127 + 15
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'exp' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m170 |[0;0m var mant = bits & 0x7F_FFFF
[0;36m171 |[0;0m if exp <= 0 {
[#VariableNeverMutated]: <https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated>
/Users/admin/builder/spi-builder-workspace/Sources/MagpieTTSCoreML/LocalTransformer.swift:30:21: [1;33mwarning: [1;39m'cblas_sgemv' was deprecated in macOS 13.3: An updated CBLAS interface supporting ILP64 is available. Please compile with -DACCELERATE_NEW_LAPACK to access the new headers and -DACCELERATE_LAPACK_ILP64 for ILP64 support.[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration\DeprecatedDeclaration]8;;\]
[0;36m28 |[0;0m hidden.withUnsafeBufferPointer { hPtr in
[0;36m29 |[0;0m out.withUnsafeMutableBufferPointer { outPtr in
[0;36m30 |[0;0m cblas_sgemv(
[0;36m|[0;0m `- [1;33mwarning: [1;39m'cblas_sgemv' was deprecated in macOS 13.3: An updated CBLAS interface supporting ILP64 is available. Please compile with -DACCELERATE_NEW_LAPACK to access the new headers and -DACCELERATE_LAPACK_ILP64 for ILP64 support.[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration\DeprecatedDeclaration]8;;\]
[0;36m31 |[0;0m CblasRowMajor, CblasNoTrans,
[0;36m32 |[0;0m Int32(D), Int32(M),
/Users/admin/builder/spi-builder-workspace/Sources/MagpieTTSCoreML/LocalTransformer.swift:172:21: [1;33mwarning: [1;39m'cblas_sgemv' was deprecated in macOS 13.3: An updated CBLAS interface supporting ILP64 is available. Please compile with -DACCELERATE_NEW_LAPACK to access the new headers and -DACCELERATE_LAPACK_ILP64 for ILP64 support.[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration\DeprecatedDeclaration]8;;\]
[0;36m170 |[0;0m lastHidden.withUnsafeBufferPointer { hPtr in
[0;36m171 |[0;0m out.withUnsafeMutableBufferPointer { oPtr in
[0;36m172 |[0;0m cblas_sgemv(
[0;36m|[0;0m `- [1;33mwarning: [1;39m'cblas_sgemv' was deprecated in macOS 13.3: An updated CBLAS interface supporting ILP64 is available. Please compile with -DACCELERATE_NEW_LAPACK to access the new headers and -DACCELERATE_LAPACK_ILP64 for ILP64 support.[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration\DeprecatedDeclaration]8;;\]
[0;36m173 |[0;0m CblasRowMajor, CblasNoTrans,
[0;36m174 |[0;0m Int32(V), Int32(D),
/Users/admin/builder/spi-builder-workspace/Sources/MagpieTTSCoreML/LocalTransformer.swift:243:21: [1;33mwarning: [1;39m'cblas_sgemm' was deprecated in macOS 13.3: An updated CBLAS interface supporting ILP64 is available. Please compile with -DACCELERATE_NEW_LAPACK to access the new headers and -DACCELERATE_LAPACK_ILP64 for ILP64 support.[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration\DeprecatedDeclaration]8;;\]
[0;36m241 |[0;0m b.withUnsafeBufferPointer { bp in
[0;36m242 |[0;0m out.withUnsafeMutableBufferPointer { op in
[0;36m243 |[0;0m cblas_sgemm(
[0;36m|[0;0m `- [1;33mwarning: [1;39m'cblas_sgemm' was deprecated in macOS 13.3: An updated CBLAS interface supporting ILP64 is available. Please compile with -DACCELERATE_NEW_LAPACK to access the new headers and -DACCELERATE_LAPACK_ILP64 for ILP64 support.[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration\DeprecatedDeclaration]8;;\]
[0;36m244 |[0;0m CblasRowMajor, CblasNoTrans, CblasNoTrans,
[0;36m245 |[0;0m Int32(M), Int32(N), Int32(K),
/Users/admin/builder/spi-builder-workspace/Sources/MagpieTTSCoreML/LocalTransformer.swift:265:21: [1;33mwarning: [1;39m'cblas_sgemm' was deprecated in macOS 13.3: An updated CBLAS interface supporting ILP64 is available. Please compile with -DACCELERATE_NEW_LAPACK to access the new headers and -DACCELERATE_LAPACK_ILP64 for ILP64 support.[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration\DeprecatedDeclaration]8;;\]
[0;36m263 |[0;0m b.withUnsafeBufferPointer { bp in
[0;36m264 |[0;0m out.withUnsafeMutableBufferPointer { op in
[0;36m265 |[0;0m cblas_sgemm(
[0;36m|[0;0m `- [1;33mwarning: [1;39m'cblas_sgemm' was deprecated in macOS 13.3: An updated CBLAS interface supporting ILP64 is available. Please compile with -DACCELERATE_NEW_LAPACK to access the new headers and -DACCELERATE_LAPACK_ILP64 for ILP64 support.[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration\DeprecatedDeclaration]8;;\]
[0;36m266 |[0;0m CblasRowMajor, CblasNoTrans, CblasTrans,
[0;36m267 |[0;0m Int32(M), Int32(N), Int32(K),
[#DeprecatedDeclaration]: <https://docs.swift.org/compiler/documentation/diagnostics/deprecated-declaration>
/Users/admin/builder/spi-builder-workspace/Sources/HibikiTranslate/HibikiTranslate.swift:345:19: [1;33mwarning: [1;39m'catch' block is unreachable because no errors are thrown in 'do' block[0;0m
[0;36m343 |[0;0m continuation.yield(chunk)
[0;36m344 |[0;0m continuation.finish()
[0;36m345 |[0;0m } catch {
[0;36m|[0;0m `- [1;33mwarning: [1;39m'catch' block is unreachable because no errors are thrown in 'do' block[0;0m
[0;36m346 |[0;0m continuation.finish(throwing: error)
[0;36m347 |[0;0m }
/Users/admin/builder/spi-builder-workspace/Sources/CSM/CSMGenerate.swift:55:13: [1;33mwarning: [1;39mvariable 'c0' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m53 |[0;0m let lastH = h[0..., -1] // [B,dim]
[0;36m54 |[0;0m let c0Logits = applyLinear(codebook0Head, lastH) // [B,audioVocab]
[0;36m55 |[0;0m var c0 = sampler(c0Logits).expandedDimensions(axis: -1) // [B,1]
[0;36m|[0;0m `- [1;33mwarning: [1;39mvariable 'c0' was never mutated; consider changing to 'let' constant[0;0m [#]8;;https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated\VariableNeverMutated]8;;\]
[0;36m56 |[0;0m var curr = embedAudio(0, c0) // [B,1,dim]
[0;36m57 |[0;0m curr = concatenated([lastH.expandedDimensions(axis: 1), curr], axis: 1) // [B,2,dim]
[#VariableNeverMutated]: <https://docs.swift.org/compiler/documentation/diagnostics/variable-never-mutated>
/Users/admin/builder/spi-builder-workspace/Sources/AudioServer/NativeInference.swift:339:90: [1;33mwarning: [1;39mconverting non-Sendable function value to '@Sendable (Gemma4GenerationTrace) -> Void' may introduce data races[0;0m
[0;36m337 |[0;0m let capture = self.failures != nil && input.operation == .chat ? NativeInferenceGenerationCapture() : nil
[0;36m338 |[0;0m do {
[0;36m339 |[0;0m let observer: (@Sendable (Gemma4GenerationTrace) -> Void)? = capture.map { capture in
[0;36m|[0;0m `- [1;33mwarning: [1;39mconverting non-Sendable function value to '@Sendable (Gemma4GenerationTrace) -> Void' may introduce data races[0;0m
[0;36m340 |[0;0m { trace in capture.store(trace) }
[0;36m341 |[0;0m }
/Users/admin/builder/spi-builder-workspace/Sources/AudioServer/NativeInference.swift:339:90: [1;33mwarning: [1;39mconverting non-Sendable function value to '@Sendable (Gemma4GenerationTrace) -> Void' may introduce data races[0;0m
[0;36m337 |[0;0m let capture = self.failures != nil && input.operation == .chat ? NativeInferenceGenerationCapture() : nil
[0;36m338 |[0;0m do {
[0;36m339 |[0;0m let observer: (@Sendable (Gemma4GenerationTrace) -> Void)? = capture.map { capture in
[0;36m|[0;0m `- [1;33mwarning: [1;39mconverting non-Sendable function value to '@Sendable (Gemma4GenerationTrace) -> Void' may introduce data races[0;0m
[0;36m340 |[0;0m { trace in capture.store(trace) }
[0;36m341 |[0;0m }
/Users/admin/builder/spi-builder-workspace/Sources/AudioCLILib/TranscribeCommand.swift:378:89: [1;33mwarning: [1;39mconverting non-Sendable function value to '@Sendable (Double, String) -> Void' may introduce data races[0;0m
[0;36m376 |[0;0m print("Loading model (\(sizeLabel)): \(modelId)")
[0;36m377 |[0;0m let runtime = Qwen3ASRRuntime(
[0;36m378 |[0;0m descriptor: Qwen3ASRDescriptors.mlx(modelID: modelId), progressHandler: reportProgress)
[0;36m|[0;0m `- [1;33mwarning: [1;39mconverting non-Sendable function value to '@Sendable (Double, String) -> Void' may introduce data races[0;0m
[0;36m379 |[0;0m try await runtime.prepare()
[0;36m380 |[0;0m
/Users/admin/builder/spi-builder-workspace/Sources/AudioCLILib/TranscribeCommand.swift:479:65: [1;33mwarning: [1;39mconverting non-Sendable function value to '@Sendable (Double, String) -> Void' may introduce data races[0;0m
[0;36m477 |[0;0m print("Loading full CoreML ASR pipeline...")
[0;36m478 |[0;0m let runtime = Qwen3ASRRuntime(descriptor: Qwen3ASRDescriptors.coreML,
[0;36m479 |[0;0m progressHandler: reportProgress)
[0;36m|[0;0m `- [1;33mwarning: [1;39mconverting non-Sendable function value to '@Sendable (Double, String) -> Void' may introduce data races[0;0m
[0;36m480 |[0;0m try await runtime.prepare()
[0;36m481 |[0;0m
Build complete.
{
"cxx_language_standard" : "c++17",
"dependencies" : [
{
"identity" : "mlx-swift",
"requirement" : {
"range" : [
{
"lower_bound" : "0.30.0",
"upper_bound" : "1.0.0"
}
]
},
"type" : "sourceControl",
"url" : "https://github.com/ml-explore/mlx-swift"
},
{
"identity" : "mlx-swift-lm",
"requirement" : {
"exact" : [
"3.31.4"
]
},
"type" : "sourceControl",
"url" : "https://github.com/ml-explore/mlx-swift-lm"
},
{
"identity" : "swift-argument-parser",
"requirement" : {
"range" : [
{
"lower_bound" : "1.5.0",
"upper_bound" : "2.0.0"
}
]
},
"type" : "sourceControl",
"url" : "https://github.com/apple/swift-argument-parser"
},
{
"identity" : "swift-sdk",
"requirement" : {
"exact" : [
"0.12.1"
]
},
"type" : "sourceControl",
"url" : "https://github.com/modelcontextprotocol/swift-sdk.git"
},
{
"identity" : "swift-system",
"requirement" : {
"range" : [
{
"lower_bound" : "1.8.0",
"upper_bound" : "2.0.0"
}
]
},
"type" : "sourceControl",
"url" : "https://github.com/apple/swift-system.git"
},
{
"identity" : "swift-transformers",
"requirement" : {
"range" : [
{
"lower_bound" : "1.1.6",
"upper_bound" : "2.0.0"
}
]
},
"type" : "sourceControl",
"url" : "https://github.com/huggingface/swift-transformers"
},
{
"identity" : "hummingbird",
"requirement" : {
"range" : [
{
"lower_bound" : "2.5.0",
"upper_bound" : "2.17.0"
}
]
},
"type" : "sourceControl",
"url" : "https://github.com/hummingbird-project/hummingbird.git"
},
{
"identity" : "hummingbird-websocket",
"requirement" : {
"range" : [
{
"lower_bound" : "2.6.0",
"upper_bound" : "2.7.0"
}
]
},
"type" : "sourceControl",
"url" : "https://github.com/hummingbird-project/hummingbird-websocket.git"
},
{
"identity" : "swift-websocket",
"requirement" : {
"range" : [
{
"lower_bound" : "1.5.0",
"upper_bound" : "1.6.0"
}
]
},
"type" : "sourceControl",
"url" : "https://github.com/hummingbird-project/swift-websocket.git"
},
{
"identity" : "whisperkit",
"requirement" : {
"range" : [
{
"lower_bound" : "1.0.0",
"upper_bound" : "2.0.0"
}
]
},
"type" : "sourceControl",
"url" : "https://github.com/argmaxinc/WhisperKit"
},
{
"identity" : "xgrammar",
"requirement" : {
"exact" : [
"0.2.7"
]
},
"type" : "sourceControl",
"url" : "https://github.com/mlc-ai/xgrammar"
}
],
"manifest_display_name" : "Qwen3Speech",
"name" : "Qwen3Speech",
"path" : "/Users/admin/builder/spi-builder-workspace",
"platforms" : [
{
"name" : "macos",
"version" : "15.0"
},
{
"name" : "ios",
"version" : "18.0"
}
],
"products" : [
{
"name" : "PocketTTS",
"targets" : [
"PocketTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "pocket-tts-bench",
"targets" : [
"PocketTTSBenchmark"
],
"type" : {
"executable" : null
}
},
{
"name" : "SpeechRuntime",
"targets" : [
"SpeechRuntime"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "VoxtralTTS",
"targets" : [
"VoxtralTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "Qwen3ASRRuntime",
"targets" : [
"Qwen3ASRRuntime"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "GLiNER",
"targets" : [
"GLiNER"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "Clef",
"targets" : [
"Clef"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "clef-decide",
"targets" : [
"ClefCLI"
],
"type" : {
"executable" : null
}
},
{
"name" : "gliner-bench",
"targets" : [
"GLiNERBenchmark"
],
"type" : {
"executable" : null
}
},
{
"name" : "Qwen3ASR",
"targets" : [
"Qwen3ASR"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "CohereTranscribeASR",
"targets" : [
"CohereTranscribeASR"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "VoxtralASR",
"targets" : [
"VoxtralASR"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "Qwen3TTS",
"targets" : [
"Qwen3TTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "AudioCommon",
"targets" : [
"AudioCommon"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "CosyVoiceTTS",
"targets" : [
"CosyVoiceTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "ChatterboxTTS",
"targets" : [
"ChatterboxTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "OmniVoiceTTS",
"targets" : [
"OmniVoiceTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "IndicMioTTS",
"targets" : [
"IndicMioTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "FishAudioTTS",
"targets" : [
"FishAudioTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "F5TTS",
"targets" : [
"F5TTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "BreezeTTS",
"targets" : [
"BreezeTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "HiggsTTS",
"targets" : [
"HiggsTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "IndexTTS2TTS",
"targets" : [
"IndexTTS2TTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "KyutaiSTT",
"targets" : [
"KyutaiSTT"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "PersonaPlex",
"targets" : [
"PersonaPlex"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "CSM",
"targets" : [
"CSM"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "HibikiTranslate",
"targets" : [
"HibikiTranslate"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "SpeechVAD",
"targets" : [
"SpeechVAD"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "SpeechLanguageID",
"targets" : [
"SpeechLanguageID"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "SpeechEnhancement",
"targets" : [
"SpeechEnhancement"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "SpeechRestoration",
"targets" : [
"SpeechRestoration"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "SourceSeparation",
"targets" : [
"SourceSeparation"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "CanaryASR",
"targets" : [
"CanaryASR"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "ParakeetASR",
"targets" : [
"ParakeetASR"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "ParakeetStreamingASR",
"targets" : [
"ParakeetStreamingASR"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "NemotronStreamingASR",
"targets" : [
"NemotronStreamingASR"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "VoiceChat",
"targets" : [
"VoiceChat"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "WhisperASR",
"targets" : [
"WhisperASR"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "MossTranscribe",
"targets" : [
"MossTranscribe"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "VibeVoiceTTS",
"targets" : [
"VibeVoiceTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "VoxCPM2TTS",
"targets" : [
"VoxCPM2TTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "MAGNeTMusicGen",
"targets" : [
"MAGNeTMusicGen"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "StableAudio3MusicGen",
"targets" : [
"StableAudio3MusicGen"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "FlashSR",
"targets" : [
"FlashSR"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "MagpieTTS",
"targets" : [
"MagpieTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "MagpieTTSCoreML",
"targets" : [
"MagpieTTSCoreML"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "MossTTSNano",
"targets" : [
"MossTTSNano"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "OmnilingualASR",
"targets" : [
"OmnilingualASR"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "SpeechCore",
"targets" : [
"SpeechCore"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "KokoroTTS",
"targets" : [
"KokoroTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "SupertonicTTS",
"targets" : [
"SupertonicTTS"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "Qwen3TTSCoreML",
"targets" : [
"Qwen3TTSCoreML"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "Qwen3Chat",
"targets" : [
"Qwen3Chat"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "FunctionGemma",
"targets" : [
"FunctionGemma"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "MADLADTranslation",
"targets" : [
"MADLADTranslation"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "SpeechUI",
"targets" : [
"SpeechUI"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "SpeechWakeWord",
"targets" : [
"SpeechWakeWord"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "Audio2Face3D",
"targets" : [
"Audio2Face3D"
],
"type" : {
"library" : [
"automatic"
]
}
},
{
"name" : "speech",
"targets" : [
"AudioCLI"
],
"type" : {
"executable" : null
}
},
{
"name" : "speech-server",
"targets" : [
"AudioServerCLI"
],
"type" : {
"executable" : null
}
},
{
"name" : "audio",
"targets" : [
"AudioCLI"
],
"type" : {
"executable" : null
}
},
{
"name" : "audio-server",
"targets" : [
"AudioServerCLI"
],
"type" : {
"executable" : null
}
},
{
"name" : "voicechat-bench",
"targets" : [
"VoiceChatBenchmark"
],
"type" : {
"executable" : null
}
},
{
"name" : "asr-bench",
"targets" : [
"AsrBenchmark"
],
"type" : {
"executable" : null
}
},
{
"name" : "vad-bench",
"targets" : [
"VadBenchmark"
],
"type" : {
"executable" : null
}
},
{
"name" : "diarization-bench",
"targets" : [
"DiarizationBenchmark"
],
"type" : {
"executable" : null
}
}
],
"targets" : [
{
"c99name" : "WhisperASRTests",
"module_type" : "SwiftTarget",
"name" : "WhisperASRTests",
"path" : "Tests/WhisperASRTests",
"sources" : [
"WhisperBundleCacheTests.swift"
],
"target_dependencies" : [
"WhisperASR"
],
"type" : "test"
},
{
"c99name" : "WhisperASR",
"module_type" : "SwiftTarget",
"name" : "WhisperASR",
"path" : "Sources/WhisperASR",
"product_memberships" : [
"WhisperASR",
"speech",
"speech-server",
"audio",
"audio-server",
"asr-bench"
],
"sources" : [
"WhisperASR.swift",
"WhisperByteLevelTokenizer.swift",
"WhisperCoreMLRuntime.swift",
"WhisperGenerationConfig.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "VoxtralTTSTests",
"module_type" : "SwiftTarget",
"name" : "VoxtralTTSTests",
"path" : "Tests/VoxtralTTSTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/VoxtralTTSTests/Resources",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"E2EVoxtralTTSTests.swift",
"VoxtralTTSDownloadTests.swift",
"VoxtralTTSTests.swift"
],
"target_dependencies" : [
"VoxtralTTS",
"Qwen3ASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "VoxtralTTS",
"module_type" : "SwiftTarget",
"name" : "VoxtralTTS",
"path" : "Sources/VoxtralTTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXRandom"
],
"product_memberships" : [
"VoxtralTTS",
"speech",
"speech-server",
"audio",
"audio-server"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/VoxtralTTS/LICENSE-reference",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"AcousticTransformer.swift",
"AudioDecoder.swift",
"Configuration.swift",
"FrameGenerator.swift",
"LanguageModel.swift",
"VoxtralTTSModel.swift",
"VoxtralTTSTokenizer.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon",
"SpeechRuntime"
],
"type" : "library"
},
{
"c99name" : "VoxtralASRTests",
"module_type" : "SwiftTarget",
"name" : "VoxtralASRTests",
"path" : "Tests/VoxtralASRTests",
"sources" : [
"E2EVoxtralASRTests.swift",
"VoxtralASRTests.swift"
],
"target_dependencies" : [
"VoxtralASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "VoxtralASR",
"module_type" : "SwiftTarget",
"name" : "VoxtralASR",
"path" : "Sources/VoxtralASR",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXFFT"
],
"product_memberships" : [
"VoxtralASR",
"speech",
"audio",
"asr-bench"
],
"sources" : [
"Configuration.swift",
"VoxtralASR.swift",
"VoxtralAudioEncoder.swift",
"VoxtralAudioFrontend.swift",
"VoxtralLanguageModel.swift",
"VoxtralTokenizer.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "VoxCPM2TTSTests",
"module_type" : "SwiftTarget",
"name" : "VoxCPM2TTSTests",
"path" : "Tests/VoxCPM2TTSTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"VoxCPM2MemoryTests.swift",
"VoxCPM2TTSTests.swift"
],
"target_dependencies" : [
"VoxCPM2TTS",
"AudioCommon",
"MLXCommon"
],
"type" : "test"
},
{
"c99name" : "VoxCPM2TTS",
"module_type" : "SwiftTarget",
"name" : "VoxCPM2TTS",
"path" : "Sources/VoxCPM2TTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXRandom",
"Transformers"
],
"product_memberships" : [
"VoxCPM2TTS",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"AudioVAE.swift",
"Configuration.swift",
"MiniCPM4.swift",
"VoxCPM2Memory.swift",
"VoxCPM2TTS.swift",
"VoxCPM2WeightPrecision.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "VoiceChatTests",
"module_type" : "SwiftTarget",
"name" : "VoiceChatTests",
"path" : "Tests/VoiceChatTests",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXLMCommon"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/VoiceChatTests/Resources/fleurs_en.wav",
"rule" : {
"process" : {
}
}
}
],
"sources" : [
"E2EVoiceChatConversationTests.swift",
"E2EVoiceChatDeferredInputTests.swift",
"E2EVoiceChatFunctionCallingTests.swift",
"E2EVoiceChatLiveTailTests.swift",
"E2EVoiceChatRNNTTurnTakingTests.swift",
"E2EVoiceChatRealtimeCaptionTests.swift",
"E2EVoiceChatSpeechGenerationTests.swift",
"E2EVoiceChatSpeechTests.swift",
"E2EVoiceChatToolResponsePerformanceTests.swift",
"VoiceChatSpeechTests.swift",
"VoiceChatTests.swift"
],
"target_dependencies" : [
"VoiceChat"
],
"type" : "test"
},
{
"c99name" : "VoiceChatBenchmarkTests",
"module_type" : "SwiftTarget",
"name" : "VoiceChatBenchmarkTests",
"path" : "Tests/VoiceChatBenchmarkTests",
"sources" : [
"FullDuplexToolBenchTests.swift"
],
"target_dependencies" : [
"VoiceChatBenchmark"
],
"type" : "test"
},
{
"c99name" : "VoiceChatBenchmark",
"module_type" : "SwiftTarget",
"name" : "VoiceChatBenchmark",
"path" : "Sources/VoiceChatBenchmark",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXRandom",
"ArgumentParser"
],
"product_memberships" : [
"voicechat-bench"
],
"sources" : [
"FullDuplexToolBench.swift",
"VoiceChatBench.swift"
],
"target_dependencies" : [
"AudioCommon",
"VoiceChat"
],
"type" : "executable"
},
{
"c99name" : "VoiceChat",
"module_type" : "SwiftTarget",
"name" : "VoiceChat",
"path" : "Sources/VoiceChat",
"product_dependencies" : [
"Hub",
"MLX",
"MLXNN",
"MLXFast",
"MLXFFT",
"MLXRandom",
"MLXLLM",
"MLXLMCommon",
"Tokenizers"
],
"product_memberships" : [
"VoiceChat",
"speech",
"audio",
"voicechat-bench"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/VoiceChat/Resources/rnnt_vocab.json",
"rule" : {
"process" : {
}
}
}
],
"sources" : [
"Configuration.swift",
"VoiceChatCodec.swift",
"VoiceChatEncoder.swift",
"VoiceChatLanguageModel.swift",
"VoiceChatMatrix.swift",
"VoiceChatModel.swift",
"VoiceChatNemotronH.swift",
"VoiceChatPerception.swift",
"VoiceChatSession.swift",
"VoiceChatSpeechConfiguration.swift",
"VoiceChatSpeechDecoder.swift",
"VoiceChatSpeechLayers.swift",
"VoiceChatTokenizer.swift",
"VoiceChatToolCalling.swift",
"VoiceChatTranscriber.swift",
"VoiceChatTurnTaking.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "VibeVoiceTTSTests",
"module_type" : "SwiftTarget",
"name" : "VibeVoiceTTSTests",
"path" : "Tests/VibeVoiceTTSTests",
"product_dependencies" : [
"MLX"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/VibeVoiceTTSTests/Resources/test_audio.wav",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"VibeVoiceTTSTests.swift"
],
"target_dependencies" : [
"VibeVoiceTTS",
"NemotronStreamingASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "VibeVoiceTTS",
"module_type" : "SwiftTarget",
"name" : "VibeVoiceTTS",
"path" : "Sources/VibeVoiceTTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXRandom",
"Transformers"
],
"product_memberships" : [
"VibeVoiceTTS",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"Constants.swift",
"Errors/VibeVoiceError.swift",
"Inference/DPMSolver.swift",
"Inference/KVCache.swift",
"Inference/WeightLoader.swift",
"Layers/Normalization.swift",
"Layers/StreamingConv1d.swift",
"Layers/TimestepEmbedder.swift",
"Models/AcousticTokenizerConfiguration.swift",
"Models/AcousticTokenizerDecoder.swift",
"Models/AcousticTokenizerEncoder.swift",
"Models/DiffusionHead.swift",
"Models/EOSClassifier.swift",
"Models/Qwen2Attention.swift",
"Models/Qwen2Configuration.swift",
"Models/Qwen2MLP.swift",
"Models/Qwen2Model.swift",
"Models/Qwen2TransformerBlock.swift",
"Models/SemanticTokenizer.swift",
"Models/VibeVoice15BModel.swift",
"Models/VibeVoiceStreamModel.swift",
"Quantization/VibeVoiceQuantization.swift",
"VibeVoice15BTTSModel.swift",
"VibeVoiceTTS+Protocols.swift",
"VibeVoiceTTSModel.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "VadBenchmark",
"module_type" : "SwiftTarget",
"name" : "VadBenchmark",
"path" : "Sources/VadBenchmark",
"product_dependencies" : [
"ArgumentParser"
],
"product_memberships" : [
"vad-bench"
],
"sources" : [
"VadBench.swift"
],
"target_dependencies" : [
"AudioCommon",
"SpeechVAD",
"BenchmarkSupport"
],
"type" : "executable"
},
{
"c99name" : "SupertonicTTSTests",
"module_type" : "SwiftTarget",
"name" : "SupertonicTTSTests",
"path" : "Tests/SupertonicTTSTests",
"sources" : [
"SupertonicE2ETests.swift",
"SupertonicTokenizerTests.swift"
],
"target_dependencies" : [
"SupertonicTTS",
"AudioCommon",
"Qwen3ASR"
],
"type" : "test"
},
{
"c99name" : "SupertonicTTS",
"module_type" : "SwiftTarget",
"name" : "SupertonicTTS",
"path" : "Sources/SupertonicTTS",
"product_memberships" : [
"SupertonicTTS"
],
"sources" : [
"SupertonicConfig.swift",
"SupertonicError.swift",
"SupertonicGraphs.swift",
"SupertonicTTS+Protocols.swift",
"SupertonicTTS.swift",
"SupertonicTokenizer.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "StableAudio3MusicGenTests",
"module_type" : "SwiftTarget",
"name" : "StableAudio3MusicGenTests",
"path" : "Tests/StableAudio3MusicGenTests",
"product_dependencies" : [
"MLX",
"MLXNN"
],
"sources" : [
"ConfigurationTests.swift",
"E2EGenerationTests.swift",
"WeightLoadingRandomStateTests.swift"
],
"target_dependencies" : [
"StableAudio3MusicGen",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "StableAudio3MusicGen",
"module_type" : "SwiftTarget",
"name" : "StableAudio3MusicGen",
"path" : "Sources/StableAudio3MusicGen",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXRandom"
],
"product_memberships" : [
"StableAudio3MusicGen",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"Conditioner.swift",
"Configuration.swift",
"DiTMedium.swift",
"FlowMatchingSampler.swift",
"SAMELDecoder.swift",
"StableAudio3Downloader.swift",
"StableAudio3MusicGen.swift",
"StableAudio3WeightLoading.swift",
"T5GemmaEncoder.swift",
"UnigramTokenizer.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "SpeechWakeWordTests",
"module_type" : "SwiftTarget",
"name" : "SpeechWakeWordTests",
"path" : "Tests/SpeechWakeWordTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/SpeechWakeWordTests/Resources/fbank_input.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/SpeechWakeWordTests/Resources/fbank_reference.bin",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/SpeechWakeWordTests/Resources/kws_light_up.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/SpeechWakeWordTests/Resources/kws_lovely_child.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/SpeechWakeWordTests/Resources/ref_encoder_light_up.bin",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"SpeechWakeWordTests.swift"
],
"target_dependencies" : [
"SpeechWakeWord",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "SpeechWakeWord",
"module_type" : "SwiftTarget",
"name" : "SpeechWakeWord",
"path" : "Sources/SpeechWakeWord",
"product_memberships" : [
"SpeechWakeWord",
"speech",
"audio"
],
"sources" : [
"BPETokenizer.swift",
"Configuration.swift",
"ContextGraph.swift",
"KaldiFbank.swift",
"SpeechWakeWord+Memory.swift",
"SpeechWakeWord+Protocols.swift",
"SpeechWakeWord.swift",
"StreamingKwsDecoder.swift",
"StreamingSession.swift",
"Vocabulary.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "SpeechVADTests",
"module_type" : "SwiftTarget",
"name" : "SpeechVADTests",
"path" : "Tests/SpeechVADTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"Community1DiarizationTests.swift",
"DERScoringTests.swift",
"DiarizationHelpersTests.swift",
"DiarizationPipelineTests.swift",
"E2ENemotron3DiarizationTests.swift",
"FireRedVADTests.swift",
"Nemotron3ConfigTests.swift",
"Nemotron3StreamingSessionTests.swift",
"ReDimNet2SpeakerTests.swift",
"SileroVADBenchmarkTests.swift",
"SileroVADTests.swift",
"SmartTurnTests.swift",
"SortformerTests.swift",
"SpeechVADTests.swift",
"WeSpeakerBenchmarkTests.swift",
"WeSpeakerTests.swift"
],
"target_dependencies" : [
"SpeechVAD",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "SpeechVAD",
"module_type" : "SwiftTarget",
"name" : "SpeechVAD",
"path" : "Sources/SpeechVAD",
"product_dependencies" : [
"MLX",
"MLXNN"
],
"product_memberships" : [
"Qwen3ASRRuntime",
"Qwen3ASR",
"SpeechVAD",
"speech",
"speech-server",
"audio",
"audio-server",
"asr-bench",
"vad-bench",
"diarization-bench"
],
"sources" : [
"BiLSTM.swift",
"Community1Clustering.swift",
"Community1Configuration.swift",
"Community1CoreML.swift",
"Community1DiarizationPipeline.swift",
"Configuration.swift",
"CoreMLSileroInference.swift",
"CoreMLWeSpeakerInference.swift",
"DERScoring.swift",
"DiarizationHelpers.swift",
"DiarizationPipeline.swift",
"FireRedVAD.swift",
"MelFeatureExtractor.swift",
"Nemotron3CoreMLModel.swift",
"Nemotron3Diarizer.swift",
"Nemotron3MLXModel.swift",
"PowersetDecoder.swift",
"PyannoteVAD+Memory.swift",
"ReDimNet2Speaker.swift",
"Segmentation.swift",
"SileroModel.swift",
"SileroVAD+Memory.swift",
"SileroVAD.swift",
"SileroWeightLoading.swift",
"SincNet.swift",
"SmartTurn.swift",
"SortformerConfig.swift",
"SortformerDiarizer.swift",
"SortformerMelExtractor.swift",
"SortformerModel.swift",
"SortformerStateUpdater.swift",
"SortformerStreamingSession.swift",
"SpeechVAD+Protocols.swift",
"SpeechVAD.swift",
"StreamingVADProcessor.swift",
"VADPipeline.swift",
"WeSpeaker+Memory.swift",
"WeSpeaker.swift",
"WeSpeakerModel.swift",
"WeSpeakerWeightLoading.swift",
"WeightLoading.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "SpeechUITests",
"module_type" : "SwiftTarget",
"name" : "SpeechUITests",
"path" : "Tests/SpeechUITests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/SpeechUITests/Resources/test_audio.wav",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"E2ESpeechUIWiringTests.swift",
"SpeechUITests.swift"
],
"target_dependencies" : [
"SpeechUI",
"ParakeetStreamingASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "SpeechUI",
"module_type" : "SwiftTarget",
"name" : "SpeechUI",
"path" : "Sources/SpeechUI",
"product_memberships" : [
"SpeechUI"
],
"sources" : [
"SpeechUI.swift",
"TranscriptionView.swift"
],
"type" : "library"
},
{
"c99name" : "SpeechRuntimeTests",
"module_type" : "SwiftTarget",
"name" : "SpeechRuntimeTests",
"path" : "Tests/SpeechRuntimeTests",
"sources" : [
"RuntimeOwnershipTests.swift"
],
"target_dependencies" : [
"SpeechRuntime"
],
"type" : "test"
},
{
"c99name" : "SpeechRuntime",
"module_type" : "SwiftTarget",
"name" : "SpeechRuntime",
"path" : "Sources/SpeechRuntime",
"product_memberships" : [
"PocketTTS",
"pocket-tts-bench",
"SpeechRuntime",
"VoxtralTTS",
"Qwen3ASRRuntime",
"FishAudioTTS",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"InferenceGate.swift",
"ModelCache.swift",
"ModelDescriptor.swift",
"ModelRuntime.swift"
],
"type" : "library"
},
{
"c99name" : "SpeechRestorationTests",
"module_type" : "SwiftTarget",
"name" : "SpeechRestorationTests",
"path" : "Tests/SpeechRestorationTests",
"sources" : [
"E2ESpeechRestorationTests.swift",
"SeamlessM4TFrontEndExtraTests.swift",
"SeamlessM4TFrontEndTests.swift",
"SidonModelResolverTests.swift",
"SpeechRestorationConfigTests.swift",
"SpeechRestorerLogicTests.swift"
],
"target_dependencies" : [
"SpeechRestoration",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "SpeechRestoration",
"module_type" : "SwiftTarget",
"name" : "SpeechRestoration",
"path" : "Sources/SpeechRestoration",
"product_memberships" : [
"IndexTTS2TTS",
"SpeechRestoration",
"speech",
"audio"
],
"sources" : [
"Configuration.swift",
"Downloader.swift",
"SeamlessM4TFrontEnd.swift",
"SidonModels.swift",
"SpeechRestoration+Memory.swift",
"SpeechRestoration+Protocols.swift",
"SpeechRestoration.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "SpeechLanguageIDTests",
"module_type" : "SwiftTarget",
"name" : "SpeechLanguageIDTests",
"path" : "Tests/SpeechLanguageIDTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"SpeechLanguageIDTests.swift"
],
"target_dependencies" : [
"SpeechLanguageID",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "SpeechLanguageID",
"module_type" : "SwiftTarget",
"name" : "SpeechLanguageID",
"path" : "Sources/SpeechLanguageID",
"product_dependencies" : [
"MLX",
"MLXNN"
],
"product_memberships" : [
"SpeechLanguageID",
"speech",
"audio"
],
"sources" : [
"LanguageIDTypes.swift",
"SpeechBrainECAPA.swift",
"SpeechBrainFbank.swift",
"SpeechLanguageIdentifier.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "SpeechEnhancementTests",
"module_type" : "SwiftTarget",
"name" : "SpeechEnhancementTests",
"path" : "Tests/SpeechEnhancementTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"DeepFilterNet3E2ETests.swift",
"E2EDeepFilterNet3ChunkingTests.swift",
"E2ELocalVQEAECTests.swift",
"LocalVQEAECTests.swift",
"MLXDeepFilterNet3E2ETests.swift",
"SpeechEnhancementTests.swift"
],
"target_dependencies" : [
"SpeechEnhancement",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "SpeechEnhancement",
"module_type" : "SwiftTarget",
"name" : "SpeechEnhancement",
"path" : "Sources/SpeechEnhancement",
"product_dependencies" : [
"MLX",
"MLXNN"
],
"product_memberships" : [
"SpeechEnhancement",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"AudioProcessing.swift",
"Configuration.swift",
"DeepFilterNet3Model.swift",
"LocalVQEAECAdaptiveFilter.swift",
"LocalVQEAECCodec.swift",
"LocalVQEAECResidualMask.swift",
"LocalVQEAECWeights.swift",
"LocalVQEEchoCanceller.swift",
"MLXDeepFilterNet3Network.swift",
"SpeechEnhancement+Memory.swift",
"SpeechEnhancement+Protocols.swift",
"SpeechEnhancement.swift",
"WeightLoading.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon",
"LocalVQEAECFrontend"
],
"type" : "library"
},
{
"c99name" : "SpeechCoreTests",
"module_type" : "SwiftTarget",
"name" : "SpeechCoreTests",
"path" : "Tests/SpeechCoreTests",
"sources" : [
"MemoryTierTests.swift",
"VoicePipelineTurnCompletionTests.swift"
],
"target_dependencies" : [
"SpeechCore",
"AudioCommon",
"SpeechVAD",
"KokoroTTS",
"ParakeetASR"
],
"type" : "test"
},
{
"c99name" : "SpeechCore",
"module_type" : "SwiftTarget",
"name" : "SpeechCore",
"path" : "Sources/SpeechCore",
"product_memberships" : [
"SpeechCore"
],
"sources" : [
"MemoryTier.swift",
"VoicePipeline.swift"
],
"target_dependencies" : [
"CSpeechCore",
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "SourceSeparationTests",
"module_type" : "SwiftTarget",
"name" : "SourceSeparationTests",
"path" : "Tests/SourceSeparationTests",
"product_dependencies" : [
"MLX",
"MLXRandom"
],
"sources" : [
"E2EHTDemucsTests.swift",
"E2EOpenUnmixBenchmarkTests.swift",
"HTDemucsConfigTests.swift",
"OpenUnmixTests.swift"
],
"target_dependencies" : [
"SourceSeparation",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "SourceSeparation",
"module_type" : "SwiftTarget",
"name" : "SourceSeparation",
"path" : "Sources/SourceSeparation",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast"
],
"product_memberships" : [
"SourceSeparation",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"HTDemucs/DemucsPrimitives.swift",
"HTDemucs/HTDemucs.swift",
"HTDemucs/HTDemucsConfig.swift",
"HTDemucs/HTDemucsLayers.swift",
"HTDemucs/HTDemucsSeparator.swift",
"HTDemucs/HTDemucsSpec.swift",
"HTDemucs/HTDemucsTransformer.swift",
"OpenUnmixConfig.swift",
"OpenUnmixModel.swift",
"STFT.swift",
"SourceSeparation.swift",
"SourceSeparationMetrics.swift",
"WienerFilter.swift",
"WienerFilterMLX.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "Qwen3TTSTests",
"module_type" : "SwiftTarget",
"name" : "Qwen3TTSTests",
"path" : "Tests/Qwen3TTSTests",
"sources" : [
"CodePredictorFrameTests.swift",
"DiagnosticsContractTests.swift",
"E2ELocalModelLoadingTests.swift",
"E2ETestCase.swift",
"EOSCapTests.swift",
"GenerationCancellationTests.swift",
"ICLVoiceCloningTests.swift",
"LocalModelLoadingTests.swift",
"MLXCacheLimitPolicyTests.swift",
"MemoryManagementTests.swift",
"PublicLocalModelLoadingAPITests.swift",
"Qwen3TTSTests.swift",
"ReferenceAudioCacheTests.swift",
"VoiceCloningTests.swift"
],
"target_dependencies" : [
"Qwen3TTS",
"Qwen3ASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "Qwen3TTSCoreMLTests",
"module_type" : "SwiftTarget",
"name" : "Qwen3TTSCoreMLTests",
"path" : "Tests/Qwen3TTSCoreMLTests",
"sources" : [
"BundleConfigurationTests.swift",
"E2EQwen3TTSCoreML17BTests.swift",
"E2EQwen3TTSCoreMLBenchmarkTests.swift",
"Qwen3TTSCoreMLTests.swift"
],
"target_dependencies" : [
"Qwen3TTSCoreML",
"Qwen3ASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "Qwen3TTSCoreML",
"module_type" : "SwiftTarget",
"name" : "Qwen3TTSCoreML",
"path" : "Sources/Qwen3TTSCoreML",
"product_memberships" : [
"Qwen3TTSCoreML",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"BundleConfiguration.swift",
"EmbedderModels.swift",
"MultiCodeDecoderChunked.swift",
"MultiCodeDecoderCoreML.swift",
"PromptBuilder.swift",
"Qwen3TTSCoreML+Protocols.swift",
"Qwen3TTSCoreML.swift",
"Sampler.swift",
"SpeechDecoderCoreML.swift",
"TalkerGenerator.swift",
"TalkerGeneratorChunked.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "Qwen3TTS",
"module_type" : "SwiftTarget",
"name" : "Qwen3TTS",
"path" : "Sources/Qwen3TTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast"
],
"product_memberships" : [
"Qwen3TTS",
"BreezeTTS",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"CodePredictor.swift",
"Configuration.swift",
"GenerationCancellation.swift",
"LocalModelConfiguration.swift",
"Qwen3TTS+ICL.swift",
"Qwen3TTS+Memory.swift",
"Qwen3TTS+Protocols.swift",
"Qwen3TTS.swift",
"ReferenceAudioCache.swift",
"Sampling.swift",
"SpeakerEncoder.swift",
"SpeechTokenizerDecoder.swift",
"SpeechTokenizerEncoder.swift",
"TTSWeightLoading+Encoder.swift",
"TTSWeightLoading.swift",
"Talker.swift",
"TextChunker.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "Qwen3ChatTests",
"module_type" : "SwiftTarget",
"name" : "Qwen3ChatTests",
"path" : "Tests/Qwen3ChatTests",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXRandom"
],
"sources" : [
"ChatModelConfigTests.swift",
"ChatSamplerDeviceTests.swift",
"ChatTokenizerDecodeTests.swift",
"E2EGemma4GenTests.swift",
"E2EGemma4JSONSchemaTests.swift",
"E2EGemma4KVCacheTests.swift",
"E2EGemma4ParityTests.swift",
"E2EGemma4PerfTests.swift",
"E2EGemma4SamplingParityTests.swift",
"E2EGemma4SlidingWindowTests.swift",
"E2EGemma4TokenizerParityTests.swift",
"E2EGemma4UnifiedTests.swift",
"E2EQwen35QuantizationLoadTests.swift",
"Gemma4AttentionBlockTests.swift",
"Gemma4TokenizerBPETests.swift",
"Gemma4ToolsTests.swift",
"Gemma4UnifiedConfigTests.swift",
"JSONSchemaConstraintTests.swift",
"Qwen35CoreMLChatE2ETests.swift",
"Qwen35MLXChatE2ETests.swift",
"Qwen35PipelineLLMTests.swift",
"Qwen35QuantizationTests.swift",
"Qwen3ChatTests.swift",
"Qwen3DenseGenTests.swift",
"Qwen3DenseParityTests.swift",
"ReferenceGemma4BPE.swift",
"XGrammarConstraintTests.swift"
],
"target_dependencies" : [
"Qwen3Chat",
"AudioCommon",
"MLXCommon"
],
"type" : "test"
},
{
"c99name" : "Qwen3Chat",
"module_type" : "SwiftTarget",
"name" : "Qwen3Chat",
"path" : "Sources/Qwen3Chat",
"product_dependencies" : [
"MLXLLM",
"MLX",
"MLXNN",
"MLXFast",
"MLXLMCommon"
],
"product_memberships" : [
"Clef",
"clef-decide",
"Qwen3Chat",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"ChatQuantization.swift",
"ChatResponseFormat.swift",
"ChatSampler.swift",
"ChatSamplerMLX.swift",
"ChatTemplate.swift",
"ChatTokenizer.swift",
"Gemma4Chat.swift",
"Gemma4Model.swift",
"Gemma4Tokenizer.swift",
"Gemma4Tools.swift",
"Gemma4UnifiedChat.swift",
"Gemma4WeightLoading.swift",
"JSONSchema/JSONPattern.swift",
"JSONSchema/JSONSchemaGrammar.swift",
"JSONSchema/JSONSchemaMatcher.swift",
"JSONSchema/JSONTokenConstraint.swift",
"JSONSchema/OrderedJSON.swift",
"JSONSchema/XGrammarConstraint.swift",
"MLXGenerator.swift",
"Qwen35CoreMLChat.swift",
"Qwen35Model.swift",
"Qwen35PipelineLLM.swift",
"Qwen35WeightLoading.swift",
"Qwen3ChatConfig.swift",
"Qwen3ChatError.swift",
"Qwen3DenseChat.swift",
"Qwen3DenseModel.swift",
"Qwen3DenseWeightLoading.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon",
"CXGrammarBridge"
],
"type" : "library"
},
{
"c99name" : "Qwen3ASRTests",
"module_type" : "SwiftTarget",
"name" : "Qwen3ASRTests",
"path" : "Tests/Qwen3ASRTests",
"product_dependencies" : [
"MLXRandom"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/Qwen3ASRTests/Resources/kokoro_continuous_stitched.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/Qwen3ASRTests/Resources/test_audio.wav",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"CoreMLEmbeddingExtractionTests.swift",
"E2ECoreMLASRTests.swift",
"E2ECoreMLForcedAlignerBenchTests.swift",
"E2ECoreMLForcedAlignerLocalSmokeTests.swift",
"E2ECoreMLForcedAlignerTests.swift",
"E2EDownloadOptionsLoadingTests.swift",
"E2EQwen3ASRBatchedDecodeCorrectnessTests.swift",
"E2EQwen3ASRCacheLimitLifecycleTests.swift",
"E2EQwen3ASRCancellationTests.swift",
"E2EQwen3ASRDecoderCancellationTests.swift",
"E2EQwen3ASRExperimentalBatchedDecodeTests.swift",
"E2EQwen3ASRGenerateTextMemoryTests.swift",
"E2EQwen3ASRGreedyDeterminismTests.swift",
"E2EQwen3ASRHarshAudioTests.swift",
"E2EQwen3ASRLargeHangTests.swift",
"E2EQwen3ASRLongFormTests.swift",
"ForcedAlignerTests.swift",
"HarshAudio.swift",
"PreQuantizedLinearTests.swift",
"QuantizedBatchInvarianceProbe.swift",
"Qwen3ASRBatchedDecodeTests.swift",
"Qwen3ASRIntegrationTests.swift",
"Qwen3ASRTests.swift",
"Qwen3AdaptiveDecodingTests.swift",
"Qwen3CacheLimitCoordinatorTests.swift",
"Qwen3DecodingOptionsTests.swift",
"Qwen3MemoryGuardTests.swift",
"SecurityHardeningTests.swift",
"StreamingASRTests.swift",
"ThrowawayRandomStateTests.swift"
],
"target_dependencies" : [
"Qwen3ASR",
"SpeechVAD",
"AudioCommon",
"MLXCommon"
],
"type" : "test"
},
{
"c99name" : "Qwen3ASRRuntimeTests",
"module_type" : "SwiftTarget",
"name" : "Qwen3ASRRuntimeTests",
"path" : "Tests/Qwen3ASRRuntimeTests",
"sources" : [
"E2EQwen3ASRRuntimeTests.swift",
"Qwen3ASRRuntimeTests.swift"
],
"target_dependencies" : [
"Qwen3ASRRuntime",
"Qwen3ASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "Qwen3ASRRuntime",
"module_type" : "SwiftTarget",
"name" : "Qwen3ASRRuntime",
"path" : "Sources/Qwen3ASRRuntime",
"product_memberships" : [
"Qwen3ASRRuntime",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"Qwen3ASRRuntime.swift"
],
"target_dependencies" : [
"SpeechRuntime",
"Qwen3ASR",
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "Qwen3ASR",
"module_type" : "SwiftTarget",
"name" : "Qwen3ASR",
"path" : "Sources/Qwen3ASR",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast"
],
"product_memberships" : [
"Qwen3ASRRuntime",
"Qwen3ASR",
"speech",
"speech-server",
"audio",
"audio-server",
"asr-bench"
],
"sources" : [
"AudioEncoder.swift",
"AudioPreprocessing.swift",
"Configuration.swift",
"CoreMLASRModel.swift",
"CoreMLEncoder.swift",
"CoreMLForcedAligner.swift",
"CoreMLTextDecoder.swift",
"FloatTextDecoder.swift",
"ForcedAligner+Protocols.swift",
"ForcedAligner.swift",
"QuantizedTextDecoder.swift",
"Qwen3ASR+Memory.swift",
"Qwen3ASR+Protocols.swift",
"Qwen3ASR.swift",
"Qwen3ASRCacheLimit.swift",
"StreamingASR.swift",
"TextPreprocessing.swift",
"TimestampCorrection.swift",
"WeightLoading.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon",
"SpeechVAD"
],
"type" : "library"
},
{
"c99name" : "PocketTTSTests",
"module_type" : "SwiftTarget",
"name" : "PocketTTSTests",
"path" : "Tests/PocketTTSTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/PocketTTSTests/Resources/NOTICE.txt",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/PocketTTSTests/Resources/tokenizer-fixtures.json",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/PocketTTSTests/Resources/tokenizer.model",
"rule" : {
"process" : {
}
}
}
],
"sources" : [
"E2EPocketMemoryTests.swift",
"PocketTTSTests.swift"
],
"target_dependencies" : [
"PocketTTS",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "PocketTTSBenchmark",
"module_type" : "SwiftTarget",
"name" : "PocketTTSBenchmark",
"path" : "Sources/PocketTTSBenchmark",
"product_memberships" : [
"pocket-tts-bench"
],
"sources" : [
"main.swift"
],
"target_dependencies" : [
"PocketTTS",
"KokoroTTS",
"AudioCommon"
],
"type" : "executable"
},
{
"c99name" : "PocketTTS",
"module_type" : "SwiftTarget",
"name" : "PocketTTS",
"path" : "Sources/PocketTTS",
"product_memberships" : [
"PocketTTS",
"pocket-tts-bench",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"PocketTTSConfig.swift",
"PocketTTSModel.swift",
"PocketTTSTokenizer.swift",
"PocketTtsLayerKeys.swift",
"PocketTtsMimiKeys.swift"
],
"target_dependencies" : [
"AudioCommon",
"SpeechRuntime"
],
"type" : "library"
},
{
"c99name" : "PersonaPlexTests",
"module_type" : "SwiftTarget",
"name" : "PersonaPlexTests",
"path" : "Tests/PersonaPlexTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/PersonaPlexTests/Resources/test_audio.wav",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"ModelCacheDirectoryTests.swift",
"PersonaPlexTests.swift"
],
"target_dependencies" : [
"PersonaPlex",
"AudioCommon",
"Qwen3ASR"
],
"type" : "test"
},
{
"c99name" : "PersonaPlex",
"module_type" : "SwiftTarget",
"name" : "PersonaPlex",
"path" : "Sources/PersonaPlex",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast"
],
"product_memberships" : [
"KyutaiSTT",
"PersonaPlex",
"CSM",
"HibikiTranslate",
"speech",
"speech-server",
"audio",
"audio-server",
"asr-bench"
],
"sources" : [
"Configuration.swift",
"Conv.swift",
"Depformer.swift",
"KVCache.swift",
"MimiCodec.swift",
"MimiTransformer.swift",
"PadMode.swift",
"PersonaPlex+Memory.swift",
"PersonaPlex+Protocols.swift",
"PersonaPlex.swift",
"Quantization.swift",
"Sampling.swift",
"Seanet.swift",
"SentencePieceDecoder.swift",
"TemporalTransformer.swift",
"WeightLoading.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "ParakeetStreamingASRTests",
"module_type" : "SwiftTarget",
"name" : "ParakeetStreamingASRTests",
"path" : "Tests/ParakeetStreamingASRTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/ParakeetStreamingASRTests/Resources/test_audio.wav",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"ParakeetStreamingASRTests.swift"
],
"target_dependencies" : [
"ParakeetStreamingASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "ParakeetStreamingASR",
"module_type" : "SwiftTarget",
"name" : "ParakeetStreamingASR",
"path" : "Sources/ParakeetStreamingASR",
"product_memberships" : [
"ParakeetStreamingASR",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"Configuration.swift",
"ParakeetStreamingASR+Memory.swift",
"ParakeetStreamingASR+Protocols.swift",
"ParakeetStreamingASR.swift",
"RNNTGreedyDecoder.swift",
"StreamingMelPreprocessor.swift",
"StreamingSession.swift",
"Vocabulary.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "ParakeetASRTests",
"module_type" : "SwiftTarget",
"name" : "ParakeetASRTests",
"path" : "Tests/ParakeetASRTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/ParakeetASRTests/Resources/nemo_mel_reference.json",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/ParakeetASRTests/Resources/test_audio.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/ParakeetASRTests/Resources/test_audio_german.wav",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"E2EParakeetANETests.swift",
"E2EParakeetMultiEncoderTests.swift",
"E2EParakeetShapeAdaptationTests.swift",
"MelContractTests.swift",
"ParakeetASRTests.swift"
],
"target_dependencies" : [
"ParakeetASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "ParakeetASR",
"module_type" : "SwiftTarget",
"name" : "ParakeetASR",
"path" : "Sources/ParakeetASR",
"product_memberships" : [
"ParakeetASR",
"speech",
"speech-server",
"audio",
"audio-server",
"asr-bench"
],
"sources" : [
"Configuration.swift",
"MelPreprocessor.swift",
"ParakeetASR+Memory.swift",
"ParakeetASR+Protocols.swift",
"ParakeetASR.swift",
"TDTGreedyDecoder.swift",
"Vocabulary.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "OmnilingualASRTests",
"module_type" : "SwiftTarget",
"name" : "OmnilingualASRTests",
"path" : "Tests/OmnilingualASRTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/OmnilingualASRTests/Resources/fleurs_ar.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/OmnilingualASRTests/Resources/fleurs_en.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/OmnilingualASRTests/Resources/fleurs_fr.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/OmnilingualASRTests/Resources/fleurs_hi.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/OmnilingualASRTests/Resources/test_audio.wav",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"E2EOmnilingualASRTests.swift",
"E2EOmnilingualMLX1BTests.swift",
"E2EOmnilingualMLXHarshAudioTests.swift",
"E2EOmnilingualMLXQualityTests.swift",
"E2EOmnilingualMLXTests.swift",
"HarshAudio.swift",
"OmnilingualASRTests.swift",
"OmnilingualMLXTests.swift"
],
"target_dependencies" : [
"OmnilingualASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "OmnilingualASR",
"module_type" : "SwiftTarget",
"name" : "OmnilingualASR",
"path" : "Sources/OmnilingualASR",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast"
],
"product_memberships" : [
"OmnilingualASR",
"speech",
"speech-server",
"audio",
"audio-server",
"asr-bench"
],
"sources" : [
"CTCGreedyDecoder.swift",
"Configuration.swift",
"MLX/OmnilingualASRMLXModel+Protocols.swift",
"MLX/OmnilingualMLXConfig.swift",
"MLX/OmnilingualMLXModel.swift",
"MLX/OmnilingualMLXWeightLoader.swift",
"MLX/Wav2Vec2Encoder.swift",
"MLX/Wav2Vec2EncoderLayer.swift",
"MLX/Wav2Vec2Frontend.swift",
"OmnilingualASR+Memory.swift",
"OmnilingualASR+Protocols.swift",
"OmnilingualASR.swift",
"SentencePieceVocabulary.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "OmniVoiceTTSTests",
"module_type" : "SwiftTarget",
"name" : "OmniVoiceTTSTests",
"path" : "Tests/OmniVoiceTTSTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"BackboneLoadTests.swift",
"CodecDecodeTests.swift",
"CodecEncodeTests.swift",
"EndToEndRoundtripTests.swift",
"InputConstructionTests.swift",
"OmniVoiceFullPipelineTests.swift",
"OmniVoiceInt8BundleTests.swift",
"OmniVoiceModelFacadeTests.swift",
"OmniVoiceQuantTests.swift",
"OmniVoiceRSSTests.swift",
"OmniVoiceStepSweepTests.swift",
"PrecisionGateTests.swift"
],
"target_dependencies" : [
"OmniVoiceTTS",
"AudioCommon",
"MLXCommon"
],
"type" : "test"
},
{
"c99name" : "OmniVoiceTTS",
"module_type" : "SwiftTarget",
"name" : "OmniVoiceTTS",
"path" : "Sources/OmniVoiceTTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXFFT",
"Hub",
"Tokenizers"
],
"product_memberships" : [
"OmniVoiceTTS"
],
"sources" : [
"Configuration.swift",
"OmniVoiceBackbone.swift",
"OmniVoiceCodec.swift",
"OmniVoiceCodecEncoder.swift",
"OmniVoiceDiffusion.swift",
"OmniVoiceInputBuilder.swift",
"OmniVoiceModel.swift",
"OmniVoiceTTSModel.swift",
"OmniVoiceTokenizer.swift",
"RuleDurationEstimator.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "NemotronStreamingASRTests",
"module_type" : "SwiftTarget",
"name" : "NemotronStreamingASRTests",
"path" : "Tests/NemotronStreamingASRTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/NemotronStreamingASRTests/Resources/english.json",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/NemotronStreamingASRTests/Resources/test_audio.wav",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"E2ENemotronHarshAudioTests.swift",
"E2ENemotronLangTagLeakTests.swift",
"E2ENemotronLanguageGuideTests.swift",
"EnglishNumberNormalizer.swift",
"HarshAudio.swift",
"HiBisectTests.swift",
"LanguageGuideTests.swift",
"MultilingualBenchTests.swift",
"NemotronMLXTests.swift",
"NemotronStreamingASRTests.swift",
"WhisperEnglishNormalizer.swift",
"WordBoostingTests.swift"
],
"target_dependencies" : [
"NemotronStreamingASR",
"AudioCommon",
"KokoroTTS"
],
"type" : "test"
},
{
"c99name" : "NemotronStreamingASR",
"module_type" : "SwiftTarget",
"name" : "NemotronStreamingASR",
"path" : "Sources/NemotronStreamingASR",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast"
],
"product_memberships" : [
"NemotronStreamingASR",
"speech",
"speech-server",
"audio",
"audio-server",
"asr-bench"
],
"sources" : [
"Configuration.swift",
"LanguageGuide.swift",
"Languages.swift",
"NemotronMLXConfiguration.swift",
"NemotronMLXNetwork.swift",
"NemotronMLXRuntime.swift",
"NemotronStreamingASR+Memory.swift",
"NemotronStreamingASR+Protocols.swift",
"NemotronStreamingASR.swift",
"RNNTGreedyDecoder.swift",
"SentencePieceUnigramTokenizer.swift",
"StreamingMelPreprocessor.swift",
"StreamingSession.swift",
"Vocabulary.swift",
"WordBoosting.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "MossTranscribeTests",
"module_type" : "SwiftTarget",
"name" : "MossTranscribeTests",
"path" : "Tests/MossTranscribeTests",
"product_dependencies" : [
"MLX",
"Tokenizers"
],
"sources" : [
"E2EMossMLXCacheLimitTests.swift",
"E2EMossTranscribeTests.swift",
"MossMLXCacheLimitTests.swift",
"MossTranscribeTests.swift"
],
"target_dependencies" : [
"MossTranscribe",
"AudioCommon",
"MLXCommon"
],
"type" : "test"
},
{
"c99name" : "MossTranscribe",
"module_type" : "SwiftTarget",
"name" : "MossTranscribe",
"path" : "Sources/MossTranscribe",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXLMCommon",
"Tokenizers"
],
"product_memberships" : [
"MossTranscribe",
"speech",
"audio",
"asr-bench",
"diarization-bench"
],
"sources" : [
"MossCoreMLRuntime.swift",
"MossMLXAudioModel.swift",
"MossMLXConfiguration.swift",
"MossMLXMemory.swift",
"MossMLXRuntime.swift",
"MossMLXTextModel.swift",
"MossPromptProcessor.swift",
"MossTypes.swift",
"MossWhisperFeatureExtractor.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "MossTTSNanoTests",
"module_type" : "SwiftTarget",
"name" : "MossTTSNanoTests",
"path" : "Tests/MossTTSNanoTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/MossTTSNanoTests/Resources/clone_cpu_and_gpu_codes.json",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/MossTTSNanoTests/Resources/clone_cpu_only_codes.json",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/MossTTSNanoTests/Resources/config.json",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/MossTTSNanoTests/Resources/hello_cpu_codes.json",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/MossTTSNanoTests/Resources/reference.wav",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/MossTTSNanoTests/Resources/sampling_fixtures.json",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/MossTTSNanoTests/Resources/tokenizer_fixtures.json",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/MossTTSNanoTests/Resources/tokenizer_swift.json",
"rule" : {
"process" : {
}
}
}
],
"sources" : [
"E2EMossTTSNanoTests.swift",
"MossTTSNanoTests.swift"
],
"target_dependencies" : [
"MossTTSNano",
"AudioCommon",
"Qwen3ASR"
],
"type" : "test"
},
{
"c99name" : "MossTTSNano",
"module_type" : "SwiftTarget",
"name" : "MossTTSNano",
"path" : "Sources/MossTTSNano",
"product_memberships" : [
"MossTTSNano",
"speech",
"audio"
],
"sources" : [
"MossTTSNanoConfiguration.swift",
"MossTTSNanoModel.swift",
"MossTTSNanoSampling.swift",
"MossTTSNanoTokenizer.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "MagpieTTSTests",
"module_type" : "SwiftTarget",
"name" : "MagpieTTSTests",
"path" : "Tests/MagpieTTSTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"E2EMagpieTests.swift",
"MagpieConfigTests.swift",
"MagpieSamplingTests.swift",
"MagpieTokenizerTests.swift",
"MagpieZhTokenizationTests.swift"
],
"target_dependencies" : [
"MagpieTTS",
"Qwen3ASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "MagpieTTSCoreMLTests",
"module_type" : "SwiftTarget",
"name" : "MagpieTTSCoreMLTests",
"path" : "Tests/MagpieTTSCoreMLTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/MagpieTTSCoreMLTests/Resources/magpie-hello-world.wav",
"rule" : {
"process" : {
}
}
}
],
"sources" : [
"E2EMagpieCoreMLTests.swift",
"MagpieCoreMLConfigTests.swift"
],
"target_dependencies" : [
"MagpieTTSCoreML",
"MagpieTTS",
"Qwen3ASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "MagpieTTSCoreML",
"module_type" : "SwiftTarget",
"name" : "MagpieTTSCoreML",
"path" : "Sources/MagpieTTSCoreML",
"product_dependencies" : [
"MLX",
"MLXRandom"
],
"product_memberships" : [
"MagpieTTSCoreML",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"Configuration.swift",
"CoreMLHelpers.swift",
"Decoder.swift",
"Downloader.swift",
"Errors.swift",
"FSQInverse.swift",
"LocalTransformer.swift",
"LocalTransformerWeights.swift",
"MagpieTTSCoreML.swift",
"NanoCodec.swift",
"NpyReader.swift",
"StatefulDecoder.swift",
"TextEncoder.swift",
"WeightCache.swift"
],
"target_dependencies" : [
"AudioCommon",
"MagpieTTS"
],
"type" : "library"
},
{
"c99name" : "MagpieTTS",
"module_type" : "SwiftTarget",
"name" : "MagpieTTS",
"path" : "Sources/MagpieTTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXRandom"
],
"product_memberships" : [
"MagpieTTS",
"MagpieTTSCoreML",
"speech",
"speech-server",
"audio",
"audio-server"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/MagpieTTS/Resources/cmudict_ipa_de.txt",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/MagpieTTS/Resources/cmudict_ipa_en.txt",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/MagpieTTS/Resources/cmudict_ipa_es.txt",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/MagpieTTS/Resources/cmudict_pinyin_zh.txt",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/MagpieTTS/Resources/heteronyms_en.txt",
"rule" : {
"process" : {
}
}
}
],
"sources" : [
"ByT5Encoder.swift",
"ChineseG2P.swift",
"Configuration.swift",
"Decoder.swift",
"Downloader.swift",
"EnglishG2P.swift",
"JapaneseG2P.swift",
"LocalTransformer.swift",
"MagpieTTS.swift",
"NanoCodec.swift",
"Sampling.swift",
"SubVocab.swift",
"TextEncoder.swift",
"Tokenizer.swift",
"Transformer.swift",
"WeightLoader.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "MLXCommonTests",
"module_type" : "SwiftTarget",
"name" : "MLXCommonTests",
"path" : "Tests/MLXCommonTests",
"product_dependencies" : [
"MLX",
"MLXRandom"
],
"sources" : [
"DefaultCacheLimitTests.swift",
"E2EDefaultCacheLimitTests.swift"
],
"target_dependencies" : [
"MLXCommon",
"AudioCommon",
"SpeechVAD",
"OmniVoiceTTS",
"GLiNER"
],
"type" : "test"
},
{
"c99name" : "MLXCommon",
"module_type" : "SwiftTarget",
"name" : "MLXCommon",
"path" : "Sources/MLXCommon",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXFFT"
],
"product_memberships" : [
"VoxtralTTS",
"Qwen3ASRRuntime",
"GLiNER",
"Clef",
"clef-decide",
"gliner-bench",
"Qwen3ASR",
"CohereTranscribeASR",
"VoxtralASR",
"Qwen3TTS",
"CosyVoiceTTS",
"ChatterboxTTS",
"OmniVoiceTTS",
"IndicMioTTS",
"FishAudioTTS",
"F5TTS",
"BreezeTTS",
"HiggsTTS",
"IndexTTS2TTS",
"KyutaiSTT",
"PersonaPlex",
"CSM",
"HibikiTranslate",
"SpeechVAD",
"SpeechLanguageID",
"SpeechEnhancement",
"SourceSeparation",
"NemotronStreamingASR",
"VoiceChat",
"MossTranscribe",
"VibeVoiceTTS",
"VoxCPM2TTS",
"MAGNeTMusicGen",
"StableAudio3MusicGen",
"FlashSR",
"MagpieTTS",
"MagpieTTSCoreML",
"OmnilingualASR",
"Qwen3Chat",
"MADLADTranslation",
"Audio2Face3D",
"speech",
"speech-server",
"audio",
"audio-server",
"voicechat-bench",
"asr-bench",
"vad-bench",
"diarization-bench"
],
"sources" : [
"EncodecLSTM.swift",
"MLXCacheLimit.swift",
"MetalBudget.swift",
"ModuleMemory.swift",
"PreQuantizedEmbedding.swift",
"PreQuantizedLinear.swift",
"QuantizedMLP.swift",
"SDPA.swift",
"SlaneyMel.swift",
"ThrowawayRandomState.swift",
"WeightLoading.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "MAGNeTMusicGenTests",
"module_type" : "SwiftTarget",
"name" : "MAGNeTMusicGenTests",
"path" : "Tests/MAGNeTMusicGenTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"E2EDiagnosticsTests.swift",
"E2EEncodecComparisonTests.swift",
"E2EMAGNeTPipelineTests.swift",
"MAGNeTConfigTests.swift",
"MAGNeTUtilitiesTests.swift"
],
"target_dependencies" : [
"MAGNeTMusicGen",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "MAGNeTMusicGen",
"module_type" : "SwiftTarget",
"name" : "MAGNeTMusicGen",
"path" : "Sources/MAGNeTMusicGen",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXRandom",
"Transformers"
],
"product_memberships" : [
"MAGNeTMusicGen",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"Configuration.swift",
"EncodecDecoder.swift",
"MAGNeTDownloader.swift",
"MAGNeTLM.swift",
"MAGNeTMusicGen.swift",
"Sampling.swift",
"T5Encoder.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "MADLADTranslationTests",
"module_type" : "SwiftTarget",
"name" : "MADLADTranslationTests",
"path" : "Tests/MADLADTranslationTests",
"sources" : [
"E2ELocalMADLADTests.swift",
"E2EMADLADTranslationTests.swift",
"MADLADTranslationTests.swift"
],
"target_dependencies" : [
"MADLADTranslation",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "MADLADTranslation",
"module_type" : "SwiftTarget",
"name" : "MADLADTranslation",
"path" : "Sources/MADLADTranslation",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"Transformers"
],
"product_memberships" : [
"MADLADTranslation",
"speech",
"audio"
],
"sources" : [
"MADLADConfig.swift",
"MADLADModel.swift",
"MADLADTokenizer.swift",
"MADLADTranslationError.swift",
"MADLADTranslator.swift",
"MADLADWeightLoading.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "LocalVQEAECFrontend",
"module_type" : "ClangTarget",
"name" : "LocalVQEAECFrontend",
"path" : "Sources/LocalVQEAECFrontend",
"product_memberships" : [
"SpeechEnhancement",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"LocalVQEAECFrontend.cpp"
],
"type" : "library"
},
{
"c99name" : "KyutaiSTTTests",
"module_type" : "SwiftTarget",
"name" : "KyutaiSTTTests",
"path" : "Tests/KyutaiSTTTests",
"sources" : [
"E2EKyutaiSTTDecoderReferenceTests.swift",
"E2EKyutaiSTTLongStreamTests.swift",
"E2EKyutaiSTTRoundTripTests.swift",
"E2EKyutaiSTTTests.swift",
"KyutaiSTTTests.swift"
],
"target_dependencies" : [
"KyutaiSTT",
"AudioCommon",
"PersonaPlex",
"KokoroTTS"
],
"type" : "test"
},
{
"c99name" : "KyutaiSTT",
"module_type" : "SwiftTarget",
"name" : "KyutaiSTT",
"path" : "Sources/KyutaiSTT",
"product_dependencies" : [
"MLX",
"MLXNN"
],
"product_memberships" : [
"KyutaiSTT",
"speech",
"audio",
"asr-bench"
],
"sources" : [
"Configuration.swift",
"KyutaiSTTCodec.swift",
"KyutaiSTTDecoder.swift",
"KyutaiSTTModel.swift",
"KyutaiSTTSession.swift",
"KyutaiSTTStreaming.swift",
"KyutaiSTTTokenizer.swift"
],
"target_dependencies" : [
"MLXCommon",
"AudioCommon",
"PersonaPlex"
],
"type" : "library"
},
{
"c99name" : "KokoroTTSTests",
"module_type" : "SwiftTarget",
"name" : "KokoroTTSTests",
"path" : "Tests/KokoroTTSTests",
"sources" : [
"GermanPhonemizerTests.swift",
"KokoroE2ETests.swift",
"KokoroGermanE2ETests.swift",
"KokoroTTSTests.swift",
"KokoroTextChunkerTests.swift"
],
"target_dependencies" : [
"KokoroTTS",
"AudioCommon",
"Qwen3ASR"
],
"type" : "test"
},
{
"c99name" : "KokoroTTS",
"module_type" : "SwiftTarget",
"name" : "KokoroTTS",
"path" : "Sources/KokoroTTS",
"product_memberships" : [
"pocket-tts-bench",
"KokoroTTS",
"speech",
"speech-server",
"audio",
"audio-server"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/KokoroTTS/Resources/dict_fr.json",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/KokoroTTS/Resources/dict_hi.json",
"rule" : {
"process" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/KokoroTTS/Resources/dict_pt.json",
"rule" : {
"process" : {
}
}
}
],
"sources" : [
"ChinesePhonemizer.swift",
"Configuration.swift",
"GermanPhonemizer.swift",
"HindiPhonemizer.swift",
"JapanesePhonemizer.swift",
"KokoroModel.swift",
"KokoroTTS+Memory.swift",
"KokoroTTS+Protocols.swift",
"KokoroTTS.swift",
"LatinPhonemizer.swift",
"Phonemizer.swift",
"PronunciationDicts.swift",
"TextChunker.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "IndicMioTTSTests",
"module_type" : "SwiftTarget",
"name" : "IndicMioTTSTests",
"path" : "Tests/IndicMioTTSTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"E2EIndicMioTTSTests.swift",
"IndicMioRuntimeTests.swift"
],
"target_dependencies" : [
"IndicMioTTS",
"AudioCommon",
"MLXCommon",
"Qwen3ASR"
],
"type" : "test"
},
{
"c99name" : "IndicMioTTS",
"module_type" : "SwiftTarget",
"name" : "IndicMioTTS",
"path" : "Sources/IndicMioTTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXFFT",
"Hub",
"Tokenizers"
],
"product_memberships" : [
"IndicMioTTS",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"IndicMioConfig.swift",
"IndicMioErrors.swift",
"IndicMioPrompt.swift",
"IndicMioQwen3.swift",
"IndicMioSampler.swift",
"IndicMioTTSModel.swift",
"IndicMioTokenizer.swift",
"IndicMioWavLM.swift",
"MioCodecContentDecoder.swift",
"MioCodecGlobalEncoder.swift",
"MioCodecWaveDecoder.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "IndexTTS2TTSTests",
"module_type" : "SwiftTarget",
"name" : "IndexTTS2TTSTests",
"path" : "Tests/IndexTTS2TTSTests",
"sources" : [
"E2EIndexTTS2BundleTests.swift",
"IndexTTS2KVCacheTests.swift",
"IndexTTS2TTSTests.swift",
"IndexTTS2TextNormalizerTests.swift",
"IndexTTS2TextSegmenterTests.swift"
],
"target_dependencies" : [
"IndexTTS2TTS",
"AudioCommon",
"Qwen3ASR"
],
"type" : "test"
},
{
"c99name" : "IndexTTS2TTS",
"module_type" : "SwiftTarget",
"name" : "IndexTTS2TTS",
"path" : "Sources/IndexTTS2TTS",
"product_dependencies" : [
"MLX"
],
"product_memberships" : [
"IndexTTS2TTS",
"speech",
"audio"
],
"sources" : [
"IndexTTS2BigVGAN.swift",
"IndexTTS2Bundle.swift",
"IndexTTS2CAMPPlusAdapter.swift",
"IndexTTS2Emotion.swift",
"IndexTTS2NativeRuntime.swift",
"IndexTTS2ReferenceConditioning.swift",
"IndexTTS2RuntimeConfig.swift",
"IndexTTS2S2Mel.swift",
"IndexTTS2S2MelFlow.swift",
"IndexTTS2SemanticCodec.swift",
"IndexTTS2SemanticGPT.swift",
"IndexTTS2Synthesis.swift",
"IndexTTS2TTSModel.swift",
"IndexTTS2TextNormalizer.swift",
"IndexTTS2TextSegmenter.swift",
"IndexTTS2Tokenizer.swift",
"IndexTTS2Wav2Vec2Bert.swift"
],
"target_dependencies" : [
"AudioCommon",
"ChatterboxTTS",
"MLXCommon",
"SpeechRestoration"
],
"type" : "library"
},
{
"c99name" : "HiggsTTSTests",
"module_type" : "SwiftTarget",
"name" : "HiggsTTSTests",
"path" : "Tests/HiggsTTSTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"E2EHiggsTTSTests.swift",
"HiggsTTSTests.swift"
],
"target_dependencies" : [
"HiggsTTS",
"AudioCommon",
"Qwen3ASR"
],
"type" : "test"
},
{
"c99name" : "HiggsTTS",
"module_type" : "SwiftTarget",
"name" : "HiggsTTS",
"path" : "Sources/HiggsTTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"Hub",
"Tokenizers"
],
"product_memberships" : [
"HiggsTTS",
"speech",
"audio"
],
"sources" : [
"HiggsTTSBackbone.swift",
"HiggsTTSCodec.swift",
"HiggsTTSCodecEncoder.swift",
"HiggsTTSConfig.swift",
"HiggsTTSDelayPattern.swift",
"HiggsTTSFusedCodebook.swift",
"HiggsTTSModel.swift",
"HiggsTTSPromptBuilder.swift",
"HiggsTTSTokenizer.swift",
"HiggsTTSWeightLoading.swift",
"HiggsTTSWeightMap.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "HibikiTranslateTests",
"module_type" : "SwiftTarget",
"name" : "HibikiTranslateTests",
"path" : "Tests/HibikiTranslateTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/HibikiTranslateTests/Resources/fleurs_de.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/HibikiTranslateTests/Resources/fleurs_fr.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/HibikiTranslateTests/Resources/fleurs_pt.wav",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/HibikiTranslateTests/Resources/hibiki_official_es_5s.wav",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"ConfigurationTests.swift",
"E2EHibikiTranslateTests.swift"
],
"target_dependencies" : [
"HibikiTranslate",
"AudioCommon",
"ParakeetASR",
"Qwen3TTS",
"MADLADTranslation"
],
"type" : "test"
},
{
"c99name" : "HibikiTranslate",
"module_type" : "SwiftTarget",
"name" : "HibikiTranslate",
"path" : "Sources/HibikiTranslate",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXRandom"
],
"product_memberships" : [
"HibikiTranslate",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"Configuration.swift",
"HibikiDepformer.swift",
"HibikiTemporalTransformer.swift",
"HibikiTranslate.swift",
"HibikiTranslateModel.swift",
"HibikiWeightLoading.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon",
"PersonaPlex"
],
"type" : "library"
},
{
"c99name" : "GLiNERTests",
"module_type" : "SwiftTarget",
"name" : "GLiNERTests",
"path" : "Tests/GLiNERTests",
"product_dependencies" : [
"MLX",
"MLXNN"
],
"sources" : [
"GLiNERTests.swift"
],
"target_dependencies" : [
"GLiNER",
"GLiNERBenchmark"
],
"type" : "test"
},
{
"c99name" : "GLiNERBenchmark",
"module_type" : "SwiftTarget",
"name" : "GLiNERBenchmark",
"path" : "Sources/GLiNERBenchmark",
"product_dependencies" : [
"MLX"
],
"product_memberships" : [
"gliner-bench"
],
"sources" : [
"GLiNERBench.swift",
"Options.swift"
],
"target_dependencies" : [
"GLiNER"
],
"type" : "executable"
},
{
"c99name" : "GLiNER",
"module_type" : "SwiftTarget",
"name" : "GLiNER",
"path" : "Sources/GLiNER",
"product_dependencies" : [
"Hub",
"MLX",
"MLXNN",
"Tokenizers"
],
"product_memberships" : [
"GLiNER",
"gliner-bench",
"speech",
"audio"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/GLiNER/LICENSE-reference",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"GLiNER.swift",
"GLiNERConfig.swift",
"GLiNERNetwork.swift"
],
"target_dependencies" : [
"MLXCommon",
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "FunctionGemmaTests",
"module_type" : "SwiftTarget",
"name" : "FunctionGemmaTests",
"path" : "Tests/FunctionGemmaTests",
"sources" : [
"FunctionGemmaParserTests.swift",
"FunctionGemmaPromptTests.swift"
],
"target_dependencies" : [
"FunctionGemma",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "FunctionGemma",
"module_type" : "SwiftTarget",
"name" : "FunctionGemma",
"path" : "Sources/FunctionGemma",
"product_dependencies" : [
"Hub",
"Tokenizers"
],
"product_memberships" : [
"FunctionGemma"
],
"sources" : [
"FunctionGemma.swift",
"FunctionGemmaParser.swift",
"FunctionGemmaPipelineLLM.swift",
"FunctionGemmaPrompt.swift",
"FunctionGemmaTypes.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "FlashSRTests",
"module_type" : "SwiftTarget",
"name" : "FlashSRTests",
"path" : "Tests/FlashSRTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"E2EFlashSRTests.swift",
"FlashSRConfigTests.swift"
],
"target_dependencies" : [
"FlashSR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "FlashSR",
"module_type" : "SwiftTarget",
"name" : "FlashSR",
"path" : "Sources/FlashSR",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXRandom"
],
"product_memberships" : [
"FlashSR",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"Configuration.swift",
"Downloader.swift",
"FlashSR+Protocols.swift",
"FlashSR.swift",
"Mel.swift",
"UNet.swift",
"VAE.swift",
"Vocoder.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "FishAudioTTSTests",
"module_type" : "SwiftTarget",
"name" : "FishAudioTTSTests",
"path" : "Tests/FishAudioTTSTests",
"product_dependencies" : [
"MLX"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/FishAudioTTSTests/Resources",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"E2EFishAudioCodecStreamingTests.swift",
"E2EFishAudioDrama3StreamingControlTests.swift",
"E2EFishAudioDrama3Tests.swift",
"E2EFishAudioTTSTests.swift",
"FishAudioDrama3ConfigTests.swift",
"FishAudioDrama3PromptTests.swift",
"FishAudioMemoryOptionsTests.swift",
"FishAudioRuntimeTests.swift"
],
"target_dependencies" : [
"FishAudioTTS",
"AudioCommon",
"MLXCommon",
"Qwen3ASR"
],
"type" : "test"
},
{
"c99name" : "FishAudioTTS",
"module_type" : "SwiftTarget",
"name" : "FishAudioTTS",
"path" : "Sources/FishAudioTTS",
"product_dependencies" : [
"MLXLMCommon",
"MLX",
"MLXNN",
"MLXFast",
"Hub",
"Tokenizers"
],
"product_memberships" : [
"FishAudioTTS",
"speech",
"audio"
],
"sources" : [
"FishAudioCodec.swift",
"FishAudioCodecStreaming.swift",
"FishAudioConfig.swift",
"FishAudioDrama3Config.swift",
"FishAudioDrama3Generation.swift",
"FishAudioDrama3Layers.swift",
"FishAudioDrama3Model.swift",
"FishAudioDrama3Options.swift",
"FishAudioDrama3Prompt+Tokenizer.swift",
"FishAudioDrama3Prompt.swift",
"FishAudioDrama3Streaming.swift",
"FishAudioDrama3Tower.swift",
"FishAudioDualAR.swift",
"FishAudioErrors.swift",
"FishAudioGeneration.swift",
"FishAudioInputBuilder.swift",
"FishAudioMemoryOptions.swift",
"FishAudioModelInput.swift",
"FishAudioPrompt.swift",
"FishAudioReferencePrompt.swift",
"FishAudioTTSModel.swift",
"FishAudioTokenizer.swift",
"FishAudioWeightLoading.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon",
"SpeechRuntime"
],
"type" : "library"
},
{
"c99name" : "F5TTSTests",
"module_type" : "SwiftTarget",
"name" : "F5TTSTests",
"path" : "Tests/F5TTSTests",
"product_dependencies" : [
"MLX"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/F5TTSTests/Resources/pinyin_fixtures.json",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/F5TTSTests/Resources/pinyin_lexicon_subset.tsv",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"E2EF5TTSFlowParityTests.swift",
"E2EF5TTSTests.swift",
"E2EF5TTSTransformerParityTests.swift",
"E2EF5TTSVocosParityTests.swift",
"F5TTSPinyinTests.swift",
"F5TTSTests.swift"
],
"target_dependencies" : [
"F5TTS",
"AudioCommon",
"MLXCommon",
"Qwen3ASR"
],
"type" : "test"
},
{
"c99name" : "F5TTS",
"module_type" : "SwiftTarget",
"name" : "F5TTS",
"path" : "Sources/F5TTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXFFT"
],
"product_memberships" : [
"F5TTS",
"speech",
"audio"
],
"sources" : [
"F5TTSConfig.swift",
"F5TTSFlow.swift",
"F5TTSModel.swift",
"F5TTSPinyin.swift",
"F5TTSPrimitives.swift",
"F5TTSTokenizer.swift",
"F5TTSVocos.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "DiarizationBenchmarkTests",
"module_type" : "SwiftTarget",
"name" : "DiarizationBenchmarkTests",
"path" : "Tests/DiarizationBenchmarkTests",
"sources" : [
"DiarizationReportPrinterTests.swift"
],
"target_dependencies" : [
"DiarizationBenchmark",
"BenchmarkSupport"
],
"type" : "test"
},
{
"c99name" : "DiarizationBenchmark",
"module_type" : "SwiftTarget",
"name" : "DiarizationBenchmark",
"path" : "Sources/DiarizationBenchmark",
"product_dependencies" : [
"ArgumentParser"
],
"product_memberships" : [
"diarization-bench"
],
"sources" : [
"DiarizationBench.swift"
],
"target_dependencies" : [
"AudioCommon",
"MossTranscribe",
"SpeechVAD",
"BenchmarkSupport"
],
"type" : "executable"
},
{
"c99name" : "CosyVoiceTTSTests",
"module_type" : "SwiftTarget",
"name" : "CosyVoiceTTSTests",
"path" : "Tests/CosyVoiceTTSTests",
"sources" : [
"CosyVoiceTTSTests.swift",
"DialogueParserTests.swift",
"DialogueSynthesizerTests.swift",
"FlowVocoderParityTests.swift",
"LongFormSplitTests.swift",
"MelExtractorTests.swift",
"MemoryManagementTests.swift",
"StopTokensConfigTests.swift",
"VoiceCloningTests.swift"
],
"target_dependencies" : [
"CosyVoiceTTS",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "CosyVoiceTTS",
"module_type" : "SwiftTarget",
"name" : "CosyVoiceTTS",
"path" : "Sources/CosyVoiceTTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast"
],
"product_memberships" : [
"CosyVoiceTTS",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"CamPlusPlusMelExtractor.swift",
"CamPlusPlusSpeaker.swift",
"Configuration.swift",
"CosyVoiceTTS+Memory.swift",
"CosyVoiceTTS+Protocols.swift",
"CosyVoiceTTS.swift",
"DiT.swift",
"DialogueParser.swift",
"DialogueSynthesizer.swift",
"FlowMatching.swift",
"FlowMelExtractor.swift",
"HiFiGAN.swift",
"LLM.swift",
"SpeechTokenizer.swift",
"VoiceCloning.swift",
"WeightLoading.swift",
"WhisperMelExtractor.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "CohereTranscribeASRTests",
"module_type" : "SwiftTarget",
"name" : "CohereTranscribeASRTests",
"path" : "Tests/CohereTranscribeASRTests",
"sources" : [
"CohereDownloadRetryTests.swift",
"CohereTranscribeASRTests.swift",
"E2ECohereTranscribeASRTests.swift"
],
"target_dependencies" : [
"CohereTranscribeASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "CohereTranscribeASR",
"module_type" : "SwiftTarget",
"name" : "CohereTranscribeASR",
"path" : "Sources/CohereTranscribeASR",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXFFT"
],
"product_memberships" : [
"CohereTranscribeASR",
"speech",
"audio",
"asr-bench"
],
"sources" : [
"CohereAudioFrontend.swift",
"CohereTranscribe.swift",
"CohereTranscribeDecoder.swift",
"CohereTranscribeEncoder.swift",
"CohereTranscribeTokenizer.swift",
"Configuration.swift",
"WeightLoading.swift"
],
"target_dependencies" : [
"MLXCommon",
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "ClefTests",
"module_type" : "SwiftTarget",
"name" : "ClefTests",
"path" : "Tests/ClefTests",
"product_dependencies" : [
"MLXNN",
"MLX"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/ClefTests/Fixtures",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"ClefTests.swift"
],
"target_dependencies" : [
"Clef",
"Qwen3Chat"
],
"type" : "test"
},
{
"c99name" : "ClefCLI",
"module_type" : "SwiftTarget",
"name" : "ClefCLI",
"path" : "Sources/ClefCLI",
"product_memberships" : [
"clef-decide"
],
"sources" : [
"main.swift"
],
"target_dependencies" : [
"Clef"
],
"type" : "executable"
},
{
"c99name" : "Clef",
"module_type" : "SwiftTarget",
"name" : "Clef",
"path" : "Sources/Clef",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"Hub",
"Tokenizers"
],
"product_memberships" : [
"Clef",
"clef-decide",
"speech",
"audio"
],
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Sources/Clef/LICENSE-reference",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"Clef.swift",
"ClefHead.swift",
"ClefJSONRequest.swift",
"ClefRequest.swift"
],
"target_dependencies" : [
"Qwen3Chat",
"MLXCommon",
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "ChatterboxTTSTests",
"module_type" : "SwiftTarget",
"name" : "ChatterboxTTSTests",
"path" : "Tests/ChatterboxTTSTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"ChatterboxFlashCoreMLTests.swift",
"ChatterboxMemoryOptionsTests.swift",
"E2EChatterboxFlashCoreMLTests.swift",
"MTLTokenizerTests.swift",
"MelFrontEndTests.swift",
"RoundtripE2ETests.swift",
"S3GenIntegrationTests.swift",
"T3Tests.swift",
"VoiceEncoderTests.swift"
],
"target_dependencies" : [
"ChatterboxTTS",
"AudioCommon",
"MLXCommon"
],
"type" : "test"
},
{
"c99name" : "ChatterboxTTS",
"module_type" : "SwiftTarget",
"name" : "ChatterboxTTS",
"path" : "Sources/ChatterboxTTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"Hub",
"Tokenizers"
],
"product_memberships" : [
"ChatterboxTTS",
"IndexTTS2TTS",
"speech",
"audio"
],
"sources" : [
"CAMPPlus.swift",
"ChatterboxFlashCoreMLConfig.swift",
"ChatterboxFlashCoreMLGraphs.swift",
"ChatterboxFlashCoreMLModel.swift",
"ChatterboxFlashNumpy.swift",
"ChatterboxFlashTokenizer.swift",
"ChatterboxMemoryOptions.swift",
"ChatterboxS3Gen.swift",
"ChatterboxTTSModel.swift",
"MTLTokenizer.swift",
"MatchaCFM.swift",
"S3GenConformer.swift",
"S3GenVocoder.swift",
"S3TokenizerV2.swift",
"T3.swift",
"T3Conditioning.swift",
"T3Model.swift",
"VoiceEncoder.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon"
],
"type" : "library"
},
{
"c99name" : "CanaryASRTests",
"module_type" : "SwiftTarget",
"name" : "CanaryASRTests",
"path" : "Tests/CanaryASRTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/CanaryASRTests/Resources/nemo_mel_reference.json",
"rule" : {
"copy" : {
}
}
},
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/CanaryASRTests/Resources/test_audio.wav",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"E2ECanaryTests.swift",
"MelContractTests.swift"
],
"target_dependencies" : [
"CanaryASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "CanaryASR",
"module_type" : "SwiftTarget",
"name" : "CanaryASR",
"path" : "Sources/CanaryASR",
"product_memberships" : [
"CanaryASR"
],
"sources" : [
"CanaryASR+Loading.swift",
"CanaryASR+Protocols.swift",
"CanaryASR.swift",
"Configuration.swift",
"MelPreprocessor.swift",
"Vocabulary.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "CXGrammarBridge",
"module_type" : "ClangTarget",
"name" : "CXGrammarBridge",
"path" : "Sources/CXGrammarBridge",
"product_dependencies" : [
"XGrammar"
],
"product_memberships" : [
"Clef",
"clef-decide",
"Qwen3Chat",
"speech",
"speech-server",
"audio",
"audio-server"
],
"sources" : [
"CXGrammarBridge.cpp"
],
"type" : "library"
},
{
"c99name" : "CSpeechCore",
"module_type" : "BinaryTarget",
"name" : "CSpeechCore",
"path" : "remote/archive/SpeechCore.xcframework.zip",
"product_memberships" : [
"SpeechCore"
],
"sources" : [
],
"type" : "binary"
},
{
"c99name" : "CSMTests",
"module_type" : "SwiftTarget",
"name" : "CSMTests",
"path" : "Tests/CSMTests",
"product_dependencies" : [
"MLX"
],
"sources" : [
"CSMLoadTests.swift",
"CSMMimiLoadingTests.swift",
"CSMPipelineTests.swift"
],
"target_dependencies" : [
"CSM",
"PersonaPlex"
],
"type" : "test"
},
{
"c99name" : "CSM",
"module_type" : "SwiftTarget",
"name" : "CSM",
"path" : "Sources/CSM",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"MLXRandom",
"Tokenizers",
"Hub"
],
"product_memberships" : [
"CSM",
"speech",
"audio"
],
"sources" : [
"CSMGenerate.swift",
"CSMMimi.swift",
"CSMModel.swift",
"CSMPipeline.swift",
"CSMWeightLoader.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon",
"PersonaPlex"
],
"type" : "library"
},
{
"c99name" : "BreezeTTSTests",
"module_type" : "SwiftTarget",
"name" : "BreezeTTSTests",
"path" : "Tests/BreezeTTSTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/BreezeTTSTests/Resources",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"BreezeMathTests.swift",
"BreezeTTSTests.swift",
"E2EBreezeTTSTests.swift"
],
"target_dependencies" : [
"BreezeTTS",
"Qwen3ASR",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "BreezeTTS",
"module_type" : "SwiftTarget",
"name" : "BreezeTTS",
"path" : "Sources/BreezeTTS",
"product_dependencies" : [
"MLX",
"MLXNN",
"MLXFast",
"Tokenizers"
],
"product_memberships" : [
"BreezeTTS",
"speech",
"audio"
],
"sources" : [
"BreezeTTSModel.swift",
"Codec.swift",
"Configuration.swift",
"Modules.swift",
"Options.swift",
"TextEncoder.swift"
],
"target_dependencies" : [
"AudioCommon",
"MLXCommon",
"Qwen3TTS"
],
"type" : "library"
},
{
"c99name" : "BenchmarkSupportTests",
"module_type" : "SwiftTarget",
"name" : "BenchmarkSupportTests",
"path" : "Tests/BenchmarkSupportTests",
"sources" : [
"DiarizationSpeakerCountAggregationTests.swift"
],
"target_dependencies" : [
"BenchmarkSupport",
"AudioCommon",
"SpeechVAD"
],
"type" : "test"
},
{
"c99name" : "BenchmarkSupport",
"module_type" : "SwiftTarget",
"name" : "BenchmarkSupport",
"path" : "Sources/BenchmarkSupport",
"product_memberships" : [
"vad-bench",
"diarization-bench"
],
"sources" : [
"BenchmarkSupport.swift",
"DiarizationSpeakerCountAggregation.swift"
],
"target_dependencies" : [
"AudioCommon",
"SpeechVAD"
],
"type" : "library"
},
{
"c99name" : "AudioServerTests",
"module_type" : "SwiftTarget",
"name" : "AudioServerTests",
"path" : "Tests/AudioServerTests",
"resources" : [
{
"path" : "/Users/admin/builder/spi-builder-workspace/Tests/AudioServerTests/Resources/test_audio.wav",
"rule" : {
"copy" : {
}
}
}
],
"sources" : [
"ASRRuntimeRegressionTests.swift",
"E2ENativeInferenceTests.swift",
"E2EOpenAISpeechTests.swift",
"E2EOpenAITranscriptionsTests.swift",
"E2EPocketTransportTests.swift",
"E2ERealtimeAPIEnginesTests.swift",
"E2EVoxtralTransportTests.swift",
"MultipartParserTests.swift",
"NativeInferenceTests.swift",
"OpenAISpeechTests.swift",
"PocketRoutingTests.swift",
"RealtimeAPIErrorHandlingTests.swift",
"RealtimeAPITests.swift",
"RealtimeVADTests.swift",
"TestAudioServer.swift",
"TestAudioServerTests.swift",
"VoxtralRoutingTests.swift",
"WebSocketTests.swift"
],
"target_dependencies" : [
"AudioServer",
"SpeechRuntime",
"VoxtralTTS",
"PocketTTS"
],
"type" : "test"
},
{
"c99name" : "AudioServerCLI",
"module_type" : "SwiftTarget",
"name" : "AudioServerCLI",
"path" : "Sources/AudioServerCLI",
"product_dependencies" : [
"ArgumentParser"
],
"product_memberships" : [
"speech-server",
"audio-server"
],
"sources" : [
"AudioServerCommand.swift"
],
"target_dependencies" : [
"AudioServer"
],
"type" : "executable"
},
{
"c99name" : "AudioServer",
"module_type" : "SwiftTarget",
"name" : "AudioServer",
"path" : "Sources/AudioServer",
"product_dependencies" : [
"MLX",
"Hummingbird",
"HummingbirdWebSocket",
"WSCore"
],
"product_memberships" : [
"speech-server",
"audio-server"
],
"sources" : [
"AudioServer.swift",
"ModelRegistry.swift",
"NativeInference.swift",
"NativeInferenceFailures.swift",
"OpenAISpeech.swift",
"OpenAITranscriptions.swift",
"RealtimeVAD.swift"
],
"target_dependencies" : [
"Qwen3Chat",
"WhisperASR",
"PocketTTS",
"VoxtralTTS",
"SpeechRuntime",
"Qwen3ASRRuntime",
"Qwen3ASR",
"Qwen3TTS",
"Qwen3TTSCoreML",
"CosyVoiceTTS",
"ParakeetASR",
"ParakeetStreamingASR",
"NemotronStreamingASR",
"OmnilingualASR",
"KokoroTTS",
"VoxCPM2TTS",
"IndicMioTTS",
"MagpieTTS",
"MagpieTTSCoreML",
"VibeVoiceTTS",
"PersonaPlex",
"HibikiTranslate",
"SpeechEnhancement",
"SpeechVAD",
"SourceSeparation",
"FlashSR",
"MAGNeTMusicGen",
"StableAudio3MusicGen",
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "AudioCommonTests",
"module_type" : "SwiftTarget",
"name" : "AudioCommonTests",
"path" : "Tests/AudioCommonTests",
"sources" : [
"AudioFileStreamTests.swift",
"AudioRingBufferTests.swift",
"AudioSampleRingBufferTests.swift",
"BackgroundTransferTests.swift",
"CapturedAudioChunkTests.swift",
"ConcurrencyTests.swift",
"CoreMLComputeUnitsTests.swift",
"CoreMLLoaderTests.swift",
"DefaultGenerationStreamTests.swift",
"DownloadErrorPropagationTests.swift",
"DownloadRetryTests.swift",
"DownloaderRobustnessTests.swift",
"E2EDownloadResolutionTests.swift",
"E2EModelDownloadSweepTests.swift",
"E2EModelInventoryTests.swift",
"E2EStreamingPlaybackTests.swift",
"E2ESystemAudioTapTests.swift",
"FullDuplexAudioIOTests.swift",
"HindiEmotionTTSCatalogTests.swift",
"HuggingFaceDownloaderTests.swift",
"InferenceCancellationTokenTests.swift",
"IsolatedTestRunnerTests.swift",
"MakefileTestSelectionTests.swift",
"MicrophoneResamplingTests.swift",
"ModelLoaderTests.swift",
"PipelineToolTests.swift",
"PlaybackSampleBufferTests.swift",
"ResampleRegressionTests.swift",
"ResampleTests.swift",
"SentencePieceModelTests.swift",
"ShardIsolationScriptTests.swift",
"StreamingAudioPlayerTests.swift",
"SystemAudioTapTests.swift",
"WAVWriterTests.swift"
],
"target_dependencies" : [
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "AudioCommon",
"module_type" : "SwiftTarget",
"name" : "AudioCommon",
"path" : "Sources/AudioCommon",
"product_dependencies" : [
"Hub"
],
"product_memberships" : [
"PocketTTS",
"pocket-tts-bench",
"VoxtralTTS",
"Qwen3ASRRuntime",
"GLiNER",
"Clef",
"clef-decide",
"gliner-bench",
"Qwen3ASR",
"CohereTranscribeASR",
"VoxtralASR",
"Qwen3TTS",
"AudioCommon",
"CosyVoiceTTS",
"ChatterboxTTS",
"OmniVoiceTTS",
"IndicMioTTS",
"FishAudioTTS",
"F5TTS",
"BreezeTTS",
"HiggsTTS",
"IndexTTS2TTS",
"KyutaiSTT",
"PersonaPlex",
"CSM",
"HibikiTranslate",
"SpeechVAD",
"SpeechLanguageID",
"SpeechEnhancement",
"SpeechRestoration",
"SourceSeparation",
"CanaryASR",
"ParakeetASR",
"ParakeetStreamingASR",
"NemotronStreamingASR",
"VoiceChat",
"WhisperASR",
"MossTranscribe",
"VibeVoiceTTS",
"VoxCPM2TTS",
"MAGNeTMusicGen",
"StableAudio3MusicGen",
"FlashSR",
"MagpieTTS",
"MagpieTTSCoreML",
"MossTTSNano",
"OmnilingualASR",
"SpeechCore",
"KokoroTTS",
"SupertonicTTS",
"Qwen3TTSCoreML",
"Qwen3Chat",
"FunctionGemma",
"MADLADTranslation",
"SpeechWakeWord",
"Audio2Face3D",
"speech",
"speech-server",
"audio",
"audio-server",
"voicechat-bench",
"asr-bench",
"vad-bench",
"diarization-bench"
],
"sources" : [
"AudioFileLoader.swift",
"AudioFileStream.swift",
"AudioIO.swift",
"AudioModelError.swift",
"AudioRingBuffer.swift",
"CoreMLComputeUnits.swift",
"CoreMLLoader.swift",
"DownloadOptions.swift",
"FullDuplexAudioIO.swift",
"HindiEmotionTTSCatalog.swift",
"HuggingFaceBackgroundTransfer.swift",
"HuggingFaceDownloader.swift",
"HuggingFaceRepoManifest.swift",
"HuggingFaceTransfer.swift",
"InferenceCancellationToken.swift",
"Logging.swift",
"ModelLoader.swift",
"PipelineLLM.swift",
"PlaybackSampleBuffer.swift",
"Protocols.swift",
"SentencePieceModel.swift",
"StreamingAudioPlayer.swift",
"SystemAudioTap.swift",
"Tokenizer.swift",
"WAVWriter.swift"
],
"type" : "library"
},
{
"c99name" : "AudioCLITests",
"module_type" : "SwiftTarget",
"name" : "AudioCLITests",
"path" : "Tests/AudioCLITests",
"product_dependencies" : [
"ArgumentParser"
],
"sources" : [
"AudioCLITests.swift",
"BreezeSpeakCommandTests.swift",
"ClefCommandTests.swift",
"E2EBreezeCLITests.swift",
"E2EPocketCLITests.swift",
"ErrorHandlingTests.swift",
"FishDrama3SpeakCommandTests.swift",
"GLiNERCommandTests.swift",
"KyutaiSTTCommandTests.swift",
"LanguageIDCommandTests.swift",
"MossNanoCommandTests.swift",
"PocketSpeakCommandTests.swift",
"Qwen3TTSCoreMLCommandTests.swift",
"RestoreCommandTests.swift",
"SendableTests.swift",
"VoiceChatCommandTests.swift",
"VoxtralSpeakCommandTests.swift"
],
"target_dependencies" : [
"AudioCLILib",
"GLiNER"
],
"type" : "test"
},
{
"c99name" : "AudioCLILib",
"module_type" : "SwiftTarget",
"name" : "AudioCLILib",
"path" : "Sources/AudioCLILib",
"product_dependencies" : [
"MLX",
"MLXRandom",
"ArgumentParser",
"MCP",
"SystemPackage"
],
"product_memberships" : [
"speech",
"audio"
],
"sources" : [
"AlignCommand.swift",
"AudioCLI.swift",
"AudioTranslateCommand.swift",
"AvatarMotionCommand.swift",
"CSMCommand.swift",
"ClefCommand.swift",
"ComposeCommand.swift",
"DenoiseCommand.swift",
"DiarizeCommand.swift",
"EmbedSpeakerCommand.swift",
"GLiNERCommand.swift",
"KokoroCommand.swift",
"LanguageIDCommand.swift",
"Qwen3TTSCoreMLCommand.swift",
"RespondCommand.swift",
"RestoreCommand.swift",
"SA3ParityCommand.swift",
"SeparateCommand.swift",
"SpeakCommand+MossNano.swift",
"SpeakCommand.swift",
"TranscribeBatchCommand.swift",
"TranscribeCommand.swift",
"TranslateCommand.swift",
"TurnCommand.swift",
"UpsampleCommand.swift",
"Utilities.swift",
"VadCommand.swift",
"VadStreamCommand.swift",
"VibeVoiceCommand.swift",
"VibeVoiceEncodeCommand.swift",
"VoiceChatCommand.swift",
"VoiceChatDemoSupport.swift",
"VoiceChatMCP.swift",
"WakeCommand.swift"
],
"target_dependencies" : [
"MLXCommon",
"FishAudioTTS",
"PocketTTS",
"VoxtralTTS",
"SpeechRuntime",
"Qwen3ASRRuntime",
"CohereTranscribeASR",
"KyutaiSTT",
"Qwen3ASR",
"Qwen3TTS",
"CosyVoiceTTS",
"Qwen3TTSCoreML",
"PersonaPlex",
"CSM",
"HibikiTranslate",
"SpeechVAD",
"SpeechLanguageID",
"SpeechEnhancement",
"SpeechRestoration",
"SourceSeparation",
"ParakeetASR",
"ParakeetStreamingASR",
"NemotronStreamingASR",
"WhisperASR",
"MossTranscribe",
"VoiceChat",
"OmnilingualASR",
"VoxtralASR",
"KokoroTTS",
"VibeVoiceTTS",
"VoxCPM2TTS",
"F5TTS",
"HiggsTTS",
"BreezeTTS",
"IndexTTS2TTS",
"IndicMioTTS",
"MAGNeTMusicGen",
"StableAudio3MusicGen",
"FlashSR",
"MagpieTTS",
"MagpieTTSCoreML",
"MADLADTranslation",
"MossTTSNano",
"GLiNER",
"Clef",
"SpeechWakeWord",
"Audio2Face3D",
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "AudioCLI",
"module_type" : "SwiftTarget",
"name" : "AudioCLI",
"path" : "Sources/AudioCLI",
"product_memberships" : [
"speech",
"audio"
],
"sources" : [
"main.swift"
],
"target_dependencies" : [
"AudioCLILib"
],
"type" : "executable"
},
{
"c99name" : "Audio2Face3DTests",
"module_type" : "SwiftTarget",
"name" : "Audio2Face3DTests",
"path" : "Tests/Audio2Face3DTests",
"sources" : [
"Audio2Face3DParityTests.swift",
"Audio2Face3DTests.swift",
"E2EAudio2Face3DTests.swift"
],
"target_dependencies" : [
"Audio2Face3D",
"AudioCommon"
],
"type" : "test"
},
{
"c99name" : "Audio2Face3D",
"module_type" : "SwiftTarget",
"name" : "Audio2Face3D",
"path" : "Sources/Audio2Face3D",
"product_dependencies" : [
"MLX"
],
"product_memberships" : [
"Audio2Face3D",
"speech",
"audio"
],
"sources" : [
"Audio2Face3DDownloader.swift",
"Audio2Face3DFrame.swift",
"Audio2Face3DMLXRuntime.swift",
"Audio2Face3DModel.swift",
"Configuration.swift"
],
"target_dependencies" : [
"MLXCommon",
"AudioCommon"
],
"type" : "library"
},
{
"c99name" : "AsrBenchmarkTests",
"module_type" : "SwiftTarget",
"name" : "AsrBenchmarkTests",
"path" : "Tests/AsrBenchmarkTests",
"sources" : [
"AsrBenchmarkTests.swift"
],
"target_dependencies" : [
"AsrBenchmark"
],
"type" : "test"
},
{
"c99name" : "AsrBenchmark",
"module_type" : "SwiftTarget",
"name" : "AsrBenchmark",
"path" : "Sources/AsrBenchmark",
"product_dependencies" : [
"MLX",
"WhisperKit",
"ArgumentParser"
],
"product_memberships" : [
"asr-bench"
],
"sources" : [
"AsrBench.swift",
"Dataset.swift",
"Engine.swift",
"EngineCohereVoxtralMLX.swift",
"EngineKyutaiSTT.swift",
"EngineMoss.swift",
"EngineNemotron.swift",
"EngineOmnilingual.swift",
"EngineOmnilingualMLX.swift",
"EngineParakeet.swift",
"EngineQwen3CoreML.swift",
"EngineQwen3MLX.swift",
"EngineWhisperASR.swift",
"EngineWhisperKit.swift",
"Memory.swift",
"Normalizer.swift",
"Report.swift",
"WER.swift"
],
"target_dependencies" : [
"AudioCommon",
"KyutaiSTT",
"CohereTranscribeASR",
"Qwen3ASR",
"ParakeetASR",
"NemotronStreamingASR",
"OmnilingualASR",
"WhisperASR",
"VoxtralASR",
"MossTranscribe"
],
"type" : "executable"
}
],
"tools_version" : "5.10"
}
Done.