car-voice 0.13.0

Voice I/O capability for CAR — mic capture, VAD, listener/speaker traits
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167

//! Voice I/O capability for CAR.
//!
//! Channel-neutral microphone capture, voice activity detection,
//! speech-to-text, text-to-speech, and audio playback. Any CAR-based
//! agent or channel (CLI, GUI, IDE plug-in) can consume this crate
//! without pulling in a UI shell.
//!
//! ## Module map
//!
//! - [`config`] — `VoiceConfig` (provider selection, VAD tuning, mode).
//! - [`error`] — `VoiceError`.
//! - [`events`] — `VoiceEvent` enum (SpeechStart/End/Transcript/BargeIn).
//! - [`stt`] — `SttProvider` trait for speech-to-text backends.
//! - [`tts`] — `Speaker` trait + raw playback helper.
//! - [`provider`] — Factory functions that build the right STT/TTS from config.
//! - [`elevenlabs_stt`] / [`elevenlabs_tts`] — ElevenLabs cloud providers.
//! - [`whisper_cpp_stt`] — In-process Whisper STT via whisper.cpp (Metal on Apple Silicon)
//! - [`local_tts`] — Local OpenAI-compatible TTS provider
//!   (e.g. MLX-Whisper, mlx-audio Kokoro/Qwen3-TTS).
//! - [`listener`] — `Listener` trait + cross-platform `CpalListener`.
//! - [`voice_processing_listener`] — macOS `VoiceProcessingIO` listener with
//!   hardware AEC, AGC, and barge-in support.
//! - [`voice_audio_mixer`] — software mixer that feeds VPIO's bus 0
//!   reference signal so AEC has something to subtract.
//! - [`vad`] — energy-based voice activity detection with adaptive noise
//!   floor and runtime threshold boost (used by the listener for barge-in).
//! - [`enrollment`] — speaker voiceprint enrollment + per-segment role
//!   classification (`TranscriptRole`, `Enrollment`, `SpeakerPipeline`).
//! - [`narration`] — TARS-style commentary helpers, pure functions.

pub mod config;
pub mod cpal_listener;
mod runtime;
pub(crate) use runtime::voice_runtime_handle;
pub mod elevenlabs_streaming_listener;
pub mod elevenlabs_stt;
pub mod elevenlabs_tts;
pub mod enrollment;
pub mod enrollment_store;
pub mod error;
pub mod events;
pub mod listener;
pub mod local_tts;
// Pure-function classification + bridge-phrase helpers used by both
// the macOS-only orchestrator and the cross-platform FFI dispatch
// helper (`car-ffi-common::voice_turn`). No platform deps.
pub mod utterance;
// Orchestrator owns a `VoiceMixerHandle` for cancellable TTS, which
// is itself gated to `target_os = "macos"` (the mixer feeds VPIO's
// bus 0). Gating the orchestrator to match keeps the workspace
// buildable on Linux/Windows. The two-track voice loop runs on
// macOS today; iOS reaches the orchestrator via the macOS-shared
// dispatch surface in `car-engine::voice_turn`, which is itself
// platform-agnostic.
/// SFSpeechRecognizer-backed STT — macOS 10.15+. On-device, free,
/// multilingual, no model download. See macOS Apple-frameworks proposal.
#[cfg(target_os = "macos")]
pub mod apple_speech_stt;
/// AVSpeechSynthesizer-backed TTS — macOS only, no model download,
/// no MLX dependency. See `docs/proposals/macos-apple-frameworks.md`.
#[cfg(target_os = "macos")]
pub mod apple_speech_tts;
#[cfg(feature = "diarization")]
pub mod diarization;
/// In-process Kokoro-82M TTS via MLX (Apple Silicon macOS only —
/// internally uses `car_inference::backend::mlx_kokoro` which is
/// itself gated on aarch64, so x86_64 Mac must not compile this).
#[cfg(all(target_os = "macos", target_arch = "aarch64", not(car_skip_mlx)))]
pub mod kokoro_tts;
pub mod narration;
#[cfg(target_os = "macos")]
pub mod orchestrator;
#[cfg(feature = "parakeet")]
pub mod parakeet_stt;
pub mod provider;
pub mod providers_info;
pub mod push_listener;
pub mod session;
pub mod streaming_whisper;
pub mod stt;
#[cfg(all(target_os = "macos", feature = "system-audio-macos"))]
pub mod system_audio_capture;
#[cfg(all(target_os = "macos", feature = "system-audio-macos"))]
pub mod system_audio_listener;
#[cfg(target_os = "linux")]
pub mod system_audio_listener_linux;
#[cfg(target_os = "windows")]
pub mod system_audio_listener_windows;
pub mod transcript;
pub mod tts;
pub mod vad;
pub mod vad_backend;
#[cfg(target_os = "macos")]
pub mod voice_audio_mixer;
#[cfg(target_os = "macos")]
pub mod voice_processing_listener;
pub mod whisper_cpp_stt;

pub use config::{
    compose_voice_context, voice_config_watch, ListenerMode, SttProvider as SttProviderKind,
    TtsProvider as TtsProviderKind, VoiceConfig, VoiceConfigHandle, VoiceConfigSender,
    DEFAULT_VOICE_PROMPT_OVERLAY,
};
// Cross-platform re-exports of the utterance helpers — every consumer
// of car-voice (incl. car-ffi-common on Linux/Windows) gets these
// without needing to reach into the macOS-only orchestrator.
#[cfg(target_os = "macos")]
pub use orchestrator::VoiceOrchestrator;
pub use utterance::{
    bridge_phrase, classify_utterance, format_for_voice, ToolKind, UtteranceClass,
};
// Re-export VoiceTelemetry + DirectDataFetcher so callers don't need a
// direct car-engine dep.
#[cfg(target_os = "macos")]
pub use apple_speech_stt::AppleSpeechSttProvider;
#[cfg(target_os = "macos")]
pub use apple_speech_tts::AppleSpeechSpeaker;
pub use car_engine::{DirectDataFetcher, VoiceTelemetry};
pub use cpal_listener::CpalListener;
#[cfg(feature = "diarization")]
pub use diarization::{
    DiarizationConfig, DiarizationError, Diarizer, SharedDiarizer, SpeakerDiarizer,
};
pub use elevenlabs_streaming_listener::ElevenLabsStreamingListener;
pub use elevenlabs_stt::ElevenLabsSttProvider;
pub use elevenlabs_tts::{ElevenLabsSpeaker, VoiceSettings};
pub use enrollment::{
    Enrollment, EnrollmentOutcome, FilterbankEmbedder, PendingEnrollment, SpeakerEmbedder,
    SpeakerEmbedding, SpeakerPipeline, TranscriptRole,
};
pub use enrollment_store::{
    enroll_from_pcm, enroll_from_wav, enrollment_dir, enrollment_path, list_enrollments,
    load_enrollment, remove_enrollment, save_enrollment, EnrollmentInfo,
};
pub use error::VoiceError;
pub use events::VoiceEvent;
#[cfg(all(target_os = "macos", target_arch = "aarch64", not(car_skip_mlx)))]
pub use kokoro_tts::KokoroSpeaker;
pub use listener::Listener;
pub use local_tts::LocalTtsSpeaker;
#[cfg(feature = "parakeet")]
pub use parakeet_stt::{ParakeetPartial, ParakeetSttProvider};
pub use provider::{build_stt_provider, build_tts_speaker};
pub use providers_info::{
    list_voice_providers, list_voice_providers_json, VoiceProviderInfo, VoiceProviderKind,
};
pub use push_listener::{PushHandle, PushListener};
pub use session::{VoiceEventSink, VoiceSession, VoiceSessionRegistry};
pub use streaming_whisper::{ChunkOverlapStreamer, StreamingConfig, StreamingPartial};
pub use stt::SttProvider;
#[cfg(all(target_os = "macos", feature = "system-audio-macos"))]
pub use system_audio_listener::SystemAudioListener;
#[cfg(target_os = "linux")]
pub use system_audio_listener_linux::{default_monitor_source, list_monitor_sources};
#[cfg(target_os = "windows")]
pub use system_audio_listener_windows::WindowsLoopbackListener;
pub use transcript::{parse_transcript, ParsedTranscript};
pub use tts::{play_audio, AudioFormat, Speaker, SynthesizedAudio};
pub use vad_backend::{EnergyVadBackend, SileroError, SileroVadBackend, VadBackend};
#[cfg(target_os = "macos")]
pub use voice_audio_mixer::{StemRole as MixerStemRole, VoiceAudioMixer, VoiceMixerHandle};
#[cfg(target_os = "macos")]
pub use voice_processing_listener::VoiceProcessingListener;
pub use whisper_cpp_stt::WhisperCppSttProvider;

/// Convenience result alias used throughout the crate.
pub type Result<T> = std::result::Result<T, VoiceError>;