"""TTS client implementations — one module per backend server.

Public API: the client classes (QwenTTSClient, FasterTTSClient,
AudioCppTTSClient, SgOmniTTSClient), the backend/voice-mode vocabulary,
and the shared helpers (normalize_language, speaker tables, whisper
transcription) that the UIs and setup wizards build on.
"""

# The TTS backends a conversion can use, in Convert-form order. Each has a
# client module here; the backends package mirrors these keys for its
# install/setup wizards.
BACKEND_QWEN = "qwen"
BACKEND_FASTER = "faster"
BACKEND_AUDIOCPP = "audiocpp"
BACKEND_SGLOMNI = "sglomni"
BACKENDS = (BACKEND_AUDIOCPP, BACKEND_QWEN, BACKEND_FASTER, BACKEND_SGLOMNI)

from .base import BaseTTSClient, ConversionCancelled, VOICE_MODE_CLONE, \
    VOICE_MODE_CUSTOM, VOICE_MODE_DESIGN, VOICE_MODES, resolve_request_seed
from .languages import LANGUAGE_CHOICES, LANGUAGE_ISO_CODES, TTS_LANGUAGES, \
    TTS_LANGUAGE_ALIASES, normalize_language
from .speakers import QWEN3_TTS_SPEAKERS, SPEAKER_DISPLAY_NAMES, \
    is_builtin_speaker, speaker_display_name_for
from .transcribe import (transcribe_reference_audio,
                         transcribe_reference_audio_detailed,
                         whisper_backend_available, whisper_backend_problem)
from .qwen import CUSTOM_VOICE_MODEL_ID, MODEL_SIZE, QwenTTSClient
from .faster import SAMPLE_RATE, FasterTTSClient
from .sglomni import SgOmniTTSClient
from .audiocpp import (
    AUDIOCPP_CLONE_ONLY_FAMILIES,
    AUDIOCPP_DEFAULT_FAMILY_PROFILE,
    AUDIOCPP_FAMILY_PROFILES,
    AUDIOCPP_FAMILY_QWEN3_TTS,
    AUDIOCPP_LANG_DISPLAY,
    AUDIOCPP_LANG_ISO,
    AUDIOCPP_LANG_OMIT,
    AUDIOCPP_SYNTHESIS_TASKS,
    AUDIOCPP_TASK_TTS,
    AUDIOCPP_TASK_VDES,
    AUDIOCPP_VOICE_CLONE,
    AUDIOCPP_VOICE_DESIGN,
    AUDIOCPP_VOICE_NONE,
    AUDIOCPP_VOICE_OPTIONAL,
    AUDIOCPP_VOICE_REQUIRED,
    AUDIOCPP_VOICE_SPEAKER,
    AudioCppFamilyProfile,
    AudioCppTTSClient,
    audiocpp_entry_supports_design,
    audiocpp_entry_voice_capability,
    audiocpp_family_narrates,
    audiocpp_family_spec_tasks,
    audiocpp_family_voice_policy,
    audiocpp_request_error,
    audiocpp_script_input,
    audiocpp_voice_for_run,
    allocation_log_note,
    nvidia_device_memory_report,
)

__all__ = [
    # vocabulary
    "BACKEND_QWEN", "BACKEND_FASTER", "BACKEND_AUDIOCPP", "BACKEND_SGLOMNI",
    "BACKENDS",
    "VOICE_MODE_CUSTOM", "VOICE_MODE_CLONE", "VOICE_MODE_DESIGN", "VOICE_MODES",
    # clients
    "BaseTTSClient", "ConversionCancelled", "resolve_request_seed",
    "QwenTTSClient", "FasterTTSClient", "AudioCppTTSClient",
    "SgOmniTTSClient",
    # model facts
    "MODEL_SIZE", "CUSTOM_VOICE_MODEL_ID", "SAMPLE_RATE",
    # languages
    "TTS_LANGUAGES", "TTS_LANGUAGE_ALIASES", "LANGUAGE_ISO_CODES",
    "LANGUAGE_CHOICES", "normalize_language",
    # speakers
    "QWEN3_TTS_SPEAKERS", "SPEAKER_DISPLAY_NAMES",
    "speaker_display_name_for", "is_builtin_speaker",
    # transcription
    "transcribe_reference_audio", "transcribe_reference_audio_detailed",
    "whisper_backend_available", "whisper_backend_problem",
    # audio.cpp family profiles
    "AUDIOCPP_LANG_DISPLAY", "AUDIOCPP_LANG_ISO", "AUDIOCPP_LANG_OMIT",
    "AUDIOCPP_FAMILY_QWEN3_TTS", "AUDIOCPP_TASK_TTS", "AUDIOCPP_TASK_VDES",
    "AUDIOCPP_SYNTHESIS_TASKS", "AUDIOCPP_VOICE_SPEAKER",
    "AUDIOCPP_VOICE_CLONE", "AUDIOCPP_VOICE_DESIGN",
    "AUDIOCPP_CLONE_ONLY_FAMILIES", "AUDIOCPP_VOICE_REQUIRED",
    "AUDIOCPP_VOICE_OPTIONAL", "AUDIOCPP_VOICE_NONE",
    "AudioCppFamilyProfile", "AUDIOCPP_DEFAULT_FAMILY_PROFILE",
    "AUDIOCPP_FAMILY_PROFILES", "audiocpp_entry_voice_capability",
    "audiocpp_entry_supports_design", "audiocpp_voice_for_run",
    "audiocpp_family_narrates",
    "audiocpp_family_spec_tasks", "audiocpp_family_voice_policy",
    "audiocpp_request_error", "audiocpp_script_input",
    "allocation_log_note", "nvidia_device_memory_report",
    "AUDIOCPP_VOICE_REQUIRED_FAMILIES", "AUDIOCPP_ALLOCATION_FRAGMENTS",
]