from __future__ import annotations


DIARIZATION_MODELS: dict[str, dict] = {
    "pyannote-3.1": {
        "name": "pyannote speaker-diarization 3.1",
        "repo_id": "pyannote/speaker-diarization-3.1",
        "backend": "pyannote",
    },
    "nemo-msdd": {
        "name": "NeMo MSDD (экспериментальный)",
        "backend": "nemo-msdd",
        "config": "diar_msdd_telephonic",
    },
}

TRANSCRIPTION_MODELS: dict[str, dict] = {
    # --- faster-whisper (CTranslate2) ---
    "whisper-large-v3": {
        "name": "Whisper large-v3",
        "repo_id": "Systran/faster-whisper-large-v3",
        "backend": "faster-whisper",
        "compute_type": "float16",
        "language": "ru",
    },
    "whisper-large-v3-turbo": {
        "name": "Whisper large-v3-turbo",
        "repo_id": "Systran/faster-whisper-large-v3-turbo",
        "backend": "faster-whisper",
        "compute_type": "float16",
        "language": "ru",
    },
    # --- transformers-whisper (HF native, Russian fine-tuned) ---
    "whisper-podlodka-turbo": {
        "name": "Whisper Podlodka Turbo (RU)",
        "repo_id": "bond005/whisper-podlodka-turbo",
        "backend": "transformers-whisper",
        "language": "ru",
    },
    "whisper-large-v3-russian": {
        "name": "Whisper large-v3 Russian",
        "repo_id": "antony66/whisper-large-v3-russian",
        "backend": "transformers-whisper",
        "language": "ru",
    },
    "whisper-russian-ties-podlodka": {
        "name": "Whisper TIES-merge Podlodka (RU)",
        "repo_id": "Apel-sin/whisper-large-v3-russian-ties-podlodka-v1.2",
        "backend": "transformers-whisper",
        "language": "ru",
    },
    # --- GigaAM ---
    "gigaam-ctc": {
        "name": "GigaAM CTC",
        "backend": "gigaam",
        "model_type": "ctc",
        "language": "ru",
    },
    "gigaam-rnnt": {
        "name": "GigaAM RNNT",
        "backend": "gigaam",
        "model_type": "rnnt",
        "language": "ru",
    },
}


def diarization_choices() -> list[tuple[str, str]]:
    return [(v["name"], k) for k, v in DIARIZATION_MODELS.items()]


def transcription_choices() -> list[tuple[str, str]]:
    return [(v["name"], k) for k, v in TRANSCRIPTION_MODELS.items()]