"""Validation for the client-supplied parts of the Deepgram request URL.

These values are interpolated into a URL authenticated with our Deepgram key,
so they are allowlisted rather than escaped. Escaping would stop a client
appending extra parameters with `&`, but would still let it choose any model —
and `_log_stt_usage` only prices nova-3 and nova-2, so any other model would be
logged at nova-3's rate and quietly under-count the daily budget.

Unknown values fall back to the default rather than rejecting the connection:
an unusual browser sample rate should not break the microphone.

The language lists are Deepgram's, per model:
https://developers.deepgram.com/docs/models-languages-overview
They are the same lists the dropdown in static/transcribe2/index.html offers —
`tests/test_transcribe2.py` fails if the two drift apart. A language nova-2
cannot do falls back to the default instead of failing the connection with a
Deepgram error the page cannot explain.
"""

# What the UI actually offers — static/transcribe2/index.html.
MODELS = frozenset({"nova-3", "nova-2"})

# "multi" is code-switching: 10 languages on nova-3, Spanish+English on nova-2.
NOVA3_LANGUAGES = frozenset({
    "multi", "af", "af-ZA", "ar", "ar-AE", "ar-SA", "ar-QA", "ar-KW",
    "ar-SY", "ar-LB", "ar-PS", "ar-JO", "ar-EG", "ar-SD", "ar-TD", "ar-MA",
    "ar-DZ", "ar-TN", "ar-IQ", "ar-IR", "hy", "as", "as-IN", "be",
    "bn", "bs", "bg", "ca", "zh-HK", "zh", "zh-CN", "zh-Hans",
    "zh-TW", "zh-Hant", "hr", "cs", "cs-CZ", "da", "da-DK", "nl",
    "en", "en-US", "en-AU", "en-GB", "en-IN", "en-NZ", "et", "fi",
    "nl-BE", "fr", "fr-CA", "ka", "ka-GE", "de", "de-CH", "el",
    "gu", "gu-IN", "he", "hi", "hu", "id", "it", "ja",
    "kn", "kk", "kk-KZ", "ko", "ko-KR", "lv", "lt", "mk",
    "ms", "mr", "mn", "ne", "no", "ps", "ps-AF", "fa",
    "pl", "pt", "pt-BR", "pt-PT", "pa", "pa-IN", "ro", "ru",
    "sr", "sk", "sl", "es", "es-419", "sv", "sv-SE", "tl",
    "ta", "te", "th", "th-TH", "tr", "tr-TR", "uk", "ur",
    "vi",
})

NOVA2_LANGUAGES = frozenset({
    "multi", "bg", "ca", "zh", "zh-CN", "zh-Hans", "zh-TW", "zh-Hant",
    "zh-HK", "cs", "da", "da-DK", "nl", "en", "en-US", "en-AU",
    "en-GB", "en-NZ", "en-IN", "et", "fi", "nl-BE", "fr", "fr-CA",
    "de", "de-CH", "el", "hi", "hu", "id", "it", "ja",
    "ko", "ko-KR", "lv", "lt", "ms", "no", "pl", "pt",
    "pt-BR", "pt-PT", "ro", "ru", "sk", "es", "es-419", "sv",
    "sv-SE", "th", "th-TH", "tr", "uk", "vi",
})

LANGUAGES = NOVA3_LANGUAGES | NOVA2_LANGUAGES

DEFAULT_MODEL = "nova-3"
DEFAULT_LANGUAGE = "cs"


def model(raw: str | None) -> str:
    return raw if raw in MODELS else DEFAULT_MODEL


def language(raw: str | None, for_model: str = DEFAULT_MODEL) -> str:
    """Validate a language code against the model that has to transcribe it."""
    supported = NOVA2_LANGUAGES if for_model == "nova-2" else NOVA3_LANGUAGES
    return raw if raw in supported else DEFAULT_LANGUAGE


def bounded_int(raw: str | None, default: int, low: int, high: int) -> int:
    """Coerce a query parameter to an int within [low, high], else the default."""
    try:
        value = int(raw)  # type: ignore[arg-type]  # None/junk raises, caught here
    except (TypeError, ValueError):
        return default
    return value if low <= value <= high else default