vocero-s2s/src/speech_to_speech/LLM/utils.py

import base64
import io
import re
from typing import Optional

import requests  # type: ignore[import-untyped]
from PIL import Image

SMART_PUNCT_TRANSLATION = str.maketrans(
    {
        "\u2018": "'",
        "\u2019": "'",
        "\u201c": '"',
        "\u201d": '"',
    }
)

SPEECHABLE_PATTERN = re.compile(
    r"[^\w\s.,!?;:'\"\-()\/\\@#%&*+=$€£¥₹₽¢\[\]{}<>~`^|…—–\n\r\t]",
    flags=re.UNICODE,
)

# Pattern per rilevare blocchi CJK (Cinese, Giapponese, Coreano) + altre lingue non ammesse
CJK_PATTERN = re.compile(
    r"[\u4e00-\u9fff\u3400-\u4dbf\uf900-\ufaff\ufe30-\ufe4f"  # Han (Unified + Extension A + Compatibility Ideographs + Compatibility Forms)
    r"\u3040-\u309f\u30a0-\u30ff\u31f0-\u31ff"  # Hiragana + Katakana (giapponese)
    r"\uac00-\ud7af\u1100-\u11ff\u3130-\u318f"  # Hangul (coreano)
    r"\u0600-\u06ff\u0750-\u077f\ufb50-\ufdff\ufe70-\ufeff]"  # Arabo
)

# Lettere accentate spagnole + base latina per spagnolo
SPANISH_ALLOWED_CHARS = set(
    "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"
    "áéíóúüñÁÉÍÓÚÜÑ"
    "0123456789"
    " ,.!?;:'\"-()/@#%&*+=$€£[]{}<>~`^|…—–\n\r\t"
)


def remove_unspeechable(text: str) -> str:
    """Keep only speechable characters: letters, digits, punctuation, whitespace.
    support unicode characters (english, arabic, chinese, japanese, korean, etc.)
    """
    text = text.translate(SMART_PUNCT_TRANSLATION)
    return SPEECHABLE_PATTERN.sub("", text)


def sanitize_language_output(
    text: str,
    target_language: str = "es",
    strict: bool = False,
) -> tuple[str, int]:
    """Ultimo strato di difesa: rimuove testo non appartenente alla lingua target.

    Rileva e rimuove/tzona blocchi di lingue non ammesse (CJK, Arabo, ecc.).
    Se strict=True, tronca il testo al PRIMO blocco non ammesso di lunghezza > soglia
    (per bloccare meta-commenti allucinati come quelli di Qwen in cinese).

    Returns:
        (testo_pulito, numero_caratteri_rimossi)
    """
    if not text:
        return text, 0

    original_len = len(text)
    removed_log = []

    # Step 1: Rimuovi blocchi CJK + Arabo (sempre, indipendentemente da strict)
    def _remove_cjk(match: re.Match[str]) -> str:
        removed_log.append(f"CJK chars: {match.group(0)[:30]}...")
        return ""

    text_clean = CJK_PATTERN.sub(_remove_cjk, text)

    # Step 2: Se strict, trova il primo punto dove c'e' una sequenza non spagnola
    # di > 15 caratteri e tronca li' (Qwen inizia il cinese di colpo senza transizione)
    if strict and target_language == "es":
        bad_run_start = -1
        bad_run_len = 0
        for i, ch in enumerate(text_clean):
            is_ok = ch in SPANISH_ALLOWED_CHARS or ch.isspace()
            if not is_ok:
                if bad_run_start == -1:
                    bad_run_start = i
                bad_run_len += 1
                if bad_run_len > 15:
                    # Tronca prima del run non ammesso
                    removed_log.append(
                        f"Troncato a pos {bad_run_start}: run di {bad_run_len} chars non spagnoli"
                    )
                    text_clean = text_clean[:bad_run_start].rstrip()
                    break
            else:
                bad_run_start = -1
                bad_run_len = 0

    removed_count = original_len - len(text_clean)
    if removed_count > 0:
        import logging as _log
        _log.getLogger(__name__).warning(
            "sanitize_language_output: rimossi %d caratteri non-%s. Motivi: %s",
            removed_count,
            target_language,
            "; ".join(removed_log[:3]) if removed_log else "sconosciuto",
        )

    return text_clean, removed_count


WHISPER_LANGUAGE_TO_LLM_LANGUAGE = {
    "en": "english",
    "fr": "french",
    "es": "spanish",
    "zh": "chinese",
    "ja": "japanese",
    "ko": "korean",
    "hi": "hindi",
    "de": "german",
    "pt": "portuguese",
    "pl": "polish",
    "it": "italian",
    "nl": "dutch",
}


def resolve_auto_language(language_code: Optional[str]) -> tuple[Optional[str], Optional[str]]:
    """Strip the ``-auto`` suffix and resolve the human-readable language name.

    Returns ``(clean_code, language_name)``.  ``language_name`` is non-None
    when the code (with or without ``-auto``) maps to a known language.
    """
    if not language_code:
        return language_code, None
    if language_code.endswith("-auto"):
        language_code = language_code[:-5]
    if language_code not in WHISPER_LANGUAGE_TO_LLM_LANGUAGE:
        return language_code, None
    return language_code, WHISPER_LANGUAGE_TO_LLM_LANGUAGE.get(language_code)


def image_url_to_pil(image_url: str) -> Image.Image:
    """Convert an image URL or base64 data URI to a PIL Image.

    Accepts:
    - 'data:image/...;base64,<b64>' data URIs
    - 'https://...`` or ``http://...' URLs (fetched with a 10s timeout)
    """
    if image_url.startswith("data:"):
        _, b64_data = image_url.split(",", 1)
        return Image.open(io.BytesIO(base64.b64decode(b64_data)))
    resp = requests.get(image_url, timeout=10)
    resp.raise_for_status()
    return Image.open(io.BytesIO(resp.content))