classifier_model/voice_language_detector.py

import os
import urllib.request
import fasttext

from app.config import config
    


fasttext.FastText.eprint = lambda x: None


class LanguageDetector:
    def __init__(self):
        if not os.path.exists(FASTTEXT_MODEL_PATH):
            print("Downloading fastText language ID model")
            urllib.request.urlretrieve(FASTTEXT_MODEL_URL, FASTTEXT_MODEL_PATH)
        self.model = fasttext.load_model(FASTTEXT_MODEL_PATH)

    def detect(self, text: str):
        clean_text = text.replace("\n", " ").strip()
        if not clean_text:
            return DEFAULT_LANGUAGE, 0.0

        labels, confidences = self.model.predict(clean_text, k=1)
        lang_code = labels[0].replace("__label__", "")
        confidence = float(confidences[0])

        
        if lang_code in VOICE_IDS and confidence >= CONFIDENCE_THRESHOLD:
            return lang_code, confidence

        return DEFAULT_LANGUAGE, confidence



detector = LanguageDetector()


def detect_language(text: str):
    """Returns (lang_code, confidence)"""
    return detector.detect(text)