Gemini Profanity Filter

A Python module for detecting and filtering profanity in text using Google's Gemini API. The module can identify standard and obfuscated profanity across multiple languages.

Features

• Detects standard profanity in multiple languages
• Identifies obfuscated profanity (character substitutions, leetspeak, etc.)
• Handles stretched or repeated characters
• Detects deliberately fragmented words
• Works with mixed case patterns
• Processes context-sensitive profanity
• Language-specific profanity patterns including Cyrillic, Asian scripts, etc.

Installation

pip install gemini-profanity-filter

Usage

from profanity_filter import ProfanityFilter
import pandas as pd
import traceback

def run_batch_tests(pf):
    test_inputs = [
        "Ebl@NiShe nahui ti che delaesh?",
        "kon4ennyj ublYd0k idi nax",
        "ty dur@k ili 4to?",
        "suka bly@t idi v sr@ku",
        "govn0ed pozornyj",
        "pid0rasina ti proklyataya",
        "naxui idi s etim",
        "zh0p0lik bl9t",
        "mraz0ta gnid@",
        "hohol eb@n@y",
        "4mo bly@dsk@ya",
        "d@un ubly@dnij",
        "pizd@nutij shelobok",
        "ebanut@ya ovca",
        "k@kazec ebanij",
        "g0vni4e ty prost@k",
        "nax d@unish4e",
        "ni4ego ti ne ponim@esh",
        "s0sni d@",
        "nahu9 provalis"
    ]

    rows = []
    for i, text in enumerate(test_inputs, 1):
        try:
            result = pf.filter_text(text)
            if result and result.detected_profanity:
                for entry in result.detected_profanity:
                    rows.append({
                        "Тест #": i,
                        "Текст": text,
                        "Оригинал": entry.original_form,
                        "Нормализовано": entry.normalized_form,
                        "Метод": entry.detection_method,
                        "Уверенность": round(entry.confidence_score, 2),
                        "Объяснение": entry.reasoning[:100] + "..." if entry.reasoning else ""
                    })
            else:
                rows.append({
                    "Тест #": i,
                    "Текст": text,
                    "Оригинал": "-",
                    "Нормализовано": "-",
                    "Метод": "-",
                    "Уверенность": 0,
                    "Объяснение": "Нецензурная лексика не найдена"
                })
        except Exception as e:
            rows.append({
                "Тест #": i,
                "Текст": text,
                "Оригинал": "❌ Ошибка",
                "Нормализовано": "-",
                "Метод": "-",
                "Уверенность": 0,
                "Объяснение": str(e)
            })
    
    df = pd.DataFrame(rows)
    print("\n=== Сводка по результатам ===\n")
    print(df.to_string(index=False))

def main():
    try:
        print("=== Запуск авто-тестов ===")
        api_key = "AIzaSyCs3Bbjq_7PeGTAEHJkkSsUtYZvj4Fbbwo"
        pf = ProfanityFilter(api_key)
        run_batch_tests(pf)
    except Exception as e:
        print(f"❌ Ошибка: {e}")
        traceback.print_exc()

if __name__ == "__main__":
    main()

Requirements

• Python 3.7 or higher
• Google GenerativeAI Python library
• A valid Google Gemini API key

License

MIT