Delete app.py

app.py

@@ -1,1136 +0,0 @@
-import gradio as gr
-import re
-import os
-import requests
-import time
-import logging
-from packaging import version
-
-# تنظیم logging
-logging.basicConfig(level=logging.INFO)
-logger = logging.getLogger(__name__)
-
-def auto_setup_models():
-    """راه‌اندازی خودکار مدل‌ها در صورت عدم وجود"""
-    models_dir = "./models"
-    required_models = {
-        'bert-fa-ner': 'HooshvareLab/bert-fa-zwnj-base-ner',
-        'bert-base-NER': 'dslim/bert-base-NER',
-    }
-    
-    missing_models = []
-    for model_name in required_models.keys():
-        model_path = os.path.join(models_dir, model_name)
-        if not os.path.exists(model_path) or not os.listdir(model_path):
-            missing_models.append(model_name)
-    
-    if not missing_models:
-        logger.info("✅ All models are already available")
-        return True
-    
-    logger.info(f"📥 Auto-downloading missing models: {missing_models}")
-    
-    try:
-        from transformers import AutoTokenizer, AutoModelForTokenClassification
-        os.makedirs(models_dir, exist_ok=True)
-        
-        for model_name in missing_models:
-            hf_repo = required_models[model_name]
-            model_path = os.path.join(models_dir, model_name)
-            logger.info(f"📥 Downloading {model_name} from {hf_repo}...")
-            try:
-                tokenizer = AutoTokenizer.from_pretrained(hf_repo)
-                model = AutoModelForTokenClassification.from_pretrained(hf_repo)
-                tokenizer.save_pretrained(model_path)
-                model.save_pretrained(model_path)
-                logger.info(f"✅ {model_name} downloaded successfully")
-                del tokenizer, model
-            except Exception as e:
-                logger.error(f"❌ Failed to download {model_name}: {e}")
-                if os.path.exists(model_path):
-                    import shutil
-                    shutil.rmtree(model_path)
-        
-        logger.info("🎉 Auto-setup completed!")
-        return True
-        
-    except ImportError:
-        logger.error("❌ transformers library not available for auto-download")
-        return False
-    except Exception as e:
-        logger.error(f"❌ Auto-setup failed: {e}")
-        return False
-
-# اجرای auto-setup در startup
-try:
-    auto_setup_models()
-except Exception as e:
-    logger.warning(f"⚠️ Auto-setup encountered an issue: {e}")
-    logger.info("ℹ️ Continuing with manual setup...")
-
-class BilingualDataAnonymizer:
-    def __init__(self):
-        self.mapping_table = {}
-        # counters به‌روزرسانی شده با دسته‌های جدید
-        self.counters = {
-            'COMPANY': 0, 'PERSON': 0, 'AMOUNT': 0, 'ACCOUNT': 0, 
-            'DATE': 0, 'STOCK_SYMBOL': 0, 'PETROCHEMICAL': 0, 
-            'PRODUCT': 0, 'PERCENTAGE': 0, 'LOCATION': 0,
-            'VOLUME': 0, 'PHONE': 0, 'EMAIL': 0, 'ID_NUMBER': 0,
-            'FINANCIAL_TERMS': 0, 'BUSINESS_TERMS': 0, 'RATIOS': 0
-        }
-        
-        self.api_key = os.getenv("OPENAI_API_KEY", "")
-        self.models_base_path = "./models"
-        self.models_loaded = False
-        self.model_status = {}
-        self.load_local_ner_models()
-    
-    def ensure_models_directory(self):
-        if not os.path.exists(self.models_base_path):
-            try:
-                os.makedirs(self.models_base_path, exist_ok=True)
-                logger.info(f"📁 Created models directory: {self.models_base_path}")
-            except Exception as e:
-                logger.error(f"❌ Failed to create models directory: {e}")
-                return False
-        return True
-    
-    def download_model_if_missing(self, local_name, hf_repo):
-        model_path = os.path.join(self.models_base_path, local_name)
-        if os.path.exists(model_path) and os.listdir(model_path):
-            return True, f"Model {local_name} already exists"
-        try:
-            logger.info(f"📥 Auto-downloading {local_name} from {hf_repo}...")
-            from transformers import AutoTokenizer, AutoModelForTokenClassification
-            tokenizer = AutoTokenizer.from_pretrained(hf_repo)
-            model = AutoModelForTokenClassification.from_pretrained(hf_repo)
-            tokenizer.save_pretrained(model_path)
-            model.save_pretrained(model_path)
-            logger.info(f"✅ {local_name} auto-downloaded successfully")
-            return True, f"Downloaded {local_name}"
-        except Exception as e:
-            logger.error(f"❌ Auto-download failed for {local_name}: {e}")
-            return False, str(e)
-    
-    def _load_pipeline(self, task, model_path, tokenizer_path=None):
-        """لود مدل با مدیریت صحیح پارامترهای ورژن مختلف transformers"""
-        try:
-            from transformers import pipeline, AutoTokenizer, AutoModelForTokenClassification, __version__ as tr_version
-            
-            # بررسی پشتیبانی از aggregation_strategy
-            supports_agg = version.parse(tr_version) >= version.parse("4.11.0")
-            
-            # لود توکنایزر و مدل به صورت جداگانه
-            if tokenizer_path:
-                tokenizer = AutoTokenizer.from_pretrained(tokenizer_path, local_files_only=True)
-            else:
-                tokenizer = AutoTokenizer.from_pretrained(model_path, local_files_only=True)
-            
-            model = AutoModelForTokenClassification.from_pretrained(model_path, local_files_only=True)
-            
-            # ایجاد pipeline با پارامترهای مناسب
-            pipeline_kwargs = {
-                "model": model,
-                "tokenizer": tokenizer,
-                "device": -1  # استفاده از CPU
-            }
-            
-            # اضافه کردن aggregation_strategy اگر پشتیبانی می‌شود
-            if supports_agg:
-                pipeline_kwargs["aggregation_strategy"] = "simple"
-            
-            return pipeline(task, **pipeline_kwargs)
-            
-        except Exception as e:
-            logger.error(f"❌ Failed to load pipeline for {model_path}: {e}")
-            return None
-
-    def load_local_ner_models(self):
-        logger.info("🔄 Loading local NER models with auto-download...")
-        if not self.ensure_models_directory():
-            self.models_loaded = False
-            self.model_status['directory'] = "❌ Cannot create models directory"
-            return
-        
-        try:
-            try:
-                import torch
-                from transformers import AutoTokenizer, AutoModelForTokenClassification
-                transformers_available = True
-                logger.info("✅ Transformers library available")
-            except ImportError as e:
-                transformers_available = False
-                self.model_status['transformers'] = f"❌ Transformers library not installed: {str(e)}"
-                self.models_loaded = False
-                return
-            
-            # Persian model
-            persian_model_path = os.path.join(self.models_base_path, "bert-fa-ner")
-            self.download_model_if_missing("bert-fa-ner", "HooshvareLab/bert-fa-zwnj-base-ner")
-            if os.path.exists(persian_model_path) and os.listdir(persian_model_path):
-                try:
-                    self.persian_ner = self._load_pipeline("ner", persian_model_path)
-                    if self.persian_ner:
-                        self.model_status['persian'] = f"✅ Local Persian NER: {persian_model_path}"
-                    else:
-                        self.model_status['persian'] = f"❌ Failed to load Persian model: {persian_model_path}"
-                except Exception as e:
-                    self.persian_ner = None
-                    self.model_status['persian'] = f"❌ Persian model loading error: {str(e)[:100]}"
-            else:
-                self.persian_ner = None
-                self.model_status['persian'] = f"❌ Persian model not found: {persian_model_path}"
-            
-            # English model
-            english_model_path = os.path.join(self.models_base_path, "bert-base-NER")
-            self.download_model_if_missing("bert-base-NER", "dslim/bert-base-NER")
-            if os.path.exists(english_model_path) and os.listdir(english_model_path):
-                try:
-                    self.english_ner = self._load_pipeline("ner", english_model_path)
-                    if self.english_ner:
-                        self.model_status['english'] = f"✅ Local English NER: {english_model_path}"
-                    else:
-                        self.model_status['english'] = f"❌ Failed to load English model: {english_model_path}"
-                except Exception as e:
-                    self.english_ner = None
-                    self.model_status['english'] = f"❌ English model loading error: {str(e)[:100]}"
-            else:
-                self.english_ner = None
-                self.model_status['english'] = f"❌ English model not found: {english_model_path}"
-            
-            loaded_models = sum(1 for status in self.model_status.values() if status.startswith("✅"))
-            self.models_loaded = loaded_models > 0
-            if loaded_models == 0:
-                self.model_status['fallback'] = "⚠️ Using regex-only mode (no local models found)"
-                
-        except Exception as e:
-            self.models_loaded = False
-            self.model_status['critical'] = f"❌ Critical error: {str(e)[:100]}..."
-
-    def detect_language(self, text):
-        """تشخیص زبان متن"""
-        if not text:
-            return 'fa'
-        
-        persian_chars = len(re.findall(r'[\u0600-\u06FF]', text))
-        english_chars = len(re.findall(r'[a-zA-Z]', text))
-        total = persian_chars + english_chars
-        
-        if total == 0:
-            return 'fa'
-        
-        if persian_chars / total > 0.6:
-            return 'fa'
-        elif english_chars / total > 0.6:
-            return 'en'
-        else:
-            return 'mixed'
-
-    def extract_entities_with_ner(self, text, lang='fa'):
-        """استخراج entities با مدل‌های NER محلی"""
-        entities = []
-        
-        if not self.models_loaded:
-            logger.info("ℹ️ Local NER models not available - using regex only")
-            return entities
-        
-        try:
-            # مدل فارسی محلی
-            if lang in ['fa', 'mixed'] and hasattr(self, 'persian_ner') and self.persian_ner:
-                try:
-                    persian_results = self.persian_ner(text)
-                    for entity in persian_results:
-                        # بررسی فرمت خروجی بر اساس ورژن transformers
-                        if isinstance(entity, dict):
-                            if 'entity_group' in entity:
-                                # ورژن جدید با aggregation_strategy
-                                entities.append({
-                                    'text': entity['word'].strip(),
-                                    'label': entity['entity_group'],
-                                    'start': entity['start'],
-                                    'end': entity['end'],
-                                    'confidence': entity['score'],
-                                    'source': 'local_persian_ner'
-                                })
-                            else:
-                                # ورژن قدیمی
-                                entities.append({
-                                    'text': entity['word'].strip(),
-                                    'label': entity['entity'],
-                                    'start': entity['start'],
-                                    'end': entity['end'],
-                                    'confidence': entity['score'],
-                                    'source': 'local_persian_ner'
-                                })
-                    logger.info(f"Local Persian NER found {len(persian_results)} entities")
-                except Exception as e:
-                    logger.error(f"Local Persian NER extraction error: {e}")
-            
-            # مدل انگلیسی محلی
-            if lang in ['en', 'mixed'] and hasattr(self, 'english_ner') and self.english_ner:
-                try:
-                    english_results = self.english_ner(text)
-                    for entity in english_results:
-                        # بررسی فرمت خروجی بر اساس ورژن transformers
-                        if isinstance(entity, dict):
-                            if 'entity_group' in entity:
-                                # ورژن جدید با aggregation_strategy
-                                entities.append({
-                                    'text': entity['word'].strip(),
-                                    'label': entity['entity_group'],
-                                    'start': entity['start'],
-                                    'end': entity['end'],
-                                    'confidence': entity['score'],
-                                    'source': 'local_english_ner'
-                                })
-                            else:
-                                # ورژن قدیمی
-                                entities.append({
-                                    'text': entity['word'].strip(),
-                                    'label': entity['entity'],
-                                    'start': entity['start'],
-                                    'end': entity['end'],
-                                    'confidence': entity['score'],
-                                    'source': 'local_english_ner'
-                                })
-                    logger.info(f"Local English NER found {len(english_results)} entities")
-                except Exception as e:
-                    logger.error(f"Local English NER extraction error: {e}")
-                    
-        except Exception as e:
-            logger.error(f"Local NER extraction general error: {e}")
-        
-        # حذف تکراری‌ها
-        unique_entities = []
-        seen = set()
-        for entity in entities:
-            key = (entity['text'].lower(), entity['start'], entity['end'])
-            if key not in seen:
-                seen.add(key)
-                unique_entities.append(entity)
-        
-        logger.info(f"Total unique entities found by local models: {len(unique_entities)}")
-        return unique_entities
-
-    def map_ner_to_categories(self, ner_label, source=''):
-        """نگاشت برچسب‌های NER به دسته‌های سیستم"""
-        mapping = {
-            'PER': 'PERSON', 'PERSON': 'PERSON',
-            'ORG': 'COMPANY', 'ORGANIZATION': 'COMPANY',
-            'LOC': 'LOCATION', 'LOCATION': 'LOCATION',
-            'MISC': 'BUSINESS_TERMS', 'MISCELLANEOUS': 'BUSINESS_TERMS',
-            'B-PER': 'PERSON', 'I-PER': 'PERSON',
-            'B-ORG': 'COMPANY', 'I-ORG': 'COMPANY',
-            'B-LOC': 'LOCATION', 'I-LOC': 'LOCATION',
-            'B-MISC': 'BUSINESS_TERMS', 'I-MISC': 'BUSINESS_TERMS',
-            'MONEY': 'AMOUNT', 'PERCENT': 'PERCENTAGE',
-            'DATE': 'DATE', 'TIME': 'DATE'
-        }
-        return mapping.get(ner_label.upper(), 'BUSINESS_TERMS')
-
-    def anonymize_text(self, original_text, lang='fa'):
-        """گام 1: ناشناس‌سازی متن"""
-        try:
-            if not original_text or not original_text.strip():
-                return "❌ Please enter input text!" if lang == 'en' else "❌ لطفاً متن ورودی را وارد کنید!"
-            
-            # ریست متغیرها
-            self.mapping_table = {}
-            self.counters = {key: 0 for key in self.counters.keys()}
-            
-            anonymized = original_text
-            found_entities = set()
-            
-            # تشخیص زبان
-            detected_lang = self.detect_language(original_text)
-            logger.info(f"Detected language: {detected_lang}")
-            
-            # مرحله 1: استخراج با Local NER
-            if self.models_loaded:
-                logger.info("🤖 Running local NER extraction...")
-                ner_entities = self.extract_entities_with_ner(original_text, detected_lang)
-                
-                for entity in ner_entities:
-                    if (entity['text'] not in found_entities and 
-                        len(entity['text'].strip()) > 1 and 
-                        entity['confidence'] > 0.5):
-                        
-                        category = self.map_ner_to_categories(entity['label'], entity['source'])
-                        
-                        if entity['text'] not in self.mapping_table:
-                            self.counters[category] += 1
-                            code = f"{category}_{self.counters[category]:03d}_LOCAL_NER"
-                            self.mapping_table[entity['text']] = code
-                            found_entities.add(entity['text'])
-                            logger.info(f"Local NER: {entity['text']} -> {code}")
-            else:
-                logger.info("ℹ️ Using regex-only mode")
-            
-            # مرحله 2: الگوهای Regex - الگوهای جدید اضافه شده
-            patterns = {
-                'STOCK_SYMBOL': [
-                    r'نماد\s+([آ-ی‌a-zA-Z0-9]+)',
-                    r'(سبهان|غدیر|شتران|شپنا|پترول|فارس|خارک|پلاسکو|جم|کرمان|مارون|اراک|رازی|شازند|کاوه|بندر|پارس|خوزستان|ماهشهر|عسلویه)(?=\s|$|،|\.|\s+—)',
-                    r'شرکت\s+([آ-ی‌a-zA-Z\s]+?)(?=\s+در|\s+که|\s+با|،|\.|\s+$|\s+را|\s+به)',
-                    r'پتروشیمی\s+([آ-ی‌a-zA-Z\s]+?)(?=\s+در|\s+که|\s+با|،|\.|\s+$|\s+توان)',
-                    r'(AAPL|GOOGL|MSFT|AMZN|TSLA|META|NVDA|SABIC)(?=\s|$|,|\.)'
-                ],
-                'COMPANY': [
-                    r'شرکت(?=\s+در|\s+که|\s+با|\s+را|\s+به|\s+طی)',
-                    r'([آ-ی‌a-zA-Z\s]+)\s+شرکت',
-                    r'این\s+شرکت(?=\s|$|،|\.)',
-                    r'(بانک\s+[آ-ی‌a-zA-Z\s]+)',
-                    r'([A-Z][a-zA-Z\s]+(?:Inc|Corp|Corporation|Company|Ltd|Limited|LLC))'
-                ],
-                'PERSON': [
-                    r'آقای\s+([آ-ی‌a-zA-Z]+(?:\s+[آ-ی‌a-zA-Z]+)*)',
-                    r'خانم\s+([آ-ی‌a-zA-Z]+(?:\s+[آ-ی‌a-zA-Z]+)*)',
-                    r'مهندس\s+([آ-ی‌a-zA-Z]+(?:\s+[آ-ی‌a-zA-Z]+)*)',
-                    r'دکتر\s+([آ-ی‌a-zA-Z]+(?:\s+[آ-ی‌a-zA-Z]+)*)',
-                    r'([آ-ی‌a-zA-Z]+\s+[آ-ی‌a-zA-Z]+)(?=،\s+مدیرعامل|\s+مدیرعامل|\s+رئیس)',
-                    r'مدیرعامل(?=\s|$|،|\.)',
-                    r'سرپرست(?=\s+و|\s|$|،|\.)',
-                    r'رئیس\s+هیأت‌مدیره',
-                    r'وی(?=\s+ادامه|\s+اظهار|\s+گفت|\s+اعلام|\s+همچنین)'
-                ],
-                'AMOUNT': [
-                    r'\d+(?:,\d{3})*\s*(?:میلیون|میلیارد|هزار)\s*تومان',
-                    r'مبلغ\s+\d+(?:,\d{3})*\s*(?:میلیون|میلیارد|هزار)?\s*تومان',
-                    r'\d+\s*تومان(?=\s+به\s+ازای|\s+فروش|\s+،)',
-                    r'رقم\s+فعلی\s+\d+(?:,\d{3})*\s*(?:میلیون|میلیارد)\s*تومان',
-                    r'رقم\s+\d+(?:,\d{3})*\s*(?:میلیون|میلیارد)\s*تومان',
-                    r'به\s+\d+(?:,\d{3})*\s*(?:میلیون|میلیارد|هزار)\s*تومان',
-                    r'از\s+\d+(?:,\d{3})*\s*(?:میلیون|میلیارد|هزار)\s*تومان',
-                    r'برابر\s+با\s+\d+(?:,\d{3})*\s*(?:میلیون|میلیارد|هزار)\s*تومان',
-                    r'\d+(?:میلیارد|میلیون)\s*تومان(?=\s+رسیده|\s+ثبت|\s+بوده|\s+،)',
-                    r'\$\d+(?:,\d{3})*(?:\.\d+)?\s*(?:million|billion|thousand|M|B|K)?',
-                    r'\d+(?:,\d{3})*\s*ریال',
-                    r'€\d+(?:,\d{3})*(?:\.\d+)?'
-                ],
-                'PERCENTAGE': [
-                    r'\d+(?:\.\d+)?\s*درصد(?:\s+افزایش|\s+رشد|\s+کاهش|\s+بالاتر|\s+پایین‌تر)?',
-                    r'\d+(?:\.\d+)?\s*%',
-                    r'معادل\s+\d+(?:\.\d+)?\s*درصد',
-                    r'حدود\s+\d+(?:\.\d+)?\s*درصد',
-                    r'با\s+\d+(?:\.\d+)?\s*درصد\s+افزایش',
-                    r'رشد\s+\d+(?:\.\d+)?\s*درصدی',
-                    r'\d+(?:\.\d+)?\s*درصدی(?=\s+همراه|\s+بوده)',
-                    r'میزان\s+رشد(?=\s+نسبت|\s+معادل)',
-                    r'افزایش\s+قابل‌توجهی',
-                    r'بهبود\s+نسبی'
-                ],
-                'PHONE': [
-                    r'(?:تلفن[\s:]*)?(?:شماره[\s:]*)?(?:0)?(?:[۰-۹0-9]{2,3}[-\s]?)?[۰-۹0-9]{7,8}',
-                    r'(?:تماس[\s:]*)?(?:شماره[\s:]*)?(?:با[\s]*)?(?:0)?(?:[۰-۹0-9]{2,3}[-\s]?)?[۰-۹0-9]{7,8}',
-                    r'(?:موبایل[\s:]*)?(?:شماره[\s:]*)?(?:0)?9[۰-۹0-9]{9}',
-                    r'[۰-۹0-9]{3,4}[-\s][۰-۹0-9]{7,8}',
-                    r'[۰-۹0-9]{11}(?!\d)',
-                    r'(?:\+98|0098)?[۰-۹0-9]{10}',
-                    r'[۰-۹0-9]{3,4}[-\s]?[۰-۹0-9]{3,4}[-\s]?[۰-۹0-9]{3,4}'
-                ],
-                'EMAIL': [
-                    r'[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}',
-                    r'ایمیل[\s:]*[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}',
-                    r'email[\s:]*[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}',
-                    r'نشانی[\s]*الکترونیک[\s:]*[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}',
-                    r'آدرس[\s]*ایمیل[\s:]*[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}'
-                ],
-                'ACCOUNT': [
-                    r'(?:شماره[\s]*)?(?:حساب[\s]*)?(?:بانکی[\s:]*)?(?:[۰-۹0-9]{1,3}[-\s]?)*[۰-۹0-9]{8,20}',
-                    r'حساب[\s]*(?:شماره[\s:]*)?(?:[۰-۹0-9]{1,3}[-\s]?)*[۰-۹0-9]{8,20}',
-                    r'شماره[\s]*حساب[\s:]*(?:[۰-۹0-9]{1,3}[-\s]?)*[۰-۹0-9]{8,20}',
-                    r'Account[\s]*(?:Number[\s:]*)?(?:[0-9]{1,3}[-\s]?)*[0-9]{8,20}',
-                    r'[۰-۹0-9]{3}[-\s]?[۰-۹0-9]{3}[-\s]?[۰-۹0-9]{6,12}',
-                    r'[۰-۹0-9]{2,4}[-\s]?[۰-۹0-9]{6,12}[-\s]?[۰-۹0-9]{2,4}',
-                    r'واریز[\s]*(?:سود[\s:]*)?(?:[۰-۹0-9]{1,3}[-\s]?)*[۰-۹0-9]{8,20}',
-                    r'سود[\s:]*(?:[۰-۹0-9]{1,3}[-\s]?)*[۰-۹0-9]{8,20}'
-                ],
-                'ID_NUMBER': [
-                    r'IR[۰-۹0-9]{24}',
-                    r'شبا[\s:]*IR[۰-۹0-9]{24}',
-                    r'IBAN[\s:]*IR[۰-۹0-9]{24}',
-                    r'شماره[\s]*شبا[\s:]*IR[۰-۹0-9]{24}',
-                    r'(?:کد[\s]*)?(?:ملی[\s:]*)?[۰-۹0-9]{10}',
-                    r'(?:شناسه[\s]*)?(?:ملی[\s:]*)?[۰-۹0-9]{10}',
-                    r'National[\s]*(?:ID[\s:]*)?[0-9]{10}',
-                    r'(?:پاسپورت[\s:]*)?[A-Z][0-9]{8}',
-                    r'(?:Passport[\s:]*)?[A-Z][0-9]{8}',
-                    r'(?:کارت[\s:]*)?(?:[۰-۹0-9]{4}[-\s]?){3}[۰-۹0-9]{4}',
-                    r'(?:Card[\s:]*)?(?:[0-9]{4}[-\s]?){3}[0-9]{4}'
-                ],
-                'DATE': [
-                    r'[۰-۹0-9]{4}[/-][۰-۹0-9]{1,2}[/-][۰-۹0-9]{1,2}',
-                    r'[۰-۹0-9]{1,2}[/-][۰-۹0-9]{1,2}[/-][۰-۹0-9]{4}',
-                    r'(?:[۰-۹0-9]{1,2})\s*(?:فروردین|اردیبهشت|خرداد|تیر|مرداد|شهریور|مهر|آبان|آذر|دی|بهمن|اسفند)\s*(?:[۰-۹0-9]{4})',
-                    r'(?:[0-9]{1,2})\s*(?:January|February|March|April|May|June|July|August|September|October|November|December)\s*(?:[0-9]{4})',
-                    r'(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)\s*[0-9]{1,2},?\s*[0-9]{4}'
-                ]
-            }
-            
-            # پردازش patterns با اولویت‌بندی - از خاص به عام
-            logger.info("🔍 Running prioritized regex extraction...")
-            
-            # پردازش به ترتیب اولویت برای جلوگیری از تداخل
-            processed_entities = set()  # برای جلوگیری از تکرار
-            
-            for category, pattern_list in patterns.items():
-                for pattern in pattern_list:
-                    matches = re.finditer(pattern, original_text, re.IGNORECASE | re.MULTILINE)
-                    for match in matches:
-                        if match.groups():
-                            item = match.group(1).strip()
-                            full_match = match.group(0).strip()
-                        else:
-                            item = match.group(0).strip()
-                            full_match = item
-                        
-                        # بررسی تداخل با entities قبلی
-                        overlaps = False
-                        match_start, match_end = match.span()
-                        
-                        for proc_start, proc_end in processed_entities:
-                            # بررسی تداخل موقعیت
-                            if not (match_end <= proc_start or match_start >= proc_end):
-                                overlaps = True
-                                break
-                        
-                        if (not overlaps and
-                            full_match not in found_entities and 
-                            full_match not in self.mapping_table and 
-                            len(full_match) >= 2):
-                            
-                            self.counters[category] += 1
-                            code = f"{category}_{self.counters[category]:03d}_REGEX"
-                            self.mapping_table[full_match] = code
-                            found_entities.add(full_match)
-                            processed_entities.add((match_start, match_end))
-                            logger.info(f"Regex ({category}): {full_match} -> {code}")
-            
-            # جایگزینی در متن با ترتیب طولانی‌ترین اول
-            sorted_items = sorted(self.mapping_table.items(), key=lambda x: len(x[0]), reverse=True)
-            for original_item, code in sorted_items:
-                anonymized = anonymized.replace(original_item, code)
-            
-            logger.info(f"✅ Anonymization completed. Found {len(self.mapping_table)} entities.")
-            return anonymized
-            
-        except Exception as e:
-            return f"❌ Error in anonymization: {str(e)}" if lang == 'en' else f"❌ خطا در ناشناس‌سازی: {str(e)}"
-
-    def send_to_chatgpt(self, anonymized_text, lang='fa'):
-        """گام 2: ارسال به ChatGPT"""
-        try:
-            if not anonymized_text or not anonymized_text.strip():
-                return "❌ Anonymized text is empty!" if lang == 'en' else "❌ متن ناشناس‌شده خالی است!"
-            
-            if not self.api_key:
-                return "❌ API Key not configured! Please set OPENAI_API_KEY environment variable." if lang == 'en' else "❌ کلید API تنظیم نشده است! لطفاً OPENAI_API_KEY را در متغیرهای محیطی تنظیم کنید."
-            
-            system_msg = "You are a professional financial analyst. The text contains anonymous codes. Answer questions accurately." if lang == 'en' else "شما یک تحلیلگر مالی حرفه‌ای هستید. متن حاوی کدهای ناشناس است. به سوالات با دقت پاسخ دهید."
-            
-            headers = {
-                "Authorization": f"Bearer {self.api_key}",
-                "Content-Type": "application/json"
-            }
-            
-            data = {
-                "model": "gpt-4o-mini",
-                "messages": [
-                    {"role": "system", "content": system_msg},
-                    {"role": "user", "content": anonymized_text}
-                ],
-                "max_tokens": 2000,
-                "temperature": 0.7
-            }
-            
-            response = requests.post(
-                "https://api.openai.com/v1/chat/completions",
-                headers=headers,
-                json=data,
-                timeout=30
-            )
-            
-            if response.status_code == 200:
-                result = response.json()
-                return result['choices'][0]['message']['content']
-            else:
-                error_data = response.json() if response.content else {}
-                error_message = error_data.get('error', {}).get('message', response.text)
-                
-                if 'Incorrect API key' in error_message:
-                    return "❌ Invalid API key." if lang == 'en' else "❌ کلید API نامعتبر است."
-                elif 'quota' in error_message:
-                    return "❌ API quota exceeded." if lang == 'en' else "❌ سهمیه API تمام شده است."
-                else:
-                    return f"❌ API Error: {error_message}"
-                
-        except Exception as e:
-            return f"❌ Error connecting to ChatGPT: {str(e)}" if lang == 'en' else f"❌ خطا در ارتباط با ChatGPT: {str(e)}"
-
-    def deanonymize_response(self, gpt_response, lang='fa'):
-        """گام 3: بازگردانی"""
-        try:
-            if not gpt_response or not gpt_response.strip():
-                return "❌ ChatGPT response is empty!" if lang == 'en' else "❌ پاسخ ChatGPT خالی است!"
-            
-            if not self.mapping_table:
-                return "❌ Mapping table is empty!" if lang == 'en' else "❌ جدول نگاشت خالی است!"
-            
-            final_result = gpt_response
-            reverse_mapping = {code: original for original, code in self.mapping_table.items()}
-            
-            sorted_codes = sorted(reverse_mapping.items(), key=lambda x: len(x[0]), reverse=True)
-            for code, original in sorted_codes:
-                final_result = final_result.replace(code, original)
-                escaped_code = code.replace('_', '\\_')
-                final_result = final_result.replace(escaped_code, original)
-            
-            return final_result
-            
-        except Exception as e:
-            return f"❌ Deanonymization error: {str(e)}" if lang == 'en' else f"❌ خطا در بازگردانی: {str(e)}"
-
-    def get_model_status(self):
-        """وضعیت مدل‌های محلی"""
-        status = "🤖 **Local Model Status (Enhanced with Priority-Based Sensitive Data Detection):**\n\n"
-        
-        if hasattr(self, 'model_status') and self.model_status:
-            for model_type, model_status in self.model_status.items():
-                if model_type == 'persian':
-                    status += f"• **Persian NER**: {model_status}\n"
-                elif model_type == 'english':
-                    status += f"• **English NER**: {model_status}\n"
-                elif model_type == 'financial':
-                    status += f"• **Financial NER**: {model_status}\n"
-                elif model_type == 'transformers':
-                    status += f"• **Transformers**: {model_status}\n"
-                elif model_type == 'fallback':
-                    status += f"• **Fallback Mode**: {model_status}\n"
-                elif model_type == 'critical':
-                    status += f"• **Critical**: {model_status}\n"
-                elif model_type == 'directory':
-                    status += f"• **Directory**: {model_status}\n"
-        
-        loaded_count = sum(1 for status in getattr(self, 'model_status', {}).values() 
-                          if status.startswith("✅"))
-        status += f"\n📊 **Summary**: {loaded_count}/2 local models loaded"
-        
-        status += f"\n📁 **Models Path**: {self.models_base_path}"
-        status += f"\n🔧 **Latest Features**: Priority-based detection with overlap prevention"
-        
-        status += f"\n\n🔍 **Enhanced Sensitive Data Detection (Priority Order):**"
-        status += f"\n   1️⃣ **ID Numbers**: IBAN/SHEBA codes, National IDs, Passport numbers"
-        status += f"\n   2️⃣ **Contact Info**: Email addresses with context keywords"
-        status += f"\n   3️⃣ **Phone Numbers**: Mobile & landline with country codes"
-        status += f"\n   4️⃣ **Bank Accounts**: Account numbers with Persian keywords"
-        status += f"\n   5️⃣ **Financial Data**: Amounts, percentages, stock symbols"
-        status += f"\n   6️⃣ **Corporate Data**: Company names, person names, dates"
-        
-        status += f"\n\n✨ **Key Improvements:**"
-        status += f"\n   🎯 Overlap detection prevents double-matching"
-        status += f"\n   🇮🇷 Full Persian digit support (۰-۹)"
-        status += f"\n   🔄 Context-aware pattern matching"
-        status += f"\n   📏 Length-based replacement order"
-        
-        return status
-
-def process_all_steps(input_text, language):
-    """پردازش خودکار تمام مراحل"""
-    lang = 'en' if language == 'English' else 'fa'
-    
-    if not input_text.strip():
-        error_msg = "❌ Please enter input text!" if lang == 'en' else "❌ لطفاً متن ورودی را وارد کنید!"
-        return error_msg, "", "", ""
-    
-    try:
-        start_time = time.time()
-        
-        anonymized_text = anonymizer.anonymize_text(input_text, lang)
-        if anonymized_text.startswith("❌"):
-            return anonymized_text, "", "", ""
-        
-        gpt_response = anonymizer.send_to_chatgpt(anonymized_text, lang)
-        if gpt_response.startswith("❌"):
-            entities_found = len(anonymizer.mapping_table)
-            local_ner_count = sum(1 for code in anonymizer.mapping_table.values() if '_LOCAL_NER' in code)
-            regex_count = sum(1 for code in anonymizer.mapping_table.values() if '_REGEX' in code)
-            
-            # آمار اطلاعات حساس
-            sensitive_categories = ['ID_NUMBER', 'EMAIL', 'PHONE', 'ACCOUNT']
-            sensitive_count = sum(1 for code in anonymizer.mapping_table.values() 
-                                if any(cat in code for cat in sensitive_categories))
-            
-            method = "Priority-Based Local NER + Regex" if anonymizer.models_loaded else "Priority-Based Regex Only"
-            success_msg = (f"✅ Anonymization completed with {method}!\n"
-                          f"🔍 Sensitive data: {sensitive_count} | 🤖 NER: {local_ner_count} | 🔎 Regex: {regex_count}\n"
-                          f"📊 Total: {entities_found} entities protected")
-            return success_msg, anonymized_text, gpt_response, ""
-        
-        final_result = anonymizer.deanonymize_response(gpt_response, lang)
-        
-        total_time = time.time() - start_time
-        entities_found = len(anonymizer.mapping_table)
-        local_ner_count = sum(1 for code in anonymizer.mapping_table.values() if '_LOCAL_NER' in code)
-        regex_count = sum(1 for code in anonymizer.mapping_table.values() if '_REGEX' in code)
-        
-        # آمار تفصیلی اطلاعات حساس
-        id_count = sum(1 for code in anonymizer.mapping_table.values() if 'ID_NUMBER' in code)
-        email_count = sum(1 for code in anonymizer.mapping_table.values() if 'EMAIL' in code)
-        phone_count = sum(1 for code in anonymizer.mapping_table.values() if 'PHONE' in code)
-        account_count = sum(1 for code in anonymizer.mapping_table.values() if 'ACCOUNT' in code)
-        
-        sensitive_details = []
-        if id_count > 0: sensitive_details.append(f"🆔 IDs: {id_count}")
-        if email_count > 0: sensitive_details.append(f"📧 Emails: {email_count}")
-        if phone_count > 0: sensitive_details.append(f"📞 Phones: {phone_count}")
-        if account_count > 0: sensitive_details.append(f"🏦 Accounts: {account_count}")
-        
-        method = "Priority-Based Local NER + Regex" if anonymizer.models_loaded else "Priority-Based Regex Only"
-        success_msg = (f"🎉 Complete anonymization & restoration successful!\n"
-                      f"🔧 Method: {method}\n"
-                      f"🔍 Sensitive data: {' | '.join(sensitive_details) if sensitive_details else '0'}\n"
-                      f"📊 Total: {entities_found} entities | ⏱️ Time: {total_time:.2f}s")
-        
-        return success_msg, anonymized_text, gpt_response, final_result
-        
-    except Exception as e:
-        error_msg = f"❌ Processing error: {str(e)}" if lang == 'en' else f"❌ خطا در پردازش: {str(e)}"
-        return error_msg, "", "", ""
-
-def get_mapping_table(language):
-    """نمایش جدول نگاشت"""
-    lang = 'en' if language == 'English' else 'fa'
-    
-    if not anonymizer.mapping_table:
-        return "❌ Mapping table is empty! Please process some text first." if lang == 'en' else "❌ جدول نگاشت خالی است! ابتدا متنی را پردازش کنید."
-    
-    result = "📋 **Priority-Based Sensitive Data Mapping Table:**\n\n" if lang == 'en' else "📋 **جدول نگاشت اطلاعات حساس با اولویت‌بندی:**\n\n"
-    
-    local_ner_items = {k: v for k, v in anonymizer.mapping_table.items() if '_LOCAL_NER' in v}
-    regex_items = {k: v for k, v in anonymizer.mapping_table.items() if '_REGEX' in v}
-    
-    # گروه‌بندی بر اساس نوع اطلاعات حساس
-    priority_categories = {
-        'ID_NUMBER': '🆔 **Identity & Financial Codes**',
-        'EMAIL': '📧 **Email Addresses**', 
-        'PHONE': '📞 **Phone Numbers**',
-        'ACCOUNT': '🏦 **Bank Account Numbers**'
-    }
-    
-    sensitive_found = False
-    for category, title in priority_categories.items():
-        category_items = {k: v for k, v in anonymizer.mapping_table.items() if category in v}
-        if category_items:
-            sensitive_found = True
-            result += f"{title}:\n"
-            for original, code in list(category_items.items())[:8]:
-                result += f"   • `{original}` → `{code}`\n"
-            if len(category_items) > 8:
-                result += f"   ... و {len(category_items) - 8} مورد دیگر\n"
-            result += "\n"
-    
-    if local_ner_items:
-        result += "🤖 **Local NER Detected**:\n"
-        for original, code in list(local_ner_items.items())[:8]:
-            result += f"   • `{original}` → `{code}`\n"
-        if len(local_ner_items) > 8:
-            result += f"   ... و {len(local_ner_items) - 8} مورد دیگر\n"
-        result += "\n"
-    
-    # سایر موارد (مالی، شرکتی و غیره)
-    other_categories = ['AMOUNT', 'PERCENTAGE', 'COMPANY', 'PERSON', 'STOCK_SYMBOL', 'DATE']
-    other_items = {k: v for k, v in regex_items.items() 
-                  if any(cat in v for cat in other_categories)}
-    
-    if other_items:
-        result += "💼 **Business & Financial Data**:\n"
-        for original, code in list(other_items.items())[:8]:
-            result += f"   • `{original}` → `{code}`\n"
-        if len(other_items) > 8:
-            result += f"   ... و {len(other_items) - 8} مورد دیگر\n"
-    
-    # آمار کلی
-    sensitive_count = sum(len({k: v for k, v in anonymizer.mapping_table.items() if cat in v}) 
-                         for cat in priority_categories.keys())
-    
-    result += f"\n📊 **Statistics**:\n"
-    result += f"🔍 **Sensitive Data**: {sensitive_count} items\n"
-    result += f"🤖 **NER Detected**: {len(local_ner_items)} items\n" 
-    result += f"💼 **Business Data**: {len(other_items)} items\n"
-    result += f"📋 **Total**: {len(anonymizer.mapping_table)} entities\n"
-    
-    result += f"\n✨ **Enhancement Applied**: Priority-based detection with overlap prevention\n"
-    result += f"🎯 **Success**: All major sensitive data types detected and anonymized!"
-    
-    return result
-
-def clear_all():
-    """پاک کردن همه"""
-    anonymizer.mapping_table = {}
-    anonymizer.counters = {key: 0 for key in anonymizer.counters.keys()}
-    return "", "", "", "", ""
-
-def update_ui_text(language):
-    """به‌روزرسانی متن‌های رابط کاربری"""
-    if language == 'English':
-        return {
-            'title': 'Priority-Based Bilingual Data Anonymization System',
-            'step1': 'Input Text & Settings',
-            'step2': 'Anonymized Text',
-            'step3': 'Raw ChatGPT Response',
-            'step4': 'Final Restored Response',
-            'input_placeholder': 'Enter your original text here...\nExample: Company reports, person names, financial amounts, phone numbers, emails, IBAN codes, bank accounts, etc.',
-            'process_btn': 'Process with Smart Priority Detection',
-            'clear_btn': 'Clear All',
-            'mapping_btn': 'Show Priority-Based Mapping Table',
-            'copy_btn': 'Copy',
-            'direction': 'ltr'
-        }
-    else:
-        return {
-            'title': 'سیستم ناشناس‌سازی هوشمند با اولویت‌بندی',
-            'step1': 'متن ورودی و تنظیمات',
-            'step2': 'متن ناشناس‌شده',
-            'step3': 'پاسخ خام ChatGPT',
-            'step4': 'پاسخ نهایی بازگردانده شده',
-            'input_placeholder': 'متن اصلی خود را اینجا وارد کنید...\nمثال: گزارش‌های شرکت، نام اشخاص، مبالغ مالی، شماره تلفن، ایمیل، شماره شبا، حساب بانکی و غیره',
-            'process_btn': 'پردازش با تشخیص هوشمند اولویت‌دار',
-            'clear_btn': 'پاک کردن همه',
-            'mapping_btn': 'نمایش جدول نگاشت اولویت‌دار',
-            'copy_btn': 'کپی',
-            'direction': 'rtl'
-        }
-
-def update_interface(language):
-    """تغییر رابط کاربری بر اساس زبان"""
-    ui_text = update_ui_text(language)
-    is_english = (language == 'English')
-    
-    # تغییر direction برای workflow
-    workflow_css = "workflow ltr" if is_english else "workflow rtl"
-    
-    return [
-        gr.update(value=f"<h1 style='text-align: center; color: #FFD700; font-size: 3.5em; font-weight: bold; text-shadow: 3px 3px 6px rgba(0,0,0,0.5); margin: 20px 0; background: linear-gradient(45deg, #FFD700, #FFA500); -webkit-background-clip: text; -webkit-text-fill-color: transparent; background-clip: text;'>📊 {ui_text['title']}</h1>"),
-        gr.update(value=f"<h2 style='direction: {ui_text['direction']};'>📝 {ui_text['step1']}</h2>"),
-        gr.update(placeholder=ui_text['input_placeholder'], rtl=not is_english),
-        gr.update(value=f"🚀 {ui_text['process_btn']}"),
-        gr.update(value=f"🗑️ {ui_text['clear_btn']}"),
-        gr.update(rtl=not is_english),
-        gr.update(value=f"<h2 style='direction: {ui_text['direction']};'>🎭 {ui_text['step2']}</h2>"),
-        gr.update(rtl=not is_english),
-        gr.update(value=f"<h2 style='direction: {ui_text['direction']};'>🤖 {ui_text['step3']}</h2>"),
-        gr.update(rtl=not is_english),
-        gr.update(value=f"<h2 style='direction: {ui_text['direction']};'>✅ {ui_text['step4']}</h2>"),
-        gr.update(rtl=not is_english),
-        gr.update(value=f"📋 {ui_text['mapping_btn']}"),
-        gr.update(rtl=not is_english),
-        gr.update(elem_classes=workflow_css)
-    ]
-
-# ایجاد instance
-anonymizer = BilingualDataAnonymizer()
-
-# CSS اصلاح شده برای ترازبندی عمودی مناسب
-custom_css = """
-body, .gradio-container {
-    font-family: 'Segoe UI', Tahoma, Arial, sans-serif !important;
-    background: linear-gradient(135deg, #667eea 0%, #764ba2 100%) !important;
-    min-height: 100vh !important;
-    padding: 20px !important;
-}
-
-.rtl {
-    direction: rtl !important;
-    text-align: right !important;
-}
-
-.ltr {
-    direction: ltr !important;
-    text-align: left !important;
-}
-
-.workflow {
-    display: grid !important;
-    grid-template-columns: 1fr 1fr 1fr 1fr !important;
-    gap: 25px !important;
-    padding: 30px !important;
-    align-items: start !important;
-    align-content: start !important;
-    grid-auto-rows: auto !important;
-}
-
-.workflow > * {
-    align-self: start !important;
-    vertical-align: top !important;
-    margin-top: 0 !important;
-}
-
-.workflow .gradio-column,
-.workflow-column {
-    display: flex !important;
-    flex-direction: column !important;
-    align-items: stretch !important;
-    justify-content: flex-start !important;
-    height: auto !important;
-    min-height: 0 !important;
-    margin-top: 0 !important;
-    padding-top: 0 !important;
-}
-
-.gradio-textbox {
-    border-radius: 10px !important;
-    box-shadow: 0 4px 15px rgba(0,0,0,0.1) !important;
-    flex-grow: 1 !important;
-    min-height: 380px !important;
-    max-height: 380px !important;
-    height: 380px !important;
-}
-
-.gradio-textbox textarea {
-    min-height: 350px !important;
-    max-height: 350px !important;
-    height: 350px !important;
-    resize: vertical !important;
-}
-
-.workflow.rtl {
-    direction: rtl !important;
-}
-
-.workflow.ltr {
-    direction: ltr !important;
-}
-
-h1, h2, h3 {
-    text-shadow: 2px 2px 4px rgba(0,0,0,0.3) !important;
-    margin-top: 0 !important;
-    margin-bottom: 10px !important;
-    padding-top: 0 !important;
-    line-height: 1.2 !important;
-}
-
-h2 {
-    min-height: 40px !important;
-    max-height: 40px !important;
-    display: flex !important;
-    align-items: center !important;
-    margin-bottom: 15px !important;
-}
-
-.status-box {
-    background: linear-gradient(135deg, #4CAF50, #45a049) !important;
-    border: 3px solid #2E7D32 !important;
-    border-radius: 15px !important;
-    padding: 15px !important;
-    margin: 10px 0 !important;
-    box-shadow: 0 8px 32px rgba(76, 175, 80, 0.3) !important;
-    animation: pulse 2s infinite !important;
-    min-height: 120px !important;
-    max-height: 120px !important;
-}
-
-.status-box textarea {
-    background: rgba(255, 255, 255, 0.95) !important;
-    border: none !important;
-    border-radius: 10px !important;
-    font-weight: bold !important;
-    font-size: 1.1em !important;
-    color: #1B5E20 !important;
-    text-shadow: 1px 1px 2px rgba(255, 255, 255, 0.8) !important;
-    min-height: 80px !important;
-    max-height: 80px !important;
-}
-
-@keyframes pulse {
-    0% { box-shadow: 0 8px 32px rgba(76, 175, 80, 0.3); }
-    50% { box-shadow: 0 8px 40px rgba(76, 175, 80, 0.6); }
-    100% { box-shadow: 0 8px 32px rgba(76, 175, 80, 0.3); }
-}
-
-.gradio-button {
-    border-radius: 25px !important;
-    font-weight: bold !important;
-    transition: all 0.3s ease !important;
-    margin: 5px 0 !important;
-    min-height: 50px !important;
-    max-height: 50px !important;
-}
-
-.gradio-button:hover {
-    transform: translateY(-2px) !important;
-    box-shadow: 0 6px 20px rgba(0,0,0,0.2) !important;
-}
-
-h1 {
-    background: linear-gradient(45deg, #FFD700, #FFA500) !important;
-    -webkit-background-clip: text !important;
-    -webkit-text-fill-color: transparent !important;
-    background-clip: text !important;
-    min-height: 80px !important;
-}
-
-@media (max-width: 1200px) {
-    .workflow {
-        grid-template-columns: 1fr 1fr !important;
-        gap: 20px !important;
-    }
-}
-
-@media (max-width: 768px) {
-    .workflow {
-        grid-template-columns: 1fr !important;
-        gap: 15px !important;
-    }
-    
-    .gradio-textbox {
-        min-height: 300px !important;
-        max-height: 300px !important;
-        height: 300px !important;
-    }
-}
-
-[data-testid="textbox"]:dir(rtl) {
-    text-align: right !important;
-    direction: rtl !important;
-}
-
-[data-testid="textbox"]:dir(ltr) {
-    text-align: left !important;
-    direction: ltr !important;
-}
-
-.gradio-container .gradio-column {
-    align-self: start !important;
-    vertical-align: top !important;
-}
-
-.gradio-container .gradio-row {
-    align-items: flex-start !important;
-}
-
-* {
-    box-sizing: border-box !important;
-}
-
-.gradio-container {
-    align-items: start !important;
-    justify-content: start !important;
-}
-"""
-
-# رابط کاربری Gradio با ترازبندی اصلاح شده
-with gr.Blocks(title="📊 Priority-Based Anonymization System", theme=gr.themes.Soft(), css=custom_css) as app:
-    
-    with gr.Row():
-        language_selector = gr.Radio(
-            choices=["فارسی", "English"],
-            value="فارسی",
-            label="Language / زبان",
-            interactive=True
-        )
-    
-    with gr.Column():
-        title = gr.HTML("<h1 style='text-align: center; color: #FFD700; font-size: 3.5em; font-weight: bold; text-shadow: 3px 3px 6px rgba(0,0,0,0.5); margin: 20px 0; background: linear-gradient(45deg, #FFD700, #FFA500); -webkit-background-clip: text; -webkit-text-fill-color: transparent; background-clip: text;'>📊 سیستم ناشناس‌سازی هوشمند با اولویت‌بندی</h1>")
-        
-        with gr.Row(elem_classes="workflow rtl") as workflow_row:
-            with gr.Column(elem_classes="workflow-column"):
-                step1_title = gr.HTML('<h2 style="direction: rtl;">📝 متن ورودی و تنظیمات</h2>')
-                
-                input_text = gr.Textbox(
-                    lines=15,
-                    placeholder="متن اصلی خود را اینجا وارد کنید...\n✨ سیستم هوشمند اطلاعات حساس مثل شماره تلفن، ایمیل، شماره شبا، حسا�� بانکی را به ترتیب اولویت تشخیص می‌دهد",
-                    label="",
-                    rtl=True
-                )
-                
-                process_btn = gr.Button("🚀 پردازش با تشخیص هوشمند اولویت‌دار", variant="primary")
-                clear_btn = gr.Button("🗑️ پاک کردن همه", variant="stop")
-                
-                status = gr.Textbox(
-                    label="وضعیت",
-                    lines=4,
-                    interactive=False,
-                    rtl=True,
-                    elem_classes=["status-box"]
-                )
-
-            with gr.Column(elem_classes="workflow-column"):
-                step2_title = gr.HTML('<h2 style="direction: rtl;">🎭 متن ناشناس‌شده</h2>')
-                
-                anonymized_output = gr.Textbox(
-                    lines=15,
-                    placeholder="متن ناشناس‌شده اینجا نمایش داده می‌شود...",
-                    label="",
-                    interactive=False,
-                    rtl=True
-                )
-
-            with gr.Column(elem_classes="workflow-column"):
-                step3_title = gr.HTML('<h2 style="direction: rtl;">🤖 پاسخ خام ChatGPT</h2>')
-                
-                gpt_output = gr.Textbox(
-                    lines=15,
-                    placeholder="پاسخ خام ChatGPT اینجا نمایش داده می‌شود...",
-                    label="",
-                    interactive=False,
-                    rtl=True
-                )
-
-            with gr.Column(elem_classes="workflow-column"):
-                step4_title = gr.HTML('<h2 style="direction: rtl;">✅ پاسخ نهایی بازگردانده شده</h2>')
-                
-                final_output = gr.Textbox(
-                    lines=15,
-                    placeholder="پاسخ نهایی اینجا نمایش داده می‌شود...",
-                    label="",
-                    interactive=False,
-                    rtl=True
-                )
-
-        with gr.Row():
-            with gr.Column():
-                mapping_title = gr.HTML('<h2>🗂️ جدول نگاشت اولویت‌دار</h2>')
-                mapping_btn = gr.Button("📋 نمایش جدول نگاشت اولویت‌دار")
-                
-                mapping_output = gr.Textbox(
-                    lines=10,
-                    label="جدول نگاشت اطلاعات",
-                    interactive=False,
-                    visible=False,
-                    rtl=True
-                )
-
-    # Event handlers
-    language_selector.change(
-        fn=update_interface,
-        inputs=[language_selector],
-        outputs=[title, step1_title, input_text, process_btn, clear_btn, 
-                status, step2_title, anonymized_output, step3_title, gpt_output, 
-                step4_title, final_output, mapping_btn, mapping_output, workflow_row]
-    )
-
-    process_btn.click(
-        fn=process_all_steps,
-        inputs=[input_text, language_selector],
-        outputs=[status, anonymized_output, gpt_output, final_output]
-    )
-
-    clear_btn.click(
-        fn=clear_all,
-        outputs=[input_text, anonymized_output, gpt_output, final_output, status]
-    )
-
-    mapping_btn.click(
-        fn=get_mapping_table,
-        inputs=[language_selector],
-        outputs=[mapping_output]
-    )
-
-    mapping_btn.click(
-        fn=lambda: gr.update(visible=True),
-        outputs=[mapping_output]
-    )
-
-if __name__ == "__main__":
-    app.launch(share=True)