Update app.py

app.py

@@ -127,29 +127,15 @@ def ai_detect_language(text: str) -> Dict[str, Any]:
                 'detection_method': 'unicode_detection'
             }
         
-        # Use Google Translate for language detection
-        try:
-            detection_result = translator.detect(text)
-            
-            # Handle async/sync versions of googletrans
-            if hasattr(detection_result, '__await__'):
-                # It's async, but we need sync - use a different approach
-                logger.warning("googletrans detect returned coroutine, falling back to sync method")
-                from googletrans import Translator
-                sync_translator = Translator()
-                detection_result = sync_translator.detect(text)
-            
-            # Handle both old and new googletrans response formats
-            if hasattr(detection_result, 'lang'):
-                detected_lang = detection_result.lang
-                confidence = getattr(detection_result, 'confidence', 0.8)
-            elif isinstance(detection_result, dict):
-                detected_lang = detection_result.get('lang', 'en')
-                confidence = detection_result.get('confidence', 0.8)
-            else:
-                logger.error(f"Unexpected detection result format: {type(detection_result)}")
-                detected_lang = 'en'
-                confidence = 0.5
+        # TEMPORARY: Skip Google Translate detection due to async issues
+        # Use enhanced rule-based detection instead
+        logger.info("Skipping Google Translate detection (async issue), using enhanced rule-based")
+        return enhanced_rule_based_detection(text)
+        
+        # TODO: Implement working language detection:
+        # - Use a different language detection library
+        # - Call Google Translate API directly with requests
+        # - Use the working parts of enhanced_rule_based_detection
             
             # Check if it's romanized Sinhala based on content analysis
             is_romanized_sinhala = (
@@ -338,58 +324,19 @@ def rule_based_transliterate_singlish_to_sinhala(text: str) -> str:
         return text  # Return original text if transliteration fails
 
 def translate_text(text: str, target_language: str = 'en') -> str:
-    """Translate text using Google Translate with robust error handling"""
+    """Temporary workaround: Skip translation due to async issues in HF environment"""
     try:
-        if translator is None:
-            logger.error("Translator not initialized")
-            return text
-            
-        # Try multiple approaches for translation
-        for attempt in range(3):
-            try:
-                # Attempt 1: Direct translation
-                if attempt == 0:
-                    result = translator.translate(text, dest=target_language)
-                # Attempt 2: Reinitialize translator
-                elif attempt == 1:
-                    logger.info("Reinitializing translator for attempt 2")
-                    from googletrans import Translator
-                    new_translator = Translator()
-                    result = new_translator.translate(text, dest=target_language)
-                # Attempt 3: Use a different service endpoint
-                else:
-                    logger.info("Using alternative translator service for attempt 3")
-                    from googletrans import Translator
-                    alt_translator = Translator(service_urls=['translate.google.com'])
-                    result = alt_translator.translate(text, dest=target_language)
-                
-                # Handle async results
-                if hasattr(result, '__await__'):
-                    logger.warning(f"Attempt {attempt + 1}: Got coroutine, skipping")
-                    continue
-                
-                # Extract text from result
-                if hasattr(result, 'text'):
-                    translated_text = result.text
-                    logger.info(f"Successfully translated on attempt {attempt + 1}")
-                    return translated_text
-                elif isinstance(result, dict) and 'text' in result:
-                    translated_text = result['text']
-                    logger.info(f"Successfully translated on attempt {attempt + 1}")
-                    return translated_text
-                else:
-                    logger.warning(f"Attempt {attempt + 1}: Unexpected result format: {type(result)}")
-                    continue
-                    
-            except Exception as attempt_error:
-                logger.warning(f"Translation attempt {attempt + 1} failed: {attempt_error}")
-                if attempt == 2:  # Last attempt
-                    raise attempt_error
-                continue
+        # TEMPORARY: Due to googletrans async issues in HF environment,
+        # we'll skip translation and return original text
+        # The chatbot can still work in the original language
         
-        # If all attempts fail, return original text
-        logger.error("All translation attempts failed")
+        logger.info(f"Translation skipped (async issue): '{text}' -> returning original")
         return text
+        
+        # TODO: Replace with a working translation service:
+        # - Use requests to call Google Translate API directly
+        # - Use a different translation library
+        # - Implement custom translation logic
             
     except Exception as e:
         logger.error(f"Translation error: {e}")
@@ -665,12 +612,12 @@ def generate_response_with_rag(user_message: str, session_id: str) -> Dict[str,
         # Create a prompt with conversation history and retrieved context
         language_instruction = ""
         if original_language == 'si':
-            language_instruction = "\n\nIMPORTANT: The user asked in Sinhala. Please respond in Sinhala using proper Sinhala script and formal language appropriate for policy discussions."
+            language_instruction = "\n\nIMPORTANT: The user asked in Sinhala. Please respond in the same language (Sinhala) using proper Sinhala script and formal language appropriate for policy discussions. The question was: '{}'".format(user_message)
         elif original_language == 'singlish':
             if transliteration_used:
-                language_instruction = "\n\nNote: The user used Romanized Sinhala (transliterated via Swabhasha). Please respond in Sinhala using proper Sinhala script and formal language appropriate for policy discussions."
+                language_instruction = "\n\nNote: The user used Romanized Sinhala (transliterated via Swabhasha). Please respond in Sinhala using proper Sinhala script and formal language appropriate for policy discussions. The original question was: '{}'".format(user_message)
             else:
-                language_instruction = "\n\nNote: The user used Singlish (Sinhala words in English letters). You may respond in English but consider using some familiar Sri Lankan terminology where appropriate."
+                language_instruction = "\n\nNote: The user used Singlish (Sinhala words in English letters). You may respond in English but consider using some familiar Sri Lankan terminology where appropriate. The original question was: '{}'".format(user_message)
         
         prompt = f"""You are a helpful assistant for budget proposals in Sri Lanka. You can communicate in English, Sinhala, and understand Singlish.