Update app.py

app.py

@@ -7,6 +7,7 @@ from dotenv import load_dotenv
 import os
 import speech_recognition as sr
 from pydub import AudioSegment
+import re
 
 load_dotenv()
 
@@ -22,39 +23,50 @@ recognizer = sr.Recognizer()
 
 MAIN_SYSTEM_PROMPT = {
     "role": "system",
-    "content": """You are Sam, an intelligent and adaptive English tutor. Your responses must be in JSON format with these keys:
-        'response': Your main response,
-        'corrections': Grammar or pronunciation corrections if needed,
-        'vocabulary': Suggested alternative words or phrases,
-        'level_assessment': Current assessment of user's English level (beginner/intermediate/advanced),
+    "content": """You are Sam, an intelligent and proactive English tutor. You drive the conversation and actively engage students. Your responses must be in JSON format with these keys:
+        'response': Your main response (keep it conversational and engaging),
+        'corrections': Specific grammar or pronunciation corrections with examples,
+        'vocabulary': Alternative words/phrases with explanations,
+        'level_assessment': Current assessment (beginner/intermediate/advanced),
         'encouragement': A motivating comment,
-        'context_memory': Important details about the user to remember (interests, job, etc.)
-        Guidelines:
-        1. Maintain natural conversation while gathering user information
-        2. Adapt language complexity to user's level
-        3. Provide gentle corrections without interrupting flow
-        4. Remember and use context from previous messages
-        5. Focus on user's interests and profession
-        6. Give specific vocabulary related to user's field
-        7. Keep responses encouraging and supportive
+        'context_memory': Important details about the user,
+        'next_question': A follow-up question to keep conversation flowing
         
-        Special behaviors:
-        - If user mentions their country → Ask about their culture/country
-        - If user mentions their job → Use profession-specific vocabulary
-        - If user mentions hobbies → Engage in that topic
-        - For beginners → Use simple words and short sentences
-        - For intermediate/advanced → Introduce more complex vocabulary
-        Always maintain conversation history and adapt based on it."""
+        Your personality:
+        - Be the conversation driver - ask follow-up questions
+        - Show genuine interest in the student's life
+        - Provide corrections naturally without stopping the flow
+        - Use the student's name frequently
+        - Build on previous topics
+        - Be encouraging but provide constructive feedback
+        - Ask about their day, work, hobbies, culture, goals
+        
+        Correction guidelines:
+        - Always provide corrections if there are grammar mistakes
+        - Suggest better vocabulary choices
+        - Give pronunciation tips when needed
+        - Use format: "Instead of 'X', try saying 'Y'"
+        
+        Conversation flow:
+        - Start with personal questions (name, country, job, hobbies)
+        - Build conversations around their interests
+        - Use profession-specific vocabulary
+        - Ask about their culture and experiences
+        - Keep the conversation natural and flowing
+        - Always end with a question to continue the dialogue
+        
+        Response length: Keep responses conversational (2-3 sentences max for response field)."""
 }
 
 WELCOME_PROMPT = {
     "role": "system",
-    "content": """Create a brief welcome message that:
-        1. Introduces you as Sam, the English tutor
-        2. Asks for the user's name only
+    "content": """Create a warm welcome message that:
+        1. Introduces you as Sam, an enthusiastic English tutor
+        2. Asks for their name and where they're from
+        3. Shows excitement about helping them
         
         Return in JSON format with key 'greeting'.
-        Keep it short, friendly and natural - maximum 2 sentences."""
+        Make it warm, friendly and enthusiastic - maximum 2 sentences."""
 }
 
 class EnglishTutor:
@@ -62,7 +74,7 @@ class EnglishTutor:
         self.chat_history = [MAIN_SYSTEM_PROMPT]
         self.user_info = {
             "name": None,
-            "level": None,
+            "level": "beginner",
             "interests": [],
             "country": None,
             "profession": None,
@@ -77,7 +89,7 @@ class EnglishTutor:
                 json={
                     "model": "deepseek-chat",
                     "messages": [WELCOME_PROMPT],
-                    "temperature": random.uniform(0.9, 1),
+                    "temperature": random.uniform(0.8, 1.0),
                     "response_format": {"type": "json_object"}
                 }
             )
@@ -85,11 +97,15 @@ class EnglishTutor:
             return welcome_json["greeting"]
         except Exception as e:
             print(f"Error in welcome message: {str(e)}")
-            return "Hi! I'm Sam, your English tutor. What's your name?"
+            return "Hi! I'm Sam, your English tutor. What's your name and where are you from?"
 
     def get_bot_response(self, user_message):
         try:
-            self.chat_history.append({"role": "user", "content": user_message})
+            # Add user context to the message
+            context_info = f"User info: {self.user_info}"
+            enhanced_message = f"{user_message}\n\n[Context: {context_info}]"
+            
+            self.chat_history.append({"role": "user", "content": enhanced_message})
             
             response = requests.post(
                 "https://api.deepseek.com/v1/chat/completions",
@@ -97,49 +113,66 @@ class EnglishTutor:
                 json={
                     "model": "deepseek-chat",
                     "messages": self.chat_history,
-                    "temperature": random.uniform(0.9, 1.0),
+                    "temperature": random.uniform(0.8, 1.0),
                     "response_format": {"type": "json_object"}
                 }
             )
             
             bot_response = json.loads(response.json()["choices"][0]["message"]["content"])
             
+            # Update user info
             if "level_assessment" in bot_response:
                 self.user_info["level"] = bot_response["level_assessment"]
             if "context_memory" in bot_response:
                 self._update_user_info(bot_response["context_memory"])
             
-            formatted_response = self._format_response(bot_response)
-            
             self.chat_history.append({"role": "assistant", "content": json.dumps(bot_response)})
             
-            return formatted_response
+            return bot_response
         except Exception as e:
             print(f"Error getting bot response: {str(e)}")
-            return "I apologize, but I couldn't process that properly. Could you try again?"
+            return {
+                "response": "I apologize, but I couldn't process that properly. Could you try again?",
+                "corrections": "",
+                "vocabulary": "",
+                "level_assessment": "beginner",
+                "encouragement": "Don't worry, let's keep practicing!",
+                "context_memory": "",
+                "next_question": "What would you like to talk about?"
+            }
 
     def _update_user_info(self, context_memory):
-        if isinstance(context_memory, dict):
+        if isinstance(context_memory, str):
+            # Try to extract name if mentioned
+            if "name" in context_memory.lower():
+                name_match = re.search(r"name[:\s]+([A-Za-z]+)", context_memory)
+                if name_match:
+                    self.user_info["name"] = name_match.group(1)
+            
+            # Try to extract country if mentioned
+            if "country" in context_memory.lower() or "from" in context_memory.lower():
+                country_match = re.search(r"(?:from|country)[:\s]+([A-Za-z\s]+)", context_memory)
+                if country_match:
+                    self.user_info["country"] = country_match.group(1).strip()
+        
+        elif isinstance(context_memory, dict):
             for key in self.user_info:
                 if key in context_memory:
                     self.user_info[key] = context_memory[key]
 
-    def _format_response(self, response_dict):
-        formatted = response_dict["response"]
+    def clean_text_for_tts(self, text):
+        # Remove emojis and special characters that might cause TTS issues
+        text = re.sub(r'[🎯🌟✨💫🎤🤖]', '', text)
         
-        if response_dict.get("encouragement"):
-            encouragement = response_dict["encouragement"]
-            # Make encouragement more impressive
-            if "doing" in encouragement.lower() and "great" in encouragement.lower():
-                formatted += f"\n\n🌟 {encouragement}"
-            elif "wonderful" in encouragement.lower() or "excellent" in encouragement.lower():
-                formatted += f"\n\n✨ {encouragement}"
-            elif "good" in encouragement.lower() or "well" in encouragement.lower():
-                formatted += f"\n\n🎯 {encouragement}"
-            else:
-                formatted += f"\n\n💫 {encouragement}"
+        # Remove extra spaces and newlines
+        text = re.sub(r'\s+', ' ', text).strip()
         
-        return formatted
+        # Remove duplicate words at the beginning
+        words = text.split()
+        if len(words) > 1 and words[0].lower() == words[1].lower():
+            text = ' '.join(words[1:])
+        
+        return text
 
 def convert_audio_to_text(audio_path):
     try:
@@ -178,14 +211,15 @@ tutor = EnglishTutor()
 def initialize_chat():
     try:
         welcome = tutor.get_welcome_message()
-        welcome_audio = text_to_speech(welcome)
+        clean_welcome = tutor.clean_text_for_tts(welcome)
+        welcome_audio = text_to_speech(clean_welcome)
         history = [{"role": "assistant", "content": welcome}]
-        return history, welcome_audio, f"🎤 Sam: {welcome}", ""
+        return history, welcome_audio, f"🤖 Sam: {welcome}", ""
     except Exception as e:
         print(f"Error initializing chat: {str(e)}")
-        welcome_msg = "Hi! I'm Sam, your English tutor. What's your name?"
+        welcome_msg = "Hi! I'm Sam, your English tutor. What's your name and where are you from?"
         history = [{"role": "assistant", "content": welcome_msg}]
-        return history, None, f"🎤 Sam: {welcome_msg}", ""
+        return history, None, f"🤖 Sam: {welcome_msg}", ""
 
 def process_audio(audio, history, transcript, corrections):
     try:
@@ -196,40 +230,50 @@ def process_audio(audio, history, transcript, corrections):
         if not user_message:
             return history, None, transcript, corrections
 
-        bot_response_json = tutor.get_bot_response(user_message)
+        bot_response = tutor.get_bot_response(user_message)
         
-        try:
-            response_dict = json.loads(bot_response_json) if isinstance(bot_response_json, str) else bot_response_json
-        except:
-            response_dict = {"response": bot_response_json}
+        # Create the main response with follow-up question
+        main_response = bot_response.get("response", "")
+        if bot_response.get("next_question"):
+            main_response += f" {bot_response['next_question']}"
         
-        tts_response = response_dict.get("response", bot_response_json)
-        if response_dict.get("encouragement"):
-            tts_response += f" {response_dict['encouragement']}"
+        # Add encouragement
+        if bot_response.get("encouragement"):
+            main_response += f" {bot_response['encouragement']}"
         
-        audio_response = text_to_speech(tts_response)
+        # Clean text for TTS
+        clean_response = tutor.clean_text_for_tts(main_response)
+        audio_response = text_to_speech(clean_response)
         
-        formatted_response = tutor._format_response(response_dict) if hasattr(tutor, '_format_response') else bot_response_json
-
+        # Update chat history
         history = history or []
         history.append({"role": "user", "content": user_message})
-        history.append({"role": "assistant", "content": formatted_response})
+        history.append({"role": "assistant", "content": main_response})
         
-        new_transcript = transcript + f"\n\n🎤 You: {user_message}\n🤖 Sam: {tts_response}"
+        # Update transcript
+        new_transcript = transcript + f"\n\n🎤 You: {user_message}\n🤖 Sam: {main_response}"
         
+        # Update corrections and vocabulary (THIS IS THE FIX!)
         new_corrections = corrections
-        if response_dict.get("corrections") or response_dict.get("vocabulary"):
-            correction_text = ""
-            if response_dict.get("corrections"):
-                correction_text += f"✍️ Grammar: {response_dict['corrections']}\n\n"
-            if response_dict.get("vocabulary"):
-                vocab = response_dict['vocabulary']
-                if isinstance(vocab, dict):
-                    vocab_text = "\n".join([f"• '{k}' → '{v}'" for k, v in vocab.items()])
-                else:
-                    vocab_text = str(vocab)
-                correction_text += f"📚 Vocabulary:\n{vocab_text}"
-            new_corrections = corrections + f"\n\n{correction_text}" if corrections else correction_text
+        correction_parts = []
+        
+        if bot_response.get("corrections") and bot_response["corrections"].strip():
+            correction_parts.append(f"✍️ **Grammar Corrections:**\n{bot_response['corrections']}")
+        
+        if bot_response.get("vocabulary") and bot_response["vocabulary"].strip():
+            vocab = bot_response['vocabulary']
+            if isinstance(vocab, dict):
+                vocab_text = "\n".join([f"• '{k}' → '{v}'" for k, v in vocab.items()])
+            else:
+                vocab_text = str(vocab)
+            correction_parts.append(f"📚 **Vocabulary Suggestions:**\n{vocab_text}")
+        
+        if bot_response.get("level_assessment"):
+            correction_parts.append(f"📊 **Current Level:** {bot_response['level_assessment'].title()}")
+        
+        if correction_parts:
+            new_correction_text = "\n\n".join(correction_parts)
+            new_corrections = new_corrections + f"\n\n--- Latest Feedback ---\n{new_correction_text}" if new_corrections else new_correction_text
         
         return history, audio_response, new_transcript, new_corrections
     except Exception as e:
@@ -245,8 +289,8 @@ def clear_chat():
     return initialize_chat()
 
 with gr.Blocks(theme=gr.themes.Soft()) as demo:
-    gr.Markdown("# 🎓 English Learning Assistant")
-    gr.Markdown("🎤 **Record your voice and click submit** - I'll help you improve your English conversation skills!")
+    gr.Markdown("# 🎓 English Learning Assistant with Sam")
+    gr.Markdown("🎤 **Record your voice and click submit** - Sam will actively guide your conversation and help improve your English!")
     
     with gr.Row():
         with gr.Column(scale=3):
@@ -291,13 +335,13 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
                 max_lines=8,
                 show_label=False,
                 interactive=False,
-                placeholder="Grammar corrections and vocabulary suggestions will appear here...",
+                placeholder="Grammar corrections, vocabulary suggestions, and level assessment will appear here...",
                 container=True
             )
     
     with gr.Row():
         clear_btn = gr.Button("🔄 Start New Conversation", variant="secondary", size="lg")
-        gr.Markdown("💡 **Tip**: Check the Learning Corner for personalized feedback on your English!")
+        gr.Markdown("💡 **Tip**: Sam will actively guide the conversation and provide personalized feedback!")
     
     submit_btn.click(
         submit_recording,