Update app.py

app.py

@@ -3,10 +3,9 @@ import gradio as gr
 import requests
 import json
 import speech_recognition as sr
+import pyttsx3
 from tempfile import NamedTemporaryFile
 from dotenv import load_dotenv
-import torchaudio
-from transformers import pipeline
 import logging
 
 # Set up logging
@@ -18,7 +17,7 @@ load_dotenv()
 
 # Groq API setup
 GROQ_API_KEY = os.getenv("GROQ_API_KEY", "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")
-GROQ_MODEL = os.getenv("GROQ_MODEL", "mixtral-8x7b-32768")  # Options: llama3-8b, gemma-7b, mixtral-8x7b-32768
+GROQ_MODEL = os.getenv("GROQ_MODEL", "mixtral-8x7b-32768")
 
 GROQ_API_URL = f"https://api.groq.com/openai/v1/chat/completions"
 
@@ -29,14 +28,6 @@ headers = {
 
 logger.info(f"Groq API configured with model: {GROQ_MODEL}")
 
-# Load YarnGPT as text-to-speech pipeline
-try:
-    tts_pipeline = pipeline("text-to-speech", model="saheedniyi/YarnGPT") # from_tf = True)
-    logger.info("Text-to-speech pipeline loaded successfully")
-except Exception as e:
-    logger.error(f"Error loading text-to-speech pipeline: {e}")
-    raise
-
 # Emotion options with descriptions
 emotion_options = {
     "neutral": "Neutral or balanced mood",
@@ -51,7 +42,6 @@ emotion_options = {
 
 # Audio transcription function
 def transcribe_audio(audio_path):
-    """Transcribe audio file to text using Google's speech recognition"""
     recognizer = sr.Recognizer()
     try:
         with sr.AudioFile(audio_path) as source:
@@ -69,17 +59,13 @@ def transcribe_audio(audio_path):
         logger.error(f"Unexpected error during transcription: {e}")
         return ""
 
-# Chat history to maintain conversation context
+# Chat history
 conversation_history = []
 
-
-# Function to call Groq API
+# Call Groq API
 def get_groq_response(prompt, conversation_messages=[]):
-    """Call Groq API to get a response"""
-    # Format messages for the Groq API (which uses OpenAI-compatible format)
     messages = [{"role": "system", "content": prompt}]
     
-    # Add conversation history
     for message in conversation_messages:
         if message.startswith("User: "):
             messages.append({"role": "user", "content": message[6:]})
@@ -95,8 +81,7 @@ def get_groq_response(prompt, conversation_messages=[]):
     
     try:
         response = requests.post(GROQ_API_URL, headers=headers, json=data)
-        response.raise_for_status()  # Raise exception for HTTP errors
-        
+        response.raise_for_status()
         result = response.json()
         return result["choices"][0]["message"]["content"]
     except requests.exceptions.RequestException as e:
@@ -105,12 +90,9 @@ def get_groq_response(prompt, conversation_messages=[]):
             logger.error(f"Response content: {e.response.text}")
         raise Exception(f"Failed to get response from Groq: {str(e)}")
 
-# Main AI interaction function
+# Main function
 def chat_with_ai(audio, text_input, emotion, history):
-    """Process user input and generate AI response with voice"""
     global conversation_history
-    
-    # Get user text from either text input or audio transcription
     user_text = text_input or ""
     
     if audio:
@@ -124,13 +106,9 @@ def chat_with_ai(audio, text_input, emotion, history):
     if not user_text.strip():
         return "No input provided. Please type a message or speak clearly.", None, history
 
-    # Update conversation history
     conversation_history.append(f"User: {user_text}")
-    
-    # Format conversation history for context (limit to last 10 exchanges)
     recent_messages = conversation_history[-20:]
-    
-    # Emotion-aware prompt
+
     system_prompt = f"""You are an empathetic AI assistant who provides supportive responses to users based on their emotional state.
 The user is currently feeling {emotion} ({emotion_options[emotion]}).
 Respond appropriately considering their emotional state.
@@ -138,33 +116,30 @@ Be supportive, empathetic, and helpful.
 Keep your responses concise and focused on helping the user."""
 
     try:
-        # Call Groq API with the prepared messages
         ai_response = get_groq_response(system_prompt, recent_messages)
         logger.info(f"Generated AI response: {ai_response[:30]}...")
     except Exception as e:
         logger.error(f"Error generating AI response: {e}")
         return "Sorry, I encountered an error generating a response. Please try again.", None, history
 
-    # Update conversation history
     conversation_history.append(f"AI: {ai_response}")
-    
-    # Limit history size
     if len(conversation_history) > 40:
         conversation_history = conversation_history[-40:]
 
-    # Generate speech from text
+    # Generate speech using pyttsx3
     try:
-        tts_output = tts_pipeline(ai_response)
         audio_path = NamedTemporaryFile(delete=False, suffix=".wav").name
-        torchaudio.save(audio_path, tts_output["audio"], sample_rate=tts_output["sampling_rate"])
-        logger.info("Text-to-speech generated successfully")
+        engine = pyttsx3.init()
+        engine.save_to_file(ai_response, audio_path)
+        engine.runAndWait()
+        logger.info("Text-to-speech generated successfully (offline)")
     except Exception as e:
         logger.error(f"Error generating speech: {e}")
         return ai_response, None, history + [[user_text, ai_response]]
 
     return ai_response, audio_path, history + [[user_text, ai_response]]
 
-# Function to clear conversation
+# Clear conversation
 def clear_conversation():
     global conversation_history
     conversation_history = []
@@ -173,22 +148,19 @@ def clear_conversation():
 # Gradio interface
 with gr.Blocks(title="Mind AID AI Assistant") as iface:
     gr.Markdown("# Mind AID: Emotion-Aware Conversational AI")
-    gr.Markdown(f"Using Groq's {GROQ_MODEL} model for AI responses and YarnGPT for voice synthesis")
+    gr.Markdown(f"Using Groq's {GROQ_MODEL} model for AI responses and offline TTS with pyttsx3")
     gr.Markdown("Talk or type to the AI assistant. Your emotional state helps tailor the response.")
     
     with gr.Row():
         with gr.Column(scale=3):
-            # Emotion selection with dropdown and descriptions
             emotion = gr.Dropdown(
                 label="How are you feeling right now?",
                 choices=list(emotion_options.keys()),
                 value="neutral",
                 type="value"
             )
-            
             emotion_description = gr.Markdown("**Current mood:** Neutral or balanced mood")
             
-            # Update the emotion description when dropdown changes
             def update_emotion_description(emotion_value):
                 return f"**Current mood:** {emotion_options.get(emotion_value, 'Unknown')}"
             
@@ -232,7 +204,6 @@ with gr.Blocks(title="Mind AID AI Assistant") as iface:
         outputs=[status_box, output_audio, chat_history]
     )
     
-    # Also allow enter key to submit
     text_input.submit(
         fn=chat_with_ai,
         inputs=[audio_input, text_input, emotion, chat_history],
@@ -245,10 +216,10 @@ with gr.Blocks(title="Mind AID AI Assistant") as iface:
         outputs=[chat_history, audio_input, text_input, status_box]
     )
 
-# Launch the application
+# Launch the app
 if __name__ == "__main__":
     try:
         logger.info("Starting Mind AID application with Groq API integration")
-        iface.launch(share=True)  # share=True creates a public link
+        iface.launch(share=True)
     except Exception as e:
         logger.error(f"Error launching Gradio interface: {e}")