backend/app/ai/voice_assistant.py

"""
Voice Banking Assistant for BankBot
Enables voice-based banking queries and responses
"""

import os
import json
import speech_recognition as sr
import pyttsx3
from gtts import gTTS
import io
import streamlit as st
from datetime import datetime

class VoiceAssistant:
    """Handles voice input and output for banking queries"""
    
    def __init__(self):
        self.recognizer = sr.Recognizer()
        self.engine = pyttsx3.init()
        self.engine.setProperty('rate', 150)  # Speaking rate
        self.microphone = sr.Microphone()
        
        # Initialize text-to-speech properties
        self.engine.setProperty('volume', 0.9)  # Volume level (0.0 to 1.0)
    
    def listen_to_user(self, timeout=10):
        """
        Capture audio input from microphone and convert to text
        Returns: Recognized text or None if recognition fails
        """
        try:
            with self.microphone as source:
                # Adjust for ambient noise
                self.recognizer.adjust_for_ambient_noise(source, duration=0.5)
                
                # Listen for audio
                audio = self.recognizer.listen(source, timeout=timeout, phrase_time_limit=15)
                
                # Try to recognize using Google Speech Recognition
                text = self.recognizer.recognize_google(audio)
                return text.lower()
        
        except sr.RequestError as e:
            return None  # Could not request results; network error
        except sr.UnknownValueError:
            return None  # Unable to recognize speech
        except Exception as e:
            print(f"Error listening to user: {e}")
            return None
    
    def speak_response(self, text, use_gtts=False):
        """
        Provide audio output for the response
        Args:
            text: Response text
            use_gtts: Use Google Text-to-Speech instead of pyttsx3
        Returns: Audio data or None
        """
        try:
            if use_gtts:
                # Use Google TTS (requires internet, better quality)
                tts = gTTS(text=text, lang='en', slow=False)
                audio_fp = io.BytesIO()
                tts.write_to_fp(audio_fp)
                audio_fp.seek(0)
                return audio_fp
            else:
                # Use pyttsx3 (offline, works locally)
                self.engine.say(text)
                self.engine.runAndWait()
                return True
        
        except Exception as e:
            print(f"Error in text-to-speech: {e}")
            return None
    
    def process_voice_query(self, transcribed_text, user_data, transactions):
        """
        Process voice query and extract banking intent
        Returns: Query type, extracted information
        """
        text_lower = transcribed_text.lower()
        
        # Balance query
        if any(word in text_lower for word in ["balance", "how much", "account balance"]):
            return "balance", None
        
        # Transaction history
        elif any(word in text_lower for word in ["transactions", "history", "recent", "last"]):
            return "transactions", None
        
        # Spending analysis
        elif any(word in text_lower for word in ["spending", "spent", "expenses", "how much did i spend"]):
            return "spending", None
        
        # Transfer query
        elif any(word in text_lower for word in ["transfer", "send", "pay"]):
            return "transfer", None
        
        # Loan info
        elif any(word in text_lower for word in ["loan", "emi", "borrow", "credit"]):
            return "loan", None
        
        # FD/Investment
        elif any(word in text_lower for word in ["fixed deposit", "fd", "invest", "investment"]):
            return "fd", None
        
        # Help/Support
        elif any(word in text_lower for word in ["help", "support", "assist", "how do i"]):
            return "help", None
        
        else:
            return "general", None
    
    def generate_voice_response(self, query_type, user_data, transactions, get_ai_response_fn=None):
        """
        Generate appropriate response for voice query
        Returns: Response text and audio
        """
        balance = user_data.get('balance', 0)
        
        if query_type == "balance":
            response = f"Your current account balance is rupees {balance:.2f}"
        
        elif query_type == "transactions":
            recent = transactions[:5] if transactions else []
            if not recent:
                response = "You have no recent transactions."
            else:
                response = f"Your last transaction was {recent[0].get('amount')} rupees for {recent[0].get('details', 'banking service')}"
        
        elif query_type == "spending":
            # Calculate spending
            debit_txns = [t for t in transactions if t.get('type') == 'debit']
            total_spent = sum(float(t.get('amount', 0)) for t in debit_txns[-10:])
            response = f"You have spent {total_spent:.2f} rupees in your recent transactions."
        
        elif query_type == "help":
            response = "I can help you with balance inquiries, transaction history, spending analysis, fund transfers, and loan information. What would you like to know?"
        
        elif query_type == "general" and get_ai_response_fn:
            # Use AI for general banking queries
            response = get_ai_response_fn("user query", [])
        
        else:
            response = "I didn't quite understand. Could you please rephrase your question?"
        
        return response
    
    def extract_voice_command(self, transcribed_text):
        """Extract command-specific parameters from voice input"""
        # Extract amounts from voice
        amount_words = {
            "hundred": 100, "thousand": 1000, "lakh": 100000,
            "rupees": 1, "rupee": 1, "paisa": 0.01
        }
        
        # Extract recipient name if present
        # Extract date references if present
        
        return None

def record_voice_query(username, users_data, get_ai_response_fn):
    """
    Record and process voice query through Streamlit UI
    """
    st.markdown("""
    <div style="background: linear-gradient(135deg, #667eea 0%, #764ba2 100%); 
                padding: 20px; border-radius: 10px; margin: 10px 0;">
        <h3 style="color: white; margin: 0;">🎤 Voice Banking Assistant</h3>
    </div>
    """, unsafe_allow_html=True)
    
    col1, col2, col3 = st.columns([1, 2, 1])
    
    with col2:
        if st.button("🎙️ Start Recording", key="voice_record", use_container_width=True):
            with st.spinner("🎧 Listening... Speak now!"):
                assistant = VoiceAssistant()
                recognized_text = assistant.listen_to_user(timeout=10)
                
                if recognized_text:
                    st.success(f"✅ Recognized: {recognized_text}")
                    
                    # Process the query
                    user_data = users_data.get(username, {})
                    transactions = user_data.get('transactions', [])
                    
                    query_type, _ = assistant.process_voice_query(recognized_text, user_data, transactions)
                    response = assistant.generate_voice_response(
                        query_type, 
                        user_data, 
                        transactions,
                        get_ai_response_fn
                    )
                    
                    # Display response
                    st.info(f"🤖 Response: {response}")
                    
                    # Provide audio feedback
                    with st.spinner("🔊 Converting to speech..."):
                        assistant.speak_response(response, use_gtts=False)
                    
                    st.success("✅ Response delivered")
                else:
                    st.error("❌ Could not recognize speech. Please try again.")

def voice_mode_demo():
    """Demo voice banking queries"""
    demo_queries = [
        "What's my balance?",
        "Show my recent transactions",
        "How much did I spend this month?",
        "Transfer 5000 to John",
        "Tell me about loan eligibility"
    ]
    
    return demo_queries