app.py

import torch
from flask import Flask, render_template, request, jsonify, redirect, session
import os
from transformers import pipeline
from gtts import gTTS
from pydub import AudioSegment
from pydub.silence import detect_nonsilent
from transformers import AutoConfig
import time
from waitress import serve
from simple_salesforce import Salesforce
import requests

# Initialize Flask app
app = Flask(__name__)

# Secret key for session management
app.secret_key = os.urandom(24)  # For session management (secure)

# Set the device for processing
device = "cuda" if torch.cuda.is_available() else "cpu"

# Whisper ASR model configuration for speech recognition
config = AutoConfig.from_pretrained("openai/whisper-small")
config.update({"timeout": 60})  # Set timeout to 60 seconds

# Salesforce Connection Setup
try:
    print("Attempting to connect to Salesforce...")
    sf = Salesforce(username='[email protected]', password='Sati@1020', security_token='sSSjyhInIsUohKpG8sHzty2q')
    print("Connected to Salesforce successfully!")
    print("User Info:", sf.UserInfo)  # Log the user info to verify the connection
except Exception as e:
    print(f"Failed to connect to Salesforce: {str(e)}")


# Functions for Salesforce Operations
def create_salesforce_record(sf, name, email, phone_number):
    try:
        customer_login = sf.Customer_Login__c.create({
            'Name': name,
            'Email__c': email,
            'Phone_Number__c': phone_number
        })
        return customer_login
    except Exception as e:
        raise Exception(f"Failed to create record: {str(e)}")


def get_menu_items(sf):
    query = "SELECT Name, Price__c, Ingredients__c, Category__c FROM Menu_Item__c"
    result = sf.query(query)
    return result['records']


# Voice-related function to generate audio prompts
def generate_audio_prompt(text, filename):
    try:
        tts = gTTS(text)
        tts.save(os.path.join("static", filename))
    except gtts.tts.gTTSError as e:
        print(f"Error: {e}")
        print("Retrying after 5 seconds...")
        time.sleep(5)  # Wait for 5 seconds before retrying
        generate_audio_prompt(text, filename)


# Utility function to convert audio files to WAV format
def convert_to_wav(input_path, output_path):
    try:
        audio = AudioSegment.from_file(input_path)
        audio = audio.set_frame_rate(16000).set_channels(1)  # Convert to 16kHz, mono
        audio.export(output_path, format="wav")
    except Exception as e:
        print(f"Error: {str(e)}")
        raise Exception(f"Audio conversion failed: {str(e)}")


# Utility function to check for silence in audio
def is_silent_audio(audio_path):
    audio = AudioSegment.from_wav(audio_path)
    nonsilent_parts = detect_nonsilent(audio, min_silence_len=500, silence_thresh=audio.dBFS-16)
    print(f"Detected nonsilent parts: {nonsilent_parts}")
    return len(nonsilent_parts) == 0  # If no speech detected


# Routes for Handling Views
@app.route("/")
def index():
    return render_template("index.html")


@app.route("/dashboard", methods=["GET"])
def dashboard():
    return render_template("dashboard.html")


@app.route('/login', methods=['POST'])
def login():
    # Get data from the voice bot (name, email, phone number)
    data = request.json  # Assuming the voice bot sends JSON data
    name = data.get('name')
    email = data.get('email')
    phone_number = data.get('phone_number')

    if not name or not email or not phone_number:
        return jsonify({'error': 'Missing required fields'}), 400

    try:
        # Create a Salesforce record for the customer
        customer_login = create_salesforce_record(sf, name, email, phone_number)
        session['customer_id'] = customer_login['id']  # Store customer ID in session
        return redirect("/menu")  # Redirect to the menu page after successful login
    except Exception as e:
        return jsonify({'error': f'Failed to create record in Salesforce: {str(e)}'}), 500


@app.route("/menu", methods=["GET"])
def menu_page():
    # Fetch menu items from Salesforce
    menu_items = get_menu_items(sf)
    menu_data = [{"name": item['Name'], "price": item['Price__c'], "ingredients": item['Ingredients__c'], "category": item['Category__c']} for item in menu_items]
    return render_template("menu_page.html", menu_items=menu_data)


@app.route("/cart", methods=["GET"])
def cart():
    # Retrieve cart items from session
    cart_items = session.get('cart_items', [])
    return render_template("cart_page.html", cart_items=cart_items)


@app.route("/order-summary", methods=["GET"])
def order_summary():
    # Retrieve the order details from session
    order_details = session.get('cart_items', [])
    total_price = sum(item['price'] * item['quantity'] for item in order_details)
    return render_template("order_summary.html", order_details=order_details, total_price=total_price)


@app.route("/final_order", methods=["GET"])
def final_order():
    # Clear cart items from session after confirming order
    session.pop('cart_items', None)
    return render_template("final_order.html")


@app.route("/order", methods=["POST"])
def place_order():
    # Adding items to the cart
    item_name = request.json.get('item_name')
    quantity = request.json.get('quantity')

    # Store the item in session cart
    cart_items = session.get('cart_items', [])
    cart_items.append({"name": item_name, "quantity": quantity, "price": 10})  # Example with fixed price
    session['cart_items'] = cart_items  # Save the updated cart to session

    return jsonify({"success": True, "message": f"Added {item_name} to cart."})


@app.route("/transcribe", methods=["POST"])
def transcribe():
    if "audio" not in request.files:
        return jsonify({"error": "No audio file provided"}), 400

    audio_file = request.files["audio"]
    input_audio_path = os.path.join("static", "temp_input.wav")
    output_audio_path = os.path.join("static", "temp.wav")
    audio_file.save(input_audio_path)

    try:
        # Convert audio to WAV format
        convert_to_wav(input_audio_path, output_audio_path)

        # Check if audio contains silence
        if is_silent_audio(output_audio_path):
            return jsonify({"error": "No speech detected. Please try again."}), 400
        else:
            print("Audio contains speech, proceeding with transcription.")

        # Use Whisper ASR model for transcription
        result = None
        retry_attempts = 3
        for attempt in range(retry_attempts):
            try:
                result = pipeline("automatic-speech-recognition", model="openai/whisper-small", device=0 if torch.cuda.is_available() else -1, config=config)
                print(f"Transcribed text: {result['text']}")
                break
            except requests.exceptions.ReadTimeout:
                print(f"Timeout occurred, retrying attempt {attempt + 1}/{retry_attempts}...")
                time.sleep(5)

        if result is None:
            return jsonify({"error": "Unable to transcribe audio after retries."}), 500

        transcribed_text = result["text"].strip().capitalize()
        print(f"Transcribed text: {transcribed_text}")

        # Extract name, email, and phone number from the transcribed text
        parts = transcribed_text.split()
        name = parts[0] if len(parts) > 0 else "Unknown Name"
        email = parts[1] if '@' in parts[1] else "[email protected]"
        phone_number = parts[2] if len(parts) > 2 else "0000000000"

        print(f"Parsed data - Name: {name}, Email: {email}, Phone Number: {phone_number}")

        # Confirm details before submission
        confirmation = f"Is this correct? Name: {name}, Email: {email}, Phone: {phone_number}"
        generate_audio_prompt(confirmation, "confirmation.mp3")

        # Simulate confirmation via user action
        user_confirms = True  # Assuming the user confirms, replace with actual logic

        if user_confirms:
            # Create a record in Salesforce
            salesforce_response = create_salesforce_record(name, email, phone_number)

            return jsonify({"text": transcribed_text, "salesforce_record": salesforce_response})

    except Exception as e:
        print(f"Error in transcribing or processing: {str(e)}")
        return jsonify({"error": f"Speech recognition error: {str(e)}"}), 500


# Start Production Server
if __name__ == "__main__":
    serve(app, host="0.0.0.0", port=7860)