app.py

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

# Initialize Flask app
app = Flask(__name__)

# Set the secret key to handle sessions securely
app.secret_key = os.getenv("SECRET_KEY", "sSSjyhInIsUohKpG8sHzty2q")

# Configure the session type
app.config["SESSION_TYPE"] = "filesystem"
app.config["SESSION_COOKIE_NAME"] = "my_session"
app.config["SESSION_COOKIE_SECURE"] = True
app.config["SESSION_COOKIE_SAMESITE"] = "None"

# Initialize the session (Make sure to import Session at the top of your file)
from flask_session import Session
Session(app)

# Set up logging
logging.basicConfig(level=logging.INFO)

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

# Create config object for Whisper model
config = AutoConfig.from_pretrained("openai/whisper-small")
config.update({"timeout": 60})

# Create voice prompts
prompts = {
    "welcome": "Welcome to Biryani Hub.",
    "ask_name": "Tell me your name.",
    "ask_email": "Please provide your email address.",
    "thank_you": "Thank you for registration."
}

# Function to generate and save 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:
        logging.error(f"Error generating audio: {e}")
        logging.info("Retrying after 5 seconds...")
        time.sleep(5)
        generate_audio_prompt(text, filename)

# Generate all prompts
for key, text in prompts.items():
    generate_audio_prompt(text, f"{key}.mp3")

# Symbol mapping for proper recognition
SYMBOL_MAPPING = {
    "at the rate": "@",
    "at": "@",
    "dot": ".",
    "underscore": "_",
    "hash": "#",
    "plus": "+",
    "dash": "-",
    "comma": ",",
    "space": " "
}

# Function to convert audio 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)
        audio.export(output_path, format="wav")
    except Exception as e:
        raise Exception(f"Audio conversion failed: {str(e)}")

# Function to check if audio contains speech
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)
    return len(nonsilent_parts) == 0

# Salesforce connection details
try:
    logging.info("Attempting to connect to Salesforce...")
    sf = Salesforce(username='[email protected]', password='Sati@1020', security_token='sSSjyhInIsUohKpG8sHzty2q')
    logging.info("Connected to Salesforce successfully!")
except Exception as e:
    logging.error(f"Failed to connect to Salesforce: {str(e)}")

# Route to validate user login
@app.route("/validate_login", methods=["POST"])
def validate_login():
    try:
        data = request.json
        email = data.get("email")
        mobile = data.get("mobile")

        query = f"SELECT Id, Name FROM Customer_Login__c WHERE Email__c = '{email}' AND Phone_Number__c = '{mobile}'"
        result = sf.query(query)

        if result['totalSize'] > 0:
            return jsonify({'success': True, 'message': 'User authenticated successfully.'}), 200
        else:
            return jsonify({'success': False, 'error': 'Invalid email or mobile number.'}), 400
    except Exception as e:
        logging.error(f"Error during login validation: {str(e)}")
        return jsonify({'error': 'Something went wrong. Please try again later.'}), 500

# Route to serve the home page for voice-based login
@app.route("/")
def index():
    return render_template("index.html")

# Route to capture email and mobile from voice
@app.route("/capture_email_and_mobile", methods=["POST"])
def capture_email_and_mobile():
    try:
        data = request.json
        email = data.get("email")
        mobile = data.get("mobile")

        if not email or not mobile:
            return jsonify({"error": "Email or mobile number is missing."}), 400

        logging.info(f"Captured Email: {email}, Mobile: {mobile}")
        return jsonify({"success": True, "message": "Email and mobile captured successfully."}), 200

    except Exception as e:
        logging.error(f"Error in capturing email and mobile: {str(e)}")
        return jsonify({"error": "Something went wrong while processing."}), 500

# Route to handle audio transcription
@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 to WAV
        convert_to_wav(input_audio_path, output_audio_path)

        # Check for silence
        if is_silent_audio(output_audio_path):
            return jsonify({"error": "No speech detected. Please try again."}), 400

        # Use Whisper ASR model for transcription
        result = pipeline("automatic-speech-recognition", model="openai/whisper-small", device=0 if torch.cuda.is_available() else -1, config=config)
        transcribed_text = result(output_audio_path)["text"].strip().capitalize()

        return jsonify({"text": transcribed_text})

    except Exception as e:
        return jsonify({"error": f"Speech recognition error: {str(e)}"}), 500

# Start the production server
if __name__ == "__main__":
    serve(app, host="0.0.0.0", port=7860)