Update app.py
Browse files
app.py
CHANGED
@@ -1,30 +1,60 @@
|
|
1 |
import torch
|
2 |
-
from flask import Flask, render_template, request, jsonify
|
|
|
3 |
import os
|
4 |
-
from transformers import pipeline
|
|
|
5 |
from pydub import AudioSegment
|
6 |
from pydub.silence import detect_nonsilent
|
|
|
7 |
import time
|
8 |
from waitress import serve
|
9 |
from simple_salesforce import Salesforce
|
10 |
-
import requests
|
11 |
|
12 |
app = Flask(__name__)
|
13 |
|
14 |
-
# Use whisper-small for faster processing
|
15 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
16 |
|
17 |
-
#
|
18 |
config = AutoConfig.from_pretrained("openai/whisper-small")
|
19 |
config.update({"timeout": 60}) # Set timeout to 60 seconds
|
20 |
|
21 |
-
#
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
except
|
27 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
|
29 |
# Function to convert audio to WAV format
|
30 |
def convert_to_wav(input_path, output_path):
|
@@ -33,28 +63,50 @@ def convert_to_wav(input_path, output_path):
|
|
33 |
audio = audio.set_frame_rate(16000).set_channels(1) # Convert to 16kHz, mono
|
34 |
audio.export(output_path, format="wav")
|
35 |
except Exception as e:
|
36 |
-
print(f"
|
37 |
-
raise
|
38 |
|
39 |
-
#
|
40 |
def is_silent_audio(audio_path):
|
41 |
audio = AudioSegment.from_wav(audio_path)
|
42 |
-
nonsilent_parts = detect_nonsilent(audio, min_silence_len=500, silence_thresh=audio.dBFS-16)
|
|
|
43 |
return len(nonsilent_parts) == 0 # If no speech detected
|
44 |
|
45 |
-
#
|
46 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
47 |
try:
|
48 |
-
|
49 |
'Name': name,
|
50 |
'Email__c': email,
|
51 |
'Phone_Number__c': phone_number
|
52 |
})
|
53 |
-
return {'success': True, 'id':
|
54 |
except Exception as e:
|
55 |
-
return {'error': f'Failed to create record: {str(e)}'}
|
56 |
|
57 |
-
# Registration API
|
58 |
@app.route("/submit", methods=["POST"])
|
59 |
def submit():
|
60 |
data = request.json
|
@@ -65,61 +117,91 @@ def submit():
|
|
65 |
if not name or not email or not phone:
|
66 |
return jsonify({'error': 'Missing data'}), 400
|
67 |
|
68 |
-
response = create_salesforce_record(name, email, phone)
|
69 |
-
return jsonify(response) if "error" not in response else (jsonify(response), 500)
|
70 |
-
|
71 |
-
# Updated Login API (Only for Registered Users)
|
72 |
-
@app.route('/login', methods=['POST'])
|
73 |
-
def login():
|
74 |
-
data = request.json
|
75 |
-
email = data.get('email')
|
76 |
-
phone_number = data.get('phone_number')
|
77 |
-
|
78 |
-
if not email or not phone_number:
|
79 |
-
return jsonify({'error': 'Email and phone number are required'}), 400
|
80 |
-
|
81 |
try:
|
82 |
-
#
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
# Debugging: Print Salesforce Query Result
|
93 |
-
print("🔍 Salesforce Query Result:", query_result)
|
94 |
-
|
95 |
-
if query_result['totalSize'] > 0:
|
96 |
-
print("✅ Login successful!")
|
97 |
-
return jsonify({
|
98 |
-
'success': True,
|
99 |
-
'message': 'Login successful',
|
100 |
-
'user': query_result['records'][0],
|
101 |
-
'redirect': '/dashboard'
|
102 |
-
}), 200
|
103 |
else:
|
104 |
-
|
105 |
-
|
106 |
except Exception as e:
|
107 |
-
|
108 |
-
|
109 |
|
110 |
-
# Serve Registration and Login Pages
|
111 |
@app.route("/")
|
112 |
def index():
|
113 |
return render_template("index.html")
|
114 |
|
115 |
-
@app.route("/
|
116 |
-
def
|
117 |
-
|
|
|
|
|
118 |
|
119 |
-
|
120 |
-
|
121 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
122 |
|
123 |
-
# Start
|
124 |
if __name__ == "__main__":
|
125 |
serve(app, host="0.0.0.0", port=7860)
|
|
|
1 |
import torch
|
2 |
+
from flask import Flask, render_template, request, jsonify
|
3 |
+
import json
|
4 |
import os
|
5 |
+
from transformers import pipeline
|
6 |
+
from gtts import gTTS
|
7 |
from pydub import AudioSegment
|
8 |
from pydub.silence import detect_nonsilent
|
9 |
+
from transformers import AutoConfig # Import AutoConfig for the config object
|
10 |
import time
|
11 |
from waitress import serve
|
12 |
from simple_salesforce import Salesforce
|
13 |
+
import requests # Import requests for exception handling
|
14 |
|
15 |
app = Flask(__name__)
|
16 |
|
17 |
+
# Use whisper-small for faster processing and better speed
|
18 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
19 |
|
20 |
+
# Create config object to set timeout and other parameters
|
21 |
config = AutoConfig.from_pretrained("openai/whisper-small")
|
22 |
config.update({"timeout": 60}) # Set timeout to 60 seconds
|
23 |
|
24 |
+
# Your function where you generate and save the audio
|
25 |
+
def generate_audio_prompt(text, filename):
|
26 |
+
try:
|
27 |
+
tts = gTTS(text)
|
28 |
+
tts.save(os.path.join("static", filename))
|
29 |
+
except gtts.tts.gTTSError as e:
|
30 |
+
print(f"Error: {e}")
|
31 |
+
print("Retrying after 5 seconds...")
|
32 |
+
time.sleep(5) # Wait for 5 seconds before retrying
|
33 |
+
generate_audio_prompt(text, filename)
|
34 |
+
|
35 |
+
# Generate required voice prompts
|
36 |
+
prompts = {
|
37 |
+
"welcome": "Welcome to Biryani Hub.",
|
38 |
+
"ask_name": "Tell me your name.",
|
39 |
+
"ask_email": "Please provide your email address.",
|
40 |
+
"thank_you": "Thank you for registration."
|
41 |
+
}
|
42 |
+
|
43 |
+
for key, text in prompts.items():
|
44 |
+
generate_audio_prompt(text, f"{key}.mp3")
|
45 |
+
|
46 |
+
# Symbol mapping for proper recognition
|
47 |
+
SYMBOL_MAPPING = {
|
48 |
+
"at the rate": "@",
|
49 |
+
"at": "@",
|
50 |
+
"dot": ".",
|
51 |
+
"underscore": "_",
|
52 |
+
"hash": "#",
|
53 |
+
"plus": "+",
|
54 |
+
"dash": "-",
|
55 |
+
"comma": ",",
|
56 |
+
"space": " "
|
57 |
+
}
|
58 |
|
59 |
# Function to convert audio to WAV format
|
60 |
def convert_to_wav(input_path, output_path):
|
|
|
63 |
audio = audio.set_frame_rate(16000).set_channels(1) # Convert to 16kHz, mono
|
64 |
audio.export(output_path, format="wav")
|
65 |
except Exception as e:
|
66 |
+
print(f"Error: {str(e)}")
|
67 |
+
raise Exception(f"Audio conversion failed: {str(e)}")
|
68 |
|
69 |
+
# Function to check if audio contains actual speech
|
70 |
def is_silent_audio(audio_path):
|
71 |
audio = AudioSegment.from_wav(audio_path)
|
72 |
+
nonsilent_parts = detect_nonsilent(audio, min_silence_len=500, silence_thresh=audio.dBFS-16) # Reduced silence duration
|
73 |
+
print(f"Detected nonsilent parts: {nonsilent_parts}")
|
74 |
return len(nonsilent_parts) == 0 # If no speech detected
|
75 |
|
76 |
+
# Salesforce connection details
|
77 |
+
try:
|
78 |
+
print("Attempting to connect to Salesforce...")
|
79 |
+
sf = Salesforce(username='[email protected]', password='Sati@1020', security_token='sSSjyhInIsUohKpG8sHzty2q')
|
80 |
+
print("Connected to Salesforce successfully!")
|
81 |
+
print("User Info:", sf.UserInfo) # Log the user info to verify the connection
|
82 |
+
except Exception as e:
|
83 |
+
print(f"Failed to connect to Salesforce: {str(e)}")
|
84 |
+
|
85 |
+
# Function to create Salesforce record
|
86 |
+
# API endpoint to receive data from voice bot
|
87 |
+
@app.route('/login', methods=['POST'])
|
88 |
+
def login():
|
89 |
+
# Get data from voice bot (name, email, phone number)
|
90 |
+
data = request.json # Assuming voice bot sends JSON data
|
91 |
+
|
92 |
+
name = data.get('name')
|
93 |
+
email = data.get('email')
|
94 |
+
phone_number = data.get('phone_number')
|
95 |
+
|
96 |
+
if not name or not email or not phone_number:
|
97 |
+
return jsonify({'error': 'Missing required fields'}), 400
|
98 |
+
|
99 |
+
# Create a record in Salesforce
|
100 |
try:
|
101 |
+
customer_login = sf.Customer_Login__c.create({
|
102 |
'Name': name,
|
103 |
'Email__c': email,
|
104 |
'Phone_Number__c': phone_number
|
105 |
})
|
106 |
+
return jsonify({'success': True, 'id': customer_login['id']}), 200
|
107 |
except Exception as e:
|
108 |
+
return jsonify({'error': f'Failed to create record in Salesforce: {str(e)}'}), 500
|
109 |
|
|
|
110 |
@app.route("/submit", methods=["POST"])
|
111 |
def submit():
|
112 |
data = request.json
|
|
|
117 |
if not name or not email or not phone:
|
118 |
return jsonify({'error': 'Missing data'}), 400
|
119 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
120 |
try:
|
121 |
+
# Create Salesforce record
|
122 |
+
customer_login = sf.Customer_Login__c.create({
|
123 |
+
'Name': name,
|
124 |
+
'Email__c': email,
|
125 |
+
'Phone_Number__c': phone
|
126 |
+
})
|
127 |
+
|
128 |
+
if customer_login.get('id'):
|
129 |
+
return jsonify({'success': True})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
130 |
else:
|
131 |
+
return jsonify({'error': 'Failed to create record'}), 500
|
132 |
+
|
133 |
except Exception as e:
|
134 |
+
return jsonify({'error': str(e)}), 500
|
135 |
+
|
136 |
|
|
|
137 |
@app.route("/")
|
138 |
def index():
|
139 |
return render_template("index.html")
|
140 |
|
141 |
+
@app.route("/transcribe", methods=["POST"])
|
142 |
+
def transcribe():
|
143 |
+
if "audio" not in request.files:
|
144 |
+
print("No audio file provided")
|
145 |
+
return jsonify({"error": "No audio file provided"}), 400
|
146 |
|
147 |
+
audio_file = request.files["audio"]
|
148 |
+
input_audio_path = os.path.join("static", "temp_input.wav")
|
149 |
+
output_audio_path = os.path.join("static", "temp.wav")
|
150 |
+
audio_file.save(input_audio_path)
|
151 |
+
|
152 |
+
try:
|
153 |
+
# Convert to WAV
|
154 |
+
convert_to_wav(input_audio_path, output_audio_path)
|
155 |
+
|
156 |
+
# Check for silence
|
157 |
+
if is_silent_audio(output_audio_path):
|
158 |
+
return jsonify({"error": "No speech detected. Please try again."}), 400
|
159 |
+
else:
|
160 |
+
print("Audio contains speech, proceeding with transcription.")
|
161 |
+
|
162 |
+
# Use Whisper ASR model for transcription
|
163 |
+
result = None
|
164 |
+
retry_attempts = 3
|
165 |
+
for attempt in range(retry_attempts):
|
166 |
+
try:
|
167 |
+
result = pipeline("automatic-speech-recognition", model="openai/whisper-small", device=0 if torch.cuda.is_available() else -1, config=config)
|
168 |
+
print(f"Transcribed text: {result['text']}")
|
169 |
+
break
|
170 |
+
except requests.exceptions.ReadTimeout:
|
171 |
+
print(f"Timeout occurred, retrying attempt {attempt + 1}/{retry_attempts}...")
|
172 |
+
time.sleep(5)
|
173 |
+
|
174 |
+
if result is None:
|
175 |
+
return jsonify({"error": "Unable to transcribe audio after retries."}), 500
|
176 |
+
|
177 |
+
transcribed_text = result["text"].strip().capitalize()
|
178 |
+
print(f"Transcribed text: {transcribed_text}")
|
179 |
+
|
180 |
+
# Extract name, email, and phone number from the transcribed text
|
181 |
+
parts = transcribed_text.split()
|
182 |
+
name = parts[0] if len(parts) > 0 else "Unknown Name"
|
183 |
+
email = parts[1] if '@' in parts[1] else "[email protected]"
|
184 |
+
phone_number = parts[2] if len(parts) > 2 else "0000000000"
|
185 |
+
print(f"Parsed data - Name: {name}, Email: {email}, Phone Number: {phone_number}")
|
186 |
+
|
187 |
+
# Create record in Salesforce
|
188 |
+
salesforce_response = create_salesforce_record(name, email, phone_number)
|
189 |
+
|
190 |
+
# Log the Salesforce response
|
191 |
+
print(f"Salesforce record creation response: {salesforce_response}")
|
192 |
+
|
193 |
+
# Check if the response contains an error
|
194 |
+
if "error" in salesforce_response:
|
195 |
+
print(f"Error creating record in Salesforce: {salesforce_response['error']}")
|
196 |
+
return jsonify(salesforce_response), 500
|
197 |
+
|
198 |
+
# If creation was successful, return the details
|
199 |
+
return jsonify({"text": transcribed_text, "salesforce_record": salesforce_response})
|
200 |
+
|
201 |
+
except Exception as e:
|
202 |
+
print(f"Error in transcribing or processing: {str(e)}")
|
203 |
+
return jsonify({"error": f"Speech recognition error: {str(e)}"}), 500
|
204 |
|
205 |
+
# Start Production Server
|
206 |
if __name__ == "__main__":
|
207 |
serve(app, host="0.0.0.0", port=7860)
|