Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,190 +1,213 @@
|
|
1 |
-
from flask import Flask, request, jsonify, render_template
|
2 |
-
from dotenv import load_dotenv
|
3 |
-
from groq import Groq
|
4 |
-
import os
|
5 |
-
import uuid
|
6 |
-
from gtts import gTTS
|
7 |
-
import io
|
8 |
-
import base64
|
9 |
-
import speech_recognition as sr
|
10 |
-
import tempfile
|
11 |
-
import json
|
12 |
-
|
13 |
-
try:
|
14 |
-
import pyaudio
|
15 |
-
except ImportError:
|
16 |
-
print("Warning: PyAudio not available, speech functionality will be limited")
|
17 |
-
|
18 |
-
# Initialize Flask app
|
19 |
-
app = Flask(__name__, static_folder='static')
|
20 |
-
|
21 |
-
# Load environment variables
|
22 |
-
load_dotenv()
|
23 |
-
|
24 |
-
# Groq API Configuration
|
25 |
-
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
|
26 |
-
client = Groq(api_key=GROQ_API_KEY)
|
27 |
-
MODEL = "llama3-70b-8192"
|
28 |
-
|
29 |
-
# Initialize speech recognition
|
30 |
-
recognizer = sr.Recognizer()
|
31 |
-
|
32 |
-
# Store conversation history
|
33 |
-
conversations = {}
|
34 |
-
|
35 |
-
def load_base_prompt():
|
36 |
-
try:
|
37 |
-
with open("base_prompt.txt", "r") as file:
|
38 |
-
return file.read().strip()
|
39 |
-
except FileNotFoundError:
|
40 |
-
print("Error: base_prompt.txt file not found.")
|
41 |
-
return "You are a helpful assistant for language learning."
|
42 |
-
|
43 |
-
# Load the base prompt
|
44 |
-
base_prompt = load_base_prompt()
|
45 |
-
|
46 |
-
def chat_with_groq(user_message, conversation_id=None):
|
47 |
-
try:
|
48 |
-
# Get conversation history or create new
|
49 |
-
messages = conversations.get(conversation_id, [])
|
50 |
-
if not messages:
|
51 |
-
messages.append({"role": "system", "content": base_prompt})
|
52 |
-
|
53 |
-
# Add user message
|
54 |
-
messages.append({"role": "user", "content": user_message})
|
55 |
-
|
56 |
-
# Get completion from Groq
|
57 |
-
completion = client.chat.completions.create(
|
58 |
-
model=MODEL,
|
59 |
-
messages=messages,
|
60 |
-
temperature=0.1,
|
61 |
-
max_tokens=1024
|
62 |
-
)
|
63 |
-
|
64 |
-
# Add assistant's response to history
|
65 |
-
assistant_message = completion.choices[0].message.content.strip()
|
66 |
-
messages.append({"role": "assistant", "content": assistant_message})
|
67 |
-
|
68 |
-
# Update conversation history
|
69 |
-
if conversation_id:
|
70 |
-
conversations[conversation_id] = messages
|
71 |
-
|
72 |
-
return assistant_message
|
73 |
-
except Exception as e:
|
74 |
-
print(f"Error in chat_with_groq: {str(e)}")
|
75 |
-
return f"I apologize, but I'm having trouble responding right now. Error: {str(e)}"
|
76 |
-
|
77 |
-
def text_to_speech(text):
|
78 |
-
try:
|
79 |
-
tts = gTTS(text=text, lang='en')
|
80 |
-
audio_io = io.BytesIO()
|
81 |
-
tts.write_to_fp(audio_io)
|
82 |
-
audio_io.seek(0)
|
83 |
-
return audio_io
|
84 |
-
except Exception as e:
|
85 |
-
print(f"Error in text_to_speech: {str(e)}")
|
86 |
-
return None
|
87 |
-
|
88 |
-
def speech_to_text(audio_file):
|
89 |
-
try:
|
90 |
-
# Save the uploaded audio to a temporary file
|
91 |
-
with tempfile.NamedTemporaryFile(delete=False, suffix='.wav') as temp_audio:
|
92 |
-
audio_file.save(temp_audio.name)
|
93 |
-
|
94 |
-
# Use SpeechRecognition to convert speech to text
|
95 |
-
with sr.AudioFile(temp_audio.name) as source:
|
96 |
-
# Adjust recognition settings
|
97 |
-
recognizer.dynamic_energy_threshold = True
|
98 |
-
recognizer.energy_threshold = 4000
|
99 |
-
|
100 |
-
# Record the entire audio file
|
101 |
-
audio = recognizer.record(source)
|
102 |
-
|
103 |
-
# Perform recognition with increased timeout
|
104 |
-
text = recognizer.recognize_google(audio, language='en-US')
|
105 |
-
return text
|
106 |
-
|
107 |
-
except sr.UnknownValueError:
|
108 |
-
return "Could not understand audio"
|
109 |
-
except sr.RequestError as e:
|
110 |
-
return f"Could not request results; {str(e)}"
|
111 |
-
except Exception as e:
|
112 |
-
print(f"Error in speech_to_text: {str(e)}")
|
113 |
-
return None
|
114 |
-
finally:
|
115 |
-
# Clean up temporary file
|
116 |
-
try:
|
117 |
-
os.unlink(temp_audio.name)
|
118 |
-
except:
|
119 |
-
pass
|
120 |
-
|
121 |
-
@app.route('/')
|
122 |
-
def index():
|
123 |
-
return render_template('index.html')
|
124 |
-
|
125 |
-
@app.route('/api/chat', methods=['POST'])
|
126 |
-
def chat():
|
127 |
-
try:
|
128 |
-
data = request.get_json()
|
129 |
-
user_message = data.get('message', '')
|
130 |
-
conversation_id = data.get('conversation_id', str(uuid.uuid4()))
|
131 |
-
|
132 |
-
if not user_message:
|
133 |
-
return jsonify({'error': 'No message provided'}), 400
|
134 |
-
|
135 |
-
# Get response from Groq
|
136 |
-
response = chat_with_groq(user_message, conversation_id)
|
137 |
-
|
138 |
-
# Generate voice response
|
139 |
-
audio_io = text_to_speech(response)
|
140 |
-
result = {
|
141 |
-
'response': response,
|
142 |
-
'conversation_id': conversation_id
|
143 |
-
}
|
144 |
-
|
145 |
-
if audio_io:
|
146 |
-
audio_base64 = base64.b64encode(audio_io.getvalue()).decode('utf-8')
|
147 |
-
result['voice_response'] = audio_base64
|
148 |
-
|
149 |
-
return jsonify(result)
|
150 |
-
|
151 |
-
except Exception as e:
|
152 |
-
return jsonify({'error': str(e)}), 500
|
153 |
-
|
154 |
-
@app.route('/api/voice', methods=['POST'])
|
155 |
-
def handle_voice():
|
156 |
-
try:
|
157 |
-
if 'audio' not in request.files:
|
158 |
-
return jsonify({'error': 'No audio file provided'}), 400
|
159 |
-
|
160 |
-
audio_file = request.files['audio']
|
161 |
-
conversation_id = request.form.get('conversation_id', str(uuid.uuid4()))
|
162 |
-
|
163 |
-
#
|
164 |
-
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
-
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
|
182 |
-
|
183 |
-
|
184 |
-
|
185 |
-
|
186 |
-
|
187 |
-
|
188 |
-
|
189 |
-
|
190 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from flask import Flask, request, jsonify, render_template
|
2 |
+
from dotenv import load_dotenv
|
3 |
+
from groq import Groq
|
4 |
+
import os
|
5 |
+
import uuid
|
6 |
+
from gtts import gTTS
|
7 |
+
import io
|
8 |
+
import base64
|
9 |
+
import speech_recognition as sr
|
10 |
+
import tempfile
|
11 |
+
import json
|
12 |
+
|
13 |
+
try:
|
14 |
+
import pyaudio
|
15 |
+
except ImportError:
|
16 |
+
print("Warning: PyAudio not available, speech functionality will be limited")
|
17 |
+
|
18 |
+
# Initialize Flask app
|
19 |
+
app = Flask(__name__, static_folder='static')
|
20 |
+
|
21 |
+
# Load environment variables
|
22 |
+
load_dotenv()
|
23 |
+
|
24 |
+
# Groq API Configuration
|
25 |
+
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
|
26 |
+
client = Groq(api_key=GROQ_API_KEY)
|
27 |
+
MODEL = "llama3-70b-8192"
|
28 |
+
|
29 |
+
# Initialize speech recognition
|
30 |
+
recognizer = sr.Recognizer()
|
31 |
+
|
32 |
+
# Store conversation history
|
33 |
+
conversations = {}
|
34 |
+
|
35 |
+
def load_base_prompt():
|
36 |
+
try:
|
37 |
+
with open("base_prompt.txt", "r") as file:
|
38 |
+
return file.read().strip()
|
39 |
+
except FileNotFoundError:
|
40 |
+
print("Error: base_prompt.txt file not found.")
|
41 |
+
return "You are a helpful assistant for language learning."
|
42 |
+
|
43 |
+
# Load the base prompt
|
44 |
+
base_prompt = load_base_prompt()
|
45 |
+
|
46 |
+
def chat_with_groq(user_message, conversation_id=None):
|
47 |
+
try:
|
48 |
+
# Get conversation history or create new
|
49 |
+
messages = conversations.get(conversation_id, [])
|
50 |
+
if not messages:
|
51 |
+
messages.append({"role": "system", "content": base_prompt})
|
52 |
+
|
53 |
+
# Add user message
|
54 |
+
messages.append({"role": "user", "content": user_message})
|
55 |
+
|
56 |
+
# Get completion from Groq
|
57 |
+
completion = client.chat.completions.create(
|
58 |
+
model=MODEL,
|
59 |
+
messages=messages,
|
60 |
+
temperature=0.1,
|
61 |
+
max_tokens=1024
|
62 |
+
)
|
63 |
+
|
64 |
+
# Add assistant's response to history
|
65 |
+
assistant_message = completion.choices[0].message.content.strip()
|
66 |
+
messages.append({"role": "assistant", "content": assistant_message})
|
67 |
+
|
68 |
+
# Update conversation history
|
69 |
+
if conversation_id:
|
70 |
+
conversations[conversation_id] = messages
|
71 |
+
|
72 |
+
return assistant_message
|
73 |
+
except Exception as e:
|
74 |
+
print(f"Error in chat_with_groq: {str(e)}")
|
75 |
+
return f"I apologize, but I'm having trouble responding right now. Error: {str(e)}"
|
76 |
+
|
77 |
+
def text_to_speech(text):
|
78 |
+
try:
|
79 |
+
tts = gTTS(text=text, lang='en')
|
80 |
+
audio_io = io.BytesIO()
|
81 |
+
tts.write_to_fp(audio_io)
|
82 |
+
audio_io.seek(0)
|
83 |
+
return audio_io
|
84 |
+
except Exception as e:
|
85 |
+
print(f"Error in text_to_speech: {str(e)}")
|
86 |
+
return None
|
87 |
+
|
88 |
+
def speech_to_text(audio_file):
|
89 |
+
try:
|
90 |
+
# Save the uploaded audio to a temporary file
|
91 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix='.wav') as temp_audio:
|
92 |
+
audio_file.save(temp_audio.name)
|
93 |
+
|
94 |
+
# Use SpeechRecognition to convert speech to text
|
95 |
+
with sr.AudioFile(temp_audio.name) as source:
|
96 |
+
# Adjust recognition settings
|
97 |
+
recognizer.dynamic_energy_threshold = True
|
98 |
+
recognizer.energy_threshold = 4000
|
99 |
+
|
100 |
+
# Record the entire audio file
|
101 |
+
audio = recognizer.record(source)
|
102 |
+
|
103 |
+
# Perform recognition with increased timeout
|
104 |
+
text = recognizer.recognize_google(audio, language='en-US')
|
105 |
+
return text
|
106 |
+
|
107 |
+
except sr.UnknownValueError:
|
108 |
+
return "Could not understand audio"
|
109 |
+
except sr.RequestError as e:
|
110 |
+
return f"Could not request results; {str(e)}"
|
111 |
+
except Exception as e:
|
112 |
+
print(f"Error in speech_to_text: {str(e)}")
|
113 |
+
return None
|
114 |
+
finally:
|
115 |
+
# Clean up temporary file
|
116 |
+
try:
|
117 |
+
os.unlink(temp_audio.name)
|
118 |
+
except:
|
119 |
+
pass
|
120 |
+
|
121 |
+
@app.route('/')
|
122 |
+
def index():
|
123 |
+
return render_template('index.html')
|
124 |
+
|
125 |
+
@app.route('/api/chat', methods=['POST'])
|
126 |
+
def chat():
|
127 |
+
try:
|
128 |
+
data = request.get_json()
|
129 |
+
user_message = data.get('message', '')
|
130 |
+
conversation_id = data.get('conversation_id', str(uuid.uuid4()))
|
131 |
+
|
132 |
+
if not user_message:
|
133 |
+
return jsonify({'error': 'No message provided'}), 400
|
134 |
+
|
135 |
+
# Get response from Groq
|
136 |
+
response = chat_with_groq(user_message, conversation_id)
|
137 |
+
|
138 |
+
# Generate voice response
|
139 |
+
audio_io = text_to_speech(response)
|
140 |
+
result = {
|
141 |
+
'response': response,
|
142 |
+
'conversation_id': conversation_id
|
143 |
+
}
|
144 |
+
|
145 |
+
if audio_io:
|
146 |
+
audio_base64 = base64.b64encode(audio_io.getvalue()).decode('utf-8')
|
147 |
+
result['voice_response'] = audio_base64
|
148 |
+
|
149 |
+
return jsonify(result)
|
150 |
+
|
151 |
+
except Exception as e:
|
152 |
+
return jsonify({'error': str(e)}), 500
|
153 |
+
|
154 |
+
@app.route('/api/voice', methods=['POST'])
|
155 |
+
def handle_voice():
|
156 |
+
try:
|
157 |
+
if 'audio' not in request.files:
|
158 |
+
return jsonify({'error': 'No audio file provided'}), 400
|
159 |
+
|
160 |
+
audio_file = request.files['audio']
|
161 |
+
conversation_id = request.form.get('conversation_id', str(uuid.uuid4()))
|
162 |
+
|
163 |
+
# Save the audio file temporarily with a .wav extension
|
164 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix='.wav') as temp_audio:
|
165 |
+
audio_file.save(temp_audio.name)
|
166 |
+
|
167 |
+
# Use FFmpeg to convert the audio to the correct format
|
168 |
+
output_path = temp_audio.name + '_converted.wav'
|
169 |
+
os.system(f'ffmpeg -i {temp_audio.name} -acodec pcm_s16le -ac 1 -ar 16000 {output_path}')
|
170 |
+
|
171 |
+
try:
|
172 |
+
# Use the converted file for speech recognition
|
173 |
+
with sr.AudioFile(output_path) as source:
|
174 |
+
audio = recognizer.record(source)
|
175 |
+
text = recognizer.recognize_google(audio)
|
176 |
+
|
177 |
+
if not text:
|
178 |
+
return jsonify({'error': 'Could not transcribe audio'}), 400
|
179 |
+
|
180 |
+
# Get response from Groq
|
181 |
+
response = chat_with_groq(text, conversation_id)
|
182 |
+
|
183 |
+
# Generate voice response
|
184 |
+
audio_io = text_to_speech(response)
|
185 |
+
result = {
|
186 |
+
'text': text,
|
187 |
+
'response': response,
|
188 |
+
'conversation_id': conversation_id
|
189 |
+
}
|
190 |
+
|
191 |
+
if audio_io:
|
192 |
+
audio_base64 = base64.b64encode(audio_io.getvalue()).decode('utf-8')
|
193 |
+
result['voice_response'] = audio_base64
|
194 |
+
|
195 |
+
return jsonify(result)
|
196 |
+
|
197 |
+
finally:
|
198 |
+
# Clean up temporary files
|
199 |
+
try:
|
200 |
+
os.remove(temp_audio.name)
|
201 |
+
os.remove(output_path)
|
202 |
+
except:
|
203 |
+
pass
|
204 |
+
|
205 |
+
except sr.UnknownValueError:
|
206 |
+
return jsonify({'error': 'Could not understand audio'}), 400
|
207 |
+
except sr.RequestError as e:
|
208 |
+
return jsonify({'error': f'Could not request results: {str(e)}'}), 400
|
209 |
+
except Exception as e:
|
210 |
+
print(f"Error in speech_to_text: {str(e)}")
|
211 |
+
return jsonify({'error': str(e)}), 400
|
212 |
+
if __name__ == '__main__':
|
213 |
+
app.run(host='0.0.0.0', port=7860)
|