Spaces:
Running
Running
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import logging
|
3 |
+
from telegram import Update
|
4 |
+
from telegram.ext import Application, MessageHandler, filters
|
5 |
+
from transformers import pipeline, AutoTokenizer, VitsModel
|
6 |
+
import torchaudio
|
7 |
+
import librosa
|
8 |
+
import soundfile as sf
|
9 |
+
|
10 |
+
# تهيئة النظام
|
11 |
+
logging.basicConfig(format='%(asctime)s - %(name)s - %(levelname)s - %(message)s', level=logging.INFO)
|
12 |
+
logger = logging.getLogger(__name__)
|
13 |
+
|
14 |
+
# تهيئة النماذج
|
15 |
+
asr_pipeline = pipeline(
|
16 |
+
"automatic-speech-recognition",
|
17 |
+
model="facebook/wav2vec2-large-xlsr-53-arabic"
|
18 |
+
)
|
19 |
+
|
20 |
+
tts_tokenizer = AutoTokenizer.from_pretrained("facebook/mms-tts-ara")
|
21 |
+
tts_model = VitsModel.from_pretrained("facebook/mms-tts-ara")
|
22 |
+
|
23 |
+
# معالجة الصوت الوارد
|
24 |
+
async def process_voice(update: Update, context):
|
25 |
+
try:
|
26 |
+
# تحميل الملف الصوتي
|
27 |
+
voice_file = await update.message.voice.get_file()
|
28 |
+
await voice_file.download_to_drive("user_voice.ogg")
|
29 |
+
|
30 |
+
# تحويل الصوت إلى نص
|
31 |
+
text = await speech_to_text("user_voice.ogg")
|
32 |
+
|
33 |
+
# توليد الرد
|
34 |
+
response = await generate_response(text)
|
35 |
+
|
36 |
+
# تحويل النص إلى صوت
|
37 |
+
await text_to_speech(response)
|
38 |
+
|
39 |
+
# إرسال الرد
|
40 |
+
await update.message.reply_voice("bot_response.wav")
|
41 |
+
|
42 |
+
except Exception as e:
|
43 |
+
logger.error(f"Error: {str(e)}")
|
44 |
+
await update.message.reply_text("عذراً، حدث خطأ ما. يرجى المحاولة لاحقاً.")
|
45 |
+
|
46 |
+
# تحويل الصوت إلى نص
|
47 |
+
async def speech_to_text(audio_path):
|
48 |
+
# تحويل الصيغة من ogg إلى wav
|
49 |
+
audio, sr = librosa.load(audio_path, sr=16000)
|
50 |
+
sf.write("temp.wav", audio, sr)
|
51 |
+
|
52 |
+
# التعرف على الكلام
|
53 |
+
result = asr_pipeline("temp.wav")
|
54 |
+
return result["text"]
|
55 |
+
|
56 |
+
# توليد الردود
|
57 |
+
async def generate_response(text):
|
58 |
+
chatbot = pipeline("text-generation", model="aubmindlab/aragpt2-base")
|
59 |
+
response = chatbot(
|
60 |
+
text,
|
61 |
+
max_length=100,
|
62 |
+
num_return_sequences=1,
|
63 |
+
pad_token_id=50256
|
64 |
+
)
|
65 |
+
return response[0]['generated_text']
|
66 |
+
|
67 |
+
# تحويل النص إلى صوت
|
68 |
+
async def text_to_speech(text):
|
69 |
+
inputs = tts_tokenizer(text, return_tensors="pt")
|
70 |
+
|
71 |
+
with torch.no_grad():
|
72 |
+
output = tts_model(**inputs)
|
73 |
+
|
74 |
+
waveform = output.waveform[0].numpy()
|
75 |
+
torchaudio.save("bot_response.wav", torch.Tensor(waveform), tts_model.config.sampling_rate)
|
76 |
+
|
77 |
+
# تهيئة التطبيق
|
78 |
+
if __name__ == "__main__":
|
79 |
+
TOKEN = os.getenv("TELEGRAM_TOKEN")
|
80 |
+
application = Application.builder().token(TOKEN).build()
|
81 |
+
application.add_handler(MessageHandler(filters.VOICE, process_voice))
|
82 |
+
application.run_polling()
|