Rename utils.py to twilio_webhook.py

twilio_webhook.py

@@ -0,0 +1,60 @@
+# twilio_webhook.py
+from flask import Flask, request
+from twilio.rest import Client
+from twilio.twiml.voice_response import VoiceResponse
+import requests
+
+app = Flask(__name__)
+
+@app.route("/voice", methods=['GET', 'POST'])
+def voice():
+    # Create TwiML response
+    resp = VoiceResponse()
+    
+    # Greet the caller
+    resp.say("Hello! Welcome to our customer support. Please state your query.")
+    
+    # Record the caller's message
+    resp.record(
+        action='/handle_recording',
+        method='POST',
+        max_length=30,
+        finish_on_key='#'
+    )
+    
+    return str(resp)
+
+@app.route("/handle_recording", methods=['POST'])
+def handle_recording():
+    # Get the recording URL from Twilio
+    recording_url = request.values.get('RecordingUrl')
+    
+    # Download the recording
+    audio_response = requests.get(recording_url)
+    
+    # Send to Hugging Face Space for processing
+    hf_space_url = "YOUR_HUGGING_FACE_SPACE_URL/api/predict"
+    
+    files = {"data": [audio_response.content]}
+    response = requests.post(hf_space_url, json=files)
+    
+    # Get the AI response
+    ai_response = response.json()
+    
+    # Create TwiML to speak the response
+    resp = VoiceResponse()
+    resp.say(ai_response['data'][0])
+    
+    # Ask if they need more help
+    resp.say("Is there anything else I can help you with?")
+    resp.record(
+        action='/handle_recording',
+        method='POST',
+        max_length=30,
+        finish_on_key='#'
+    )
+    
+    return str(resp)
+
+if __name__ == "__main__":
+    app.run(debug=True)

utils.py

@@ -1,50 +0,0 @@
-from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM, VitsModel
-import soundfile as sf
-import torch
-import io
-import os
-
-# Speech-to-Text (Whisper)
-def transcribe_audio(audio_path):
-    try:
-        whisper = pipeline("automatic-speech-recognition", model="openai/whisper-tiny")
-        audio, sample_rate = sf.read(audio_path)
-        if sample_rate != 8000:  # Convert to 8kHz for Twilio compatibility
-            audio = sf.read(audio_path, samplerate=8000)[0]
-            sf.write(audio_path, audio, 8000)
-        result = whisper(audio_path)
-        return result["text"]
-    except Exception as e:
-        print(f"STT Error: {e}")
-        return "Sorry, I couldn't understand that."
-
-# NLP (Falcon-7B-Instruct)
-def generate_response(text):
-    try:
-        tokenizer = AutoTokenizer.from_pretrained("tiiuae/falcon-7b-instruct")
-        model = AutoModelForCausalLM.from_pretrained("tiiuae/falcon-7b-instruct")
-        prompt = (
-            "You are a polite and helpful customer support agent. Respond professionally.\n"
-            f"User: {text}\nAgent:"
-        )
-        inputs = tokenizer(prompt, return_tensors="pt")
-        outputs = model.generate(**inputs, max_length=200, do_sample=True, top_p=0.9)
-        response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-        return response.split("Agent:")[1].strip()
-    except Exception as e:
-        print(f"NLP Error: {e}")
-        return "I'm having trouble processing your request. Please try again."
-
-# Text-to-Speech (VITS)
-def text_to_speech(text, output_path="output.wav"):
-    try:
-        tts_model = VitsModel.from_pretrained("facebook/mms-tts-eng")
-        tts_tokenizer = AutoTokenizer.from_pretrained("facebook/mms-tts-eng")
-        inputs = tts_tokenizer(text, return_tensors="pt")
-        with torch.no_grad():
-            waveform = tts_model(**inputs).waveform
-        sf.write(output_path, waveform.squeeze().numpy(), 8000)  # 8kHz for Twilio
-        return output_path
-    except Exception as e:
-        print(f"TTS Error: {e}")
-        return None