Update LLMwithvoice.py

LLMwithvoice.py

@@ -1,14 +1,9 @@
 import requests
-import torch
-from transformers import AutoTokenizer
 from pydub import AudioSegment
-from IPython.display import Audio, display
-import io
+from io import BytesIO
 
-# Hugging Face API URL for RoBERTa model
+# Hugging Face API URLs
 API_URL_ROBERTA = "https://api-inference.huggingface.co/models/deepset/roberta-base-squad2"
-
-# ESPnet TTS API URL
 API_URL_TTS = "https://api-inference.huggingface.co/models/espnet/english_male_ryanspeech_tacotron"
 
 # Function to query the RoBERTa model
@@ -41,17 +36,21 @@ def generate_speech(api_token, answer):
     try:
         response.raise_for_status()  # Raise an error for bad responses
         audio = response.content
-        audio_segment = AudioSegment.from_wav(io.BytesIO(audio))
-        display(Audio(audio))
+        audio_segment = AudioSegment.from_file(BytesIO(audio), format="wav")
+        audio_file = BytesIO()
+        audio_segment.export(audio_file, format="wav")
+        audio_file.seek(0)
+        return audio_file
     except requests.exceptions.HTTPError as e:
         print(f"HTTP error occurred: {e}")
     except Exception as e:
         print(f"An unexpected error occurred: {e}")
 
 # Function to interface with Gradio
-def gradio_interface(api_token, prompt, context):
+def gradio_interface(api_token, context, prompt):
     answer = query_roberta(api_token, prompt, context)
     if 'error' in answer:
         return answer['error'], None
-    generate_speech(api_token, answer.get('answer', ''))
-    return answer.get('answer', 'No answer found'), None
+    answer_text = answer.get('answer', 'No answer found')
+    audio_file = generate_speech(api_token, answer_text)
+    return answer_text, audio_file