Update app.py

app.py

@@ -1,34 +1,38 @@
 import streamlit as st
 from transformers import pipeline
+from io import BytesIO
 
-# Load a text generation model
+# Load Hugging Face models for question-answering and text-to-speech
 @st.cache_resource
-def load_model():
-    return pipeline("text-generation", model="EleutherAI/gpt-neo-125M")
+def load_qa_pipeline():
+    return pipeline("question-answering", model="distilbert-base-uncased-distilled-squad")
 
-# Initialize the model
-generator = load_model()
+@st.cache_resource
+def load_tts_pipeline():
+    return pipeline("text-to-speech", model="espnet/kan-bayashi-ljspeech-vits")
 
-st.title("Trump-Inspired Text Generator")
-st.write("Enter a prompt to receive a response.")
+qa_pipeline = load_qa_pipeline()
+tts_pipeline = load_tts_pipeline()
 
-# User input
-user_input = st.text_input("Enter your prompt:")
+# Streamlit interface
+st.title("Virtual Assistant")
+st.write("Ask me anything!")
 
-if user_input:
-    # Hidden style prompt for model guidance, but not shown to user
-    trump_prompt = f"In the style of Donald Trump: {user_input}"
-    
-    # Generate the response with adjusted parameters
-    response = generator(
-        trump_prompt, 
-        max_length=30,   
-        num_return_sequences=1, 
-        do_sample=True, 
-        temperature=0.9, 
-        top_p=0.85       
-    )
+# User query
+user_query = st.text_input("Type your question here:")
 
-    # Extract and display the generated text
-    generated_text = response[0]["generated_text"]
-    st.write("Response:", generated_text)
+if user_query:
+    # Generate answer using the QA model
+    context = "This is the context of the assistant. The assistant will answer general knowledge questions."  # Customize context for better QA accuracy
+    qa_result = qa_pipeline({"question": user_query, "context": context})
+    answer = qa_result['answer']
+    
+    # Display answer as text
+    st.write(f"Answer: {answer}")
+    
+    # Convert answer to audio using TTS model
+    tts_audio = tts_pipeline(answer, return_tensors="pt").audio
+    
+    # Streamlit audio player for TTS output
+    audio_bytes = BytesIO(tts_audio)
+    st.audio(audio_bytes, format="audio/wav")