app.py

import streamlit as st
from huggingface_hub import InferenceClient
from gtts import gTTS
import os

# Initialize the Hugging Face InferenceClient
client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")

def generate_response(message, system_message, max_tokens, temperature, top_p):
    # Prepare the conversation history
    messages = [{"role": "system", "content": system_message}]
    messages.append({"role": "user", "content": message})

    response = ""

    # Get the response from the Hugging Face model
    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content
        response += token

    return response

# Streamlit app layout
st.title("Hugging Face Chat with Voice Response")

system_message = st.text_input("System Message", value="You are a friendly chatbot.")
user_message = st.text_input("Your Message", value="")
max_tokens = st.slider("Max Tokens", 1, 2048, 512)
temperature = st.slider("Temperature", 0.1, 4.0, 0.7)
top_p = st.slider("Top-p (nucleus sampling)", 0.1, 1.0, 0.95)

if st.button("Send Message"):
    # Generate response from Hugging Face model
    response_text = generate_response(user_message, system_message, max_tokens, temperature, top_p)

    # Display the text response
    st.write("Response:", response_text)

    # Convert text to speech
    tts = gTTS(text=response_text, lang='en')
    audio_file = "response.mp3"
    tts.save(audio_file)

    # Play the audio file
    audio_bytes = open(audio_file, "rb").read()
    st.audio(audio_bytes, format="audio/mp3")