app.py

import streamlit as st
from huggingface_hub import InferenceClient
from gtts import gTTS
import IPython.display as ipd

client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
system_prompt = "Tu nombre es Xaman 3.0"
system_prompt_sent = False

def format_prompt(message, history):
    global system_prompt_sent
    prompt = "<s>"

    if history is not None and isinstance(history, list):
        if not any(f"[INST] {system_prompt} [/INST]" in user_prompt for user_prompt, _ in history):
            prompt += f"[INST] {system_prompt} [/INST]"
            system_prompt_sent = True

        for user_prompt, bot_response in history:
            prompt += f"[INST] {user_prompt} [/INST]"
            prompt += f" {bot_response}</s> "

    prompt += f"[INST] {message} [/INST]"
    return prompt

def text_to_speech(text):
    tts = gTTS(text=text, lang='es')
    tts.save('output.mp3')
    return 'output.mp3'

def generate(
    user_input, history, temperature=None, max_new_tokens=2048, top_p=0.95, repetition_penalty=1.0,
):
    global system_prompt_sent
    temperature = float(temperature) if temperature is not None else 0.9
    if temperature < 1e-2:
        temperature = 1e-2
    top_p = float(top_p)

    generate_kwargs = dict(
        temperature=temperature,
        max_new_tokens=max_new_tokens,
        top_p=top_p,
        repetition_penalty=repetition_penalty,
        do_sample=True,
        seed=42,
    )

    formatted_prompt = format_prompt(user_input, history)
    
    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
    output = ""

    for response in stream:
        output += response.token.text
        st.text(output)
        ipd.Audio(text_to_speech(output), autoplay=True)
    
    return output

user_input = st.text_input(label="Usuario", value="Escribe aquí tu mensaje")
output = generate(user_input, history=None)  # Ajusta 'history' según sea necesario