app.py

import streamlit as st
from huggingface_hub import InferenceClient
import base64
from pydub import AudioSegment
from io import BytesIO
from gtts import gTTS
from streamlit_webrtc import webrtc_streamer, WebRtcMode
import speech_recognition as sr
import sounddevice as sd  

client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
pre_prompt = ""
pre_prompt_sent = False
webrtc_ctx = None 

def take_user_input():
    r = sr.Recognizer()

    def audio_callback(in_data, frame_count, time_info, status):
        global webrtc_ctx 
        audio = sr.AudioData(
            in_data.tobytes(),
            sample_rate=webrtc_ctx.audio_sample_rate,
            sample_width=sd.default.dtype.itemsize
        )

        st.info('Reconociendo...')
        query = transcribe_speech(audio)

        if 'salir' in query or 'detener' in query:
            speak("Hasta luego.")
            exit()
        return query

    global webrtc_ctx 
    webrtc_ctx = webrtc_streamer(
        key="microphone",
        mode=WebRtcMode.SENDRECV,
        audio_receiver=audio_callback,
        async_processing=True,
    )

    if not webrtc_ctx:
        st.warning("Por favor, habilita el micrófono.")
        return 'None'

    st.info('Escuchando...')

    try:
        with sd.InputStream(callback=lambda indata, frames, time, status: None):
            while True:
                audio_data = webrtc_ctx.audio_receiver_stream.get()
                if audio_data:
                    audio = sr.AudioData(
                        audio_data.tobytes(),
                        sample_rate=webrtc_ctx.audio_sample_rate,
                        sample_width=audio_data.itemsize
                    )

                    st.info('Reconociendo...')

                    query = transcribe_speech(audio)

                    if 'salir' in query or 'detener' in query:
                        speak("Hasta luego.")
                        exit()
                    return query

    except sr.UnknownValueError:
        speak('No se ha reconocido nada. Intenta de nuevo...')
    except sr.RequestError as e:
        st.error(f"Error en la solicitud al reconocimiento de voz: {e}")

    return 'None'


def audio_callback(in_data, frame_count, time_info, status):
    return in_data, webrtc_ctx.audio_sample_rate

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

    if not pre_prompt_sent and all(f"[INST] {pre_prompt} [/INST]" not in user_prompt for user_prompt, _ in history):
        prompt += f"[INST] {pre_prompt} [/INST]"
        pre_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 generate(user_input, history, temperature=None, max_new_tokens=512, top_p=0.95, repetition_penalty=1.0):
    global pre_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)
    
    try:
        stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=True)
        response = ""
        for response_token in stream:
            response += response_token.token.text
        response = ' '.join(response.split()).replace('</s>', '')
        audio_bytes = text_to_speech(response)
        return response, audio_bytes
    except Exception as e:
        return str(e), None

def text_to_speech(text):
    tts = gTTS(text=text, lang='es')
    audio_stream = BytesIO()
    tts.save(audio_stream)
    audio_stream.seek(0)
    return audio_stream.read()

def speak(text):
    audio_bytes = text_to_speech(text)
    st.audio(audio_bytes, format="audio/mp3", start_time=0, key="audio_player")

if "history" not in st.session_state:
    st.session_state.history = []

user_input = take_user_input()
output, audio_bytes = generate(user_input, history=st.session_state.history)

with st.container(width=900, height=400):
    user_input_container = st.text_input("Tu entrada de usuario", value=user_input)
    st.text_area("Respuesta", value=output, key="output_text", disabled=True)

if audio_bytes is not None:
    st.markdown(
        f"""
        <audio autoplay="autoplay" controls="controls" src="data:audio/mp3;base64,{base64.b64encode(audio_bytes).decode()}" type="audio/mp3" id="audio_player"></audio>
        """,
        unsafe_allow_html=True
    )