app.py

import streamlit as st
import os
from PyPDF2 import PdfReader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings import HuggingFaceHubEmbeddings  # Cambiado desde HuggingFaceEmbeddings
from langchain.vectorstores import FAISS
from langchain.chat_models import ChatOpenAI
from langchain.chains.question_answering import load_qa_chain

st.set_page_config(page_title='preguntaDOC')
st.header("Pregunta a tu PDF")

OPENAI_API_KEY = st.text_input('OpenAI API Key', type='password')
HUGGINGFACE_API_KEY = st.text_input('Hugging Face API Key', type='password')  # Añadido para la API de Hugging Face

pdf_obj = st.file_uploader("Carga tu documento", type="pdf", on_change=st.cache_resource.clear)

@st.cache_resource 
def create_embeddings(pdf, hf_api_key):
    os.environ["HUGGINGFACEHUB_API_TOKEN"] = hf_api_key  # Configurar token de HF
    
    pdf_reader = PdfReader(pdf)
    text = ""
    for page in pdf_reader.pages:
        text += page.extract_text()
    
    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size=800,
        chunk_overlap=100,
        length_function=len
    )        
    chunks = text_splitter.split_text(text)
    
    # Usar HuggingFaceHubEmbeddings en lugar de HuggingFaceEmbeddings
    embeddings = HuggingFaceHubEmbeddings(
        repo_id="sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2",
        huggingfacehub_api_token=hf_api_key
    )
    
    knowledge_base = FAISS.from_texts(chunks, embeddings)
    return knowledge_base

if pdf_obj and HUGGINGFACE_API_KEY:
    knowledge_base = create_embeddings(pdf_obj, HUGGINGFACE_API_KEY)
    user_question = st.text_input("Haz una pregunta sobre tu PDF:")
    
    if user_question and OPENAI_API_KEY:
        os.environ["OPENAI_API_KEY"] = OPENAI_API_KEY
        docs = knowledge_base.similarity_search(user_question, 3)
        llm = ChatOpenAI(model_name='gpt-3.5-turbo')
        chain = load_qa_chain(llm, chain_type="stuff")
        
        with st.spinner("Procesando tu pregunta..."):
            try:
                respuesta = chain.run(input_documents=docs, question=user_question)
                st.write(respuesta)
            except Exception as e:
                st.error(f"Error: {str(e)}")
elif pdf_obj and not HUGGINGFACE_API_KEY:
    st.warning("Por favor, introduce una clave API de Hugging Face para procesar el documento.")