app.py

# -*- coding: utf-8 -*-
"""app

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/drive/1ZybFOpX1r-SAA-RslP5WJkQ9gdI6JCCj
"""
import streamlit as st
import os
from langchain.chat_models import ChatOpenAI
from langchain.document_loaders import PyPDFLoader
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import FAISS
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.chains import RetrievalQA
import tempfile

st.set_page_config(page_title="Chat com PDF", layout="centered")
st.title("📄 Chat com PDF usando LangChain")

uploaded_file = st.file_uploader("📤 Envie um arquivo PDF", type="pdf")

if uploaded_file is not None:
    with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp:
        tmp.write(uploaded_file.read())
        pdf_path = tmp.name

    with st.spinner("🔍 Processando o PDF..."):
        try:
            # Carregar e dividir o PDF
            loader = PyPDFLoader(pdf_path)
            documents = loader.load()

            text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
            docs = text_splitter.split_documents(documents)

            # Gerar embeddings
            embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")
            vectorstore = FAISS.from_documents(docs, embeddings)

            # Criar modelo LLM
            llm = ChatOpenAI(
                openai_api_base="https://openrouter.ai/api/v1",
                openai_api_key=os.environ["OPENROUTER_API_KEY"],
                model='deepseek/deepseek-r1-zero:free'
            )

            # Criar a cadeia de QA
            qa_chain = RetrievalQA.from_chain_type(
                llm=llm,
                retriever=vectorstore.as_retriever(),
                return_source_documents=True
            )

            # Interface para pergunta
            pergunta = st.text_input("❓ Faça uma pergunta sobre o PDF:")

            if pergunta:
                resposta = qa_chain.invoke({"query": pergunta})

                st.success("✅ Resposta:")
                st.write(resposta['result'])

                with st.expander("📄 Fontes usadas"):
                    for i, doc in enumerate(resposta['source_documents']):
                        st.markdown(f"**Fonte {i+1}:**\n\n{doc.page_content[:500]}...")

        except Exception as e:
            st.error(f"Erro: {str(e)}")