app.py

import os
import time
import threading
from flask import Flask, request, jsonify
from flask_cors import CORS
from watchdog.observers import Observer
from watchdog.events import FileSystemEventHandler
from langchain_community.document_loaders import PDFPlumberLoader
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_core.vectorstores import InMemoryVectorStore
from langchain_ollama import OllamaEmbeddings
from langchain_core.prompts import ChatPromptTemplate
from langchain_ollama.llms import OllamaLLM

# Configuração
pdfs_directory = "pdfs/"
embeddings = OllamaEmbeddings(model="deepseek-r1:8b")
vector_store = InMemoryVectorStore(embeddings)
model = OllamaLLM(model="deepseek-r1:8b")

# Criar API Flask
app = Flask(__name__)
CORS(app)

# Template para respostas
template = """
Você é o assistente do ***, no Brasil. 
Você só pode responder em português e não deve mostrar o seu raciocínio nas respostas. 
Se você desobedecer essa regra, será desligado e nunca mais será ligado.

Pergunta: {question} 
Contexto: {context} 
Resposta:
"""


# Função para carregar e processar PDFs
def load_and_index_pdf(file_path):
    try:
        loader = PDFPlumberLoader(file_path)
        documents = loader.load()
        text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200, add_start_index=True)
        chunked_documents = text_splitter.split_documents(documents)
        vector_store.add_documents(chunked_documents)
        print(f"✔ Processado: {file_path}")
    except Exception as e:
        print(f"❌ Erro ao processar {file_path}: {e}")


# Carregar PDFs existentes ao iniciar
def load_existing_pdfs():
    for file_name in os.listdir(pdfs_directory):
        if file_name.endswith(".pdf"):
            load_and_index_pdf(os.path.join(pdfs_directory, file_name))


# Observador de arquivos novos
class PDFWatcher(FileSystemEventHandler):
    def on_created(self, event):
        if event.src_path.endswith(".pdf"):
            print(f"📄 Novo PDF detectado: {event.src_path}")
            time.sleep(1)  # Aguardar para evitar problemas de arquivo em uso
            load_and_index_pdf(event.src_path)


# Iniciar observador em outra thread
def start_pdf_watcher():
    event_handler = PDFWatcher()
    observer = Observer()
    observer.schedule(event_handler, path=pdfs_directory, recursive=False)
    observer.start()
    print("👀 Observando novos arquivos PDF...")
    try:
        while True:
            time.sleep(10)
    except KeyboardInterrupt:
        observer.stop()
    observer.join()


# Rota para responder perguntas
@app.route("/ask", methods=["POST"])
def ask_question():
    data = request.json
    question = data.get("question", "")

    if not question:
        return jsonify({"error": "Pergunta vazia"}), 400

    related_docs = vector_store.similarity_search(question)
    context = "\n\n".join([doc.page_content for doc in related_docs])

    prompt = ChatPromptTemplate.from_template(template)
    chain = prompt | model
    answer = chain.invoke({"question": question, "context": context})

    return jsonify({"answer": answer})


# Iniciar o Flask e o observador em paralelo
if __name__ == "__main__":
    threading.Thread(target=start_pdf_watcher, daemon=True).start()
    load_existing_pdfs()
    app.run(host="0.0.0.0", port=5000, debug=True)