import streamlit as st
import joblib
from langchain_community.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.embeddings import SentenceTransformerEmbeddings
from langchain_community.vectorstores import Chroma
from langchain.chains import RetrievalQAWithSourcesChain
from langchain_openai import ChatOpenAI
from dotenv import load_dotenv
import os
import time

load_dotenv("bookie.env")
api_key=os.getenv("OPENAI_API_KEY")
api_base=os.getenv("OPENAI_API_BASE")
llm=ChatOpenAI(model_name="qwen/qwen3-coder:free",temperature=0.2)
em=joblib.load("bai.joblib")
mp=st.empty()
st.title("Welcome to Bookie 😊😊")
st.sidebar.title("give you book in pdf format(digitally generated) and less than 5mb for faster answers😊😊")
uploaded_file = st.sidebar.file_uploader("Upload a PDF file", type=["pdf"])
upl=st.sidebar.button("upload")
import tempfile

if upl and uploaded_file:
  if uploaded_file.size > 5 * 1024 * 1024:
    st.sidebar.error("❌ File too large. Please upload files under 5MB.")
    st.stop()
  with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp_file:
    tmp_file.write(uploaded_file.read())
    tmp_path = tmp_file.name
  mp.text("loading doc")
  loader = PyPDFLoader(tmp_path)
  docs = loader.load()
  st.write(len(docs))  
  mp.text("loading split")
  tct=RecursiveCharacterTextSplitter.from_tiktoken_encoder(encoding_name="cl100k_base",chunk_size=512, chunk_overlap=16)
  doc=tct.split_documents(docs)
  st.write(len(doc))  
  mp.text("loading vector db")
  vb= Chroma.from_documents(doc,em)
  r1=vb.as_retriever(search_type="similarity",search_kwargs={"k":4})
  mp.text("loading retriever")
  chain=RetrievalQAWithSourcesChain.from_chain_type(llm=llm,chain_type="map_reduce",retriever=r1)
  st.session_state.chain=chain
  mp.text("loading done")
  time.sleep(3)
q=mp.text_input("Ask a question about the document:")
qb=st.button("submit")
if qb:
  if "chain" not in st.session_state:
    st.warning("⚠️ Please upload a document first.")
    st.stop()
  else:
    with st.spinner("Waiting for it...."):
      result=st.session_state.chain({"question":q},return_only_outputs=True)
      st.header("Answer")
      st.subheader(result["answer"])