import os import io import streamlit as st from PyPDF2 import PdfReader from dotenv import load_dotenv from langchain.embeddings.openai import OpenAIEmbeddings from langchain.text_splitter import CharacterTextSplitter from langchain.vectorstores import FAISS # from constants import api_key from langchain.chains.question_answering import load_qa_chain from langchain.llms import OpenAI def main(): # set the openai enviroment using api_key load_dotenv() st.title("PDF Reader") # File uploader uploaded_file = st.file_uploader("Upload a PDF file", type=["pdf"]) if uploaded_file is not None: st.write("PDF uploaded successfully!") # Read PDF contents pdf_contents = uploaded_file.read() pdf_file_like_object = io.BytesIO(pdf_contents) pdfreader = PdfReader(pdf_file_like_object) from typing_extensions import Concatenate # read text from pdf raw_text = '' for i, page in enumerate(pdfreader.pages): content = page.extract_text() if content: raw_text += content # We need to split the text using Character Text Split such that it sshould not increse token size text_splitter = CharacterTextSplitter( separator = "\n", chunk_size = 800, chunk_overlap = 200, length_function = len, ) texts = text_splitter.split_text(raw_text) embeddings = OpenAIEmbeddings() document_search = FAISS.from_texts(texts, embeddings) chain = load_qa_chain(OpenAI(), chain_type="stuff") def bot(query, i): if query: docs = document_search.similarity_search(query) response = chain.run(input_documents=docs, question=query) st.text_area("Chatbot Response:", value=response, key=f"response_ker_{i}") next_query = st.text_input("Enter your next question about the PDF : ", key=i+1) i += 1 bot(query=next_query, i=i) i = 0 query = st.text_input("Enter your question about the PDF : ", key=i) bot(query=query, i=i) i = 0 # # Display number of pages st.write(f"Number of pages in PDF: {len(pdfreader.pages)}") if __name__ == "__main__": main()