app.py

import streamlit as st
from haystack.document_stores import InMemoryDocumentStore
from haystack.nodes import TransformersSummarizer, PreProcessor, PDFToTextConverter, TfidfRetriever
from haystack.schema import Document
import logging

@st.cache(hash_funcs={"builtins.SwigPyObject": lambda _: None},allow_output_mutation=True)
def start_haystack():
    document_store = InMemoryDocumentStore()
    preprocessor = PreProcessor(
        clean_empty_lines=True,
        clean_whitespace=True,
        clean_header_footer=True,
        split_by="word",
        split_length=100,
        split_respect_sentence_boundary=True,
    )
    summarizer = TransformersSummarizer(model_name_or_path="google/pegasus-newsroom")
    return document_store, summarizer, preprocessor


def pdf_to_document_store(pdf_files):
    converter = PDFToTextConverter(remove_numeric_tables=True, valid_languages=["en"])
    documents = []
    for pdf in pdf_files:
        doc = converter.convert(file_path=pdf.name, meta=None)[0]
        st.write(len(doc))
        st.write(doc)
        preprocessed_doc=preprocessor.process([doc])
        st.write(len(preprocessed_doc))
        documents.append(preprocessed_doc)
    document_store.write_documents(documents)
    st.write('Document count: ', document_store.get_document_count())


def summarize(files):
    pdf_to_document_store(files)
    summary = summarizer.predict(documents=document_store.get_all_documents(), generate_single_summary=False)
    st.write(summary)

document_store, summarizer, preprocessor = start_haystack()

uploaded_files = st.file_uploader("Choose PDF files", accept_multiple_files=True)

if uploaded_files is not None:
    st.write(len(uploaded_files))
    if st.button('Summarize Documents'):
        summarize(uploaded_files)

if st.button('Calculate num of docs'):
    st.write(document_store.get_document_count())

if st.button('Clear DocumentStore'):
    document_store.delete_all_documents()