app.py

import streamlit as st
from markdown import markdown
from annotated_text import annotation
import logging

from haystack.document_stores import InMemoryDocumentStore
from haystack.nodes import TfidfRetriever
from haystack.pipelines import ExtractiveQAPipeline
from haystack.nodes import FARMReader
import time
import joblib

@st.cache(hash_funcs={"builtins.SwigPyObject": lambda _: None},allow_output_mutation=True)
def create_pipeline():
    docs = joblib.load('docs.joblib')

    document_store = InMemoryDocumentStore()
    document_store.write_documents(docs)
    
    retriever = TfidfRetriever(document_store)
    reader = FARMReader(model_name_or_path="ixa-ehu/SciBERT-SQuAD-QuAC")
    
    pipeline = ExtractiveQAPipeline(reader, retriever)
    
    return pipeline
    
pipeline = create_pipeline()

def set_state_if_absent(key, value):
    if key not in st.session_state:
        st.session_state[key] = value

set_state_if_absent("question", 'Applications of AI and deep learning')
set_state_if_absent("results", None)

def reset_results(*args):
    st.session_state.results = None

st.markdown('''# Welcome to **SRM RP explorer**!
This QA demo uses a [Haystack Extractive QA Pipeline](https://haystack.deepset.ai/components/ready-made-pipelines#extractiveqapipeline) with 
an [InMemoryDocumentStore](https://haystack.deepset.ai/components/document-store) which contains abstracts of 17k+ research papers associated with SRM university''')

query = st.text_input('Enter a query to get started:', value=st.session_state.question, max_chars=100, on_change=reset_results)

def ask_question(query):
    start = time.time()
    prediction = pipeline.run(query=query, params={"Retriever": {"top_k": 10}, "Reader": {"top_k": 5}})
    st.write('Time taken: %s s' % round(time.time()-start, 2))
    
    results = []
    for answer in prediction["answers"]:
        answer = answer.to_dict()
        if answer["answer"]:
            results.append(
                {
                    "title":answer["meta"]["name"],
                    "link":answer["meta"]["link"],
                    "context": "..." + answer["context"] + "...",
                    "answer": answer["answer"],
                    "score": round(answer["score"] * 100, 2),
                    "offset_start_in_doc": answer["offsets_in_document"][0]["start"],
                }
            )
        else:
            results.append(
                {
                    "title":None,
                    "link":None,
                    "context": None,
                    "answer": None,
                    "score": round(answer["score"] * 100, 2),
                }
            )
    return results

if query:
    with st.spinner("👑    Performing semantic search on abstracts..."):
        try:
            msg = 'Asked ' + query
            logging.info(msg)
            st.session_state.results = ask_question(query)    
        except Exception as e:
            logging.exception(e)

if st.session_state.results:
    st.write('## Top Results')
    for count, result in enumerate(st.session_state.results):
        if result["answer"]:
            answer, context = result["answer"], result["context"]
            start_idx = context.find(answer)
            end_idx = start_idx + len(answer)
            st.write(
                markdown(context[:start_idx] + str(annotation(body=answer, label="RELEVANT", background="#964448", color='#ffffff')) + context[end_idx:]),
                unsafe_allow_html=True,
            )
            st.markdown(f"**Title:** [{result['title']}]({result['link']})\n**Relevance:** {result['score']}")
        else:
            st.info(
                "🤔    Haystack is unsure whether any of the documents contain an answer to your question. Try to reformulate it!"
            )