app.py

import time
import streamlit as st
import logging
import pandas as pd
from json import JSONDecodeError
from markdown import markdown
import random
from typing import List, Dict, Any, Tuple, Optional

from haystack.document_stores import FAISSDocumentStore
from haystack.nodes import EmbeddingRetriever
from haystack.pipelines import ExtractiveQAPipeline
from haystack.nodes import FARMReader
from haystack.pipelines import ExtractiveQAPipeline
from annotated_text import annotation
import shutil
from urllib.parse import unquote


# FAISS index directory
INDEX_DIR = 'data/index'


# the following function is cached to make index and models load only at start
@st.cache(hash_funcs={"builtins.SwigPyObject": lambda _: None}, allow_output_mutation=True)
def start_haystack():
  """
  load document store, retriever, reader and create pipeline
  """
  shutil.copy(f'{INDEX_DIR}/faiss_document_store.db','.')
  document_store = FAISSDocumentStore(
      faiss_index_path=f'{INDEX_DIR}/my_faiss_index.faiss',
      faiss_config_path=f'{INDEX_DIR}/my_faiss_index.json')
  print (f'Index size: {document_store.get_document_count()}')  
  retriever = EmbeddingRetriever(
      document_store=document_store,
    embedding_model="sentence-transformers/multi-qa-mpnet-base-dot-v1",
    model_format="sentence_transformers"
  )
  reader = FARMReader(model_name_or_path="deepset/roberta-base-squad2",
                        use_gpu=False,
                        confidence_threshold=0.15)
  pipe = ExtractiveQAPipeline(reader, retriever)
  return pipe

@st.cache()
def load_questions():
    with open('./data/questions.txt') as fin:
        questions = [line.strip() for line in fin.readlines()
                    if not line.startswith('#')]
    return questions    

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

def query(pipe, question):
    """Run query and get answers"""
    return (pipe.run(question, params={"Retriever": {"top_k": 10}, "Reader": {"top_k": 5}}), None)


def main():
    # st.set_page_config(page_title='Who killed Laura Palmer?',
    # page_icon="https://static.wikia.nocookie.net/twinpeaks/images/4/4a/Site-favicon.ico/revision/latest?cb=20210710003705")
    
    pipe=start_haystack()
    questions = load_questions()

    # Persistent state
    set_state_if_absent('question', "Where is Twin Peaks?")
    set_state_if_absent('answer', '')
    set_state_if_absent('results', None)
    set_state_if_absent('raw_json', None)
    set_state_if_absent('random_question_requested', False)

    # Small callback to reset the interface in case the text of the question changes
    def reset_results(*args):
        st.session_state.answer = None
        st.session_state.results = None
        st.session_state.raw_json = None

    # sidebar style
    st.markdown(
    """
    <style>
    [data-testid="stSidebar"][aria-expanded="true"] > div:first-child{
        width: 350px;
    }
    [data-testid="stSidebar"][aria-expanded="false"] > div:first-child{
        width: 350px;
        margin-left: -350px;
    }
    """,
    unsafe_allow_html=True,
    )
    # Title
    st.write("# Who killed Laura Palmer?")
    st.write("### The first Twin Peaks Question Answering system!")
    
    st.markdown("""
Ask any question about Twin Peaks [Twin Peaks] (https://twinpeaks.fandom.com/wiki/Twin_Peaks) 
and see if the AI ​​can find an answer...

*Note: do not use keywords, but full-fledged questions.*
""")

    # Sidebar
    st.sidebar.header("Who killed Laura Palmer?")   
    st.sidebar.image("https://upload.wikimedia.org/wikipedia/it/3/39/Twin-peaks-1990.jpg")
    st.sidebar.markdown('<p align="center"><b>Twin Peaks Question Answering system</b></p>', unsafe_allow_html=True)
    st.sidebar.markdown(f"""
    <style>
        a {{
            text-decoration: none;
        }}
        .haystack-footer {{
            text-align: center;
        }}
        .haystack-footer h4 {{
            margin: 0.1rem;
            padding:0;
        }}
        footer {{
            opacity: 0;
        }}
        .haystack-footer img {{
            display: block;
            margin-left: auto;
            margin-right: auto;
            width: 85%;
        }}
    </style>
    <div class="haystack-footer">
        <p><a href="https://github.com/anakin87/who-killed-laura-palmer">GitHub</a> - 
        Built with <a href="https://github.com/deepset-ai/haystack/">Haystack</a><br/>
        <small>Data crawled from <a href="https://twinpeaks.fandom.com/wiki/Twin_Peaks_Wiki">Twin Peaks Wiki</a>.</small>       
    </p>
    <img src = 'https://static.wikia.nocookie.net/twinpeaks/images/e/ef/Laura_Palmer%2C_the_Queen_Of_Hearts.jpg'/>
    <br/>
    </div>
    """, unsafe_allow_html=True)

    # spotify webplayer
    st.sidebar.markdown("""
    <p align="center">
    <iframe style="border-radius:12px" src="https://open.spotify.com/embed/playlist/38rrtWgflrw7grB37aMlsO?utm_source=generator" width="85%" height="380" frameBorder="0" allowfullscreen="" allow="autoplay; clipboard-write; encrypted-media; fullscreen; picture-in-picture"></iframe>
    </p>""", unsafe_allow_html=True)   

    # Search bar
    question = st.text_input("",
        value=st.session_state.question,
        max_chars=100,
        on_change=reset_results
    )
    col1, col2 = st.columns(2)
    col1.markdown("<style>.stButton button {width:100%;}</style>", unsafe_allow_html=True)
    col2.markdown("<style>.stButton button {width:100%;}</style>", unsafe_allow_html=True)

    # Run button
    run_pressed = col1.button("Run")

    # Get next random question from the CSV
    if col2.button("Random question"):
        reset_results()
        question = random.choice(questions)
        while question == st.session_state.question:  # Avoid picking the same question twice (the change is not visible on the UI)
            question = random.choice(questions)
        st.session_state.question = question
        # st.session_state.answer = new_row["Answer"].values[0]
        st.session_state.random_question_requested = True
        # Re-runs the script setting the random question as the textbox value
        # Unfortunately necessary as the Random Question button is _below_ the textbox
        raise st.script_runner.RerunException(st.script_request_queue.RerunData(None))
    else:
        st.session_state.random_question_requested = False
    
    run_query = (run_pressed or question != st.session_state.question) and not st.session_state.random_question_requested

    # Get results for query
    if run_query and question:
        time_start=time.time()
        reset_results()
        st.session_state.question = question

        with st.spinner(
            "🧠 &nbsp;&nbsp; Performing neural search on documents..."

        ):
            try:
                st.session_state.results, st.session_state.raw_json = query(pipe, question)
                time_end=time.time()
                print(f'elapsed time: {time_end - time_start}')
            except JSONDecodeError as je:
                st.error("👓 &nbsp;&nbsp; An error occurred reading the results. Is the document store working?")
                return
            except Exception as e:
                logging.exception(e)
                st.error("🐞 &nbsp;&nbsp; An error occurred during the request.")
                return

    if st.session_state.results:
        st.write("## Results:")

        alert_irrelevance=True
        if len(st.session_state.results['answers'])==0:
            st.info("🤔 &nbsp;&nbsp; Haystack is unsure whether any of the documents contain an answer to your question. Try to reformulate it!")

        for count, result in enumerate(st.session_state.results['answers']):
            result=result.to_dict()
            if result["answer"]:
                if alert_irrelevance and result['score']<0.50:
                    alert_irrelevance = False
                    st.write("""
                    <h4 style='color: darkred'>Attention, the 
                    following answers have low relevance:</h4>""",
                    unsafe_allow_html=True)

            answer, context = result["answer"], result["context"]
            start_idx = context.find(answer)
            end_idx = start_idx + len(answer)
            # Hack due to this bug: https://github.com/streamlit/streamlit/issues/3190
            st.write(markdown("- ..."+context[:start_idx] + str(annotation(answer, "ANSWER", "#3e1c21")) + context[end_idx:]+"..."), unsafe_allow_html=True)
            source = ""
            name = unquote(result['meta']['name']).replace('_',' ')
            url = result['meta']['url']
            source = f"[{name}]({url})"
            st.markdown(f"**Score:** {result['score']:.2f} -  **Source:** {source}")
main()