app.py

from dotenv import load_dotenv
import os
import streamlit as st
import openai
from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Document
from sentence_transformers import CrossEncoder
import fitz  # PyMuPDF library for PDF processing
import tempfile

load_dotenv()

openai.api_key = st.secrets['api_key']

# Create a sidebar
st.sidebar.title("Model Configuration")

# File uploader moved to the sidebar
uploaded_file = st.sidebar.file_uploader("Upload a PDF", type=["pdf"])

# Option menu for model selection
model_selection = st.sidebar.selectbox("Model Selection", ["GPT 3.5", "LLama 2"])

# Slider for selecting model temperature
model_temperature = st.sidebar.slider("Select model temperature", 0.0, 0.5, 1.0)

# Initialize LLM response storage
llm_responses = []

# Initialize HHEM model
hhem_model = CrossEncoder('vectara/hallucination_evaluation_model')

if uploaded_file is not None:
    # Save the uploaded PDF file to a temporary location
    with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as temp_pdf:
        temp_pdf.write(uploaded_file.read())
        temp_pdf_path = temp_pdf.name

    # Open the PDF file using PyMuPDF
    pdf_document = fitz.open(temp_pdf_path)

    text = ""
    for page_number in range(pdf_document.page_count):
        page = pdf_document[page_number]
        text += page.get_text()

    documents = [Document(text=text)]

    index = VectorStoreIndex.from_documents(documents)
    query_engine = index.as_query_engine()

    query = st.text_input("Ask your question")

    button = st.button("Ask")

    if button:
        print(query)
        response = query_engine.query(query)
        st.write(response.response)

        # Record LLM response
        llm_responses.append(response.response)

        # Calculate and display HHEM score for each LLM response
        for i, llm_response in enumerate(llm_responses):
            score = hhem_model.predict([text, llm_response])
            st.sidebar.write(f"Response {i + 1} - HHEM Score: {score}")

    # Close and remove the temporary PDF file
    pdf_document.close()
    os.remove(temp_pdf_path)

# Display LLM responses
if llm_responses:
    st.sidebar.markdown("## LLM Responses")
    for i, llm_response in enumerate(llm_responses):
        st.sidebar.write(f"Response {i + 1}: {llm_response}")