app.py

import streamlit as st
from PyPDF2 import PdfReader
from anthropic import Anthropic
from prompts import DIFFERENTIATE_PROMPT

def extract_differences(input_text):
        input_text = input_text.strip()
        qa_pairs = input_text.split('</difference>')
        
        # Initialize an empty list to hold the parsed dictionary objects
        parsed_data = []
        
        # Iterate over each QA pair
        for pair in qa_pairs:
            # Check if the pair has both question and answer (ignoring the last one)
            if '<text1_section>' in pair and '</text1_section>' in pair and '<text2_section>' in pair and '</text2_section>' in pair and '<explanation>' in pair and '</explanation>' in pair:
                # Extract the question and answer text
                text1 = pair.split('<text1_section>')[1].split('</text1_section>')[0]
                text2 = pair.split('<text2_section>')[1].split('</text2_section>')[0]
                explanation = pair.split('<explanation>')[1].split('</explanation>')[0]
                
                # Create a dictionary for the current pair and append it to the list
                parsed_data.append({'text1': text1.strip(), 'text2': text2.strip(), 'explanation': explanation.strip()})
        
        return parsed_data

def make_llm_api_call(prompt):
    client = Anthropic()
    message = client.messages.create(
        model="claude-3-haiku-20240307",
        max_tokens=4096,
        temperature=0,
        messages=[{"role": "user", "content": [{"type": "text", "text": prompt}]}],
    )
    return message

def get_llm_response(extractedtext1, extractedtext2):
    prompt = DIFFERENTIATE_PROMPT.format(text1=extractedtext1, text2=extractedtext2)
    
    message = make_llm_api_call(prompt)
    
    message_text = message.content[0].text

    try:
        try:
            differences_list = extract_differences(message_text.split("<differences>")[1].split("</differences>")[0].strip())
        except Exception as e:
            differences_list = extract_differences(message_text)
    except Exception as e:
        print("Error:", e)
        return message_text, []
    
    # display_text = f"{before_differences}\n\n{difference_content}\n\n{after_differences}"
    return differences_list



def extract_text_with_pypdf(pdf_path):
    reader = PdfReader(pdf_path)
    text = []
    for page in reader.pages:
        text.append(page.extract_text())
    return text


def main():
    st.set_page_config(layout="wide")  # Enable wide layout
    if "differences_data" not in  st.session_state:
        st.session_state.differences_data = []
    if "display_data" not in st.session_state:
        st.session_state.display_data = {"file1": None, "file2": None, "i": 0}
    if "file1" not in st.session_state:
        st.session_state.file1 = None
    if "file2" not in st.session_state:
        st.session_state.file2 = None
    if "extracted_texts" not in st.session_state:
        st.session_state.extracted_texts = {"file1": None, "file2": None, "extracted_text_1": [], "extracted_text_2": []}
    st.markdown('<div style="text-align: center;">' + '<h1>PDF Upload and Compare App</h1>' + '</div>', unsafe_allow_html=True)
    
    # Create columns for side-by-side buttons
    col1, col2 = st.columns([2, 2])
        
    # Add upload button to left column
    st.session_state.file1 = col1.file_uploader("**PDF 1**", type="pdf")
    
    # Add upload button to right column
    st.session_state.file2 = col2.file_uploader("**PDF 2**", type="pdf")

    
    # Check if both files are uploaded
    if st.session_state.file1 and st.session_state.file2:
        # Get filenames from uploaded files
        filename1 = st.session_state.file1.name
        filename2 = st.session_state.file2.name

        with st.spinner("Extracting text from PDFs"):
            if st.session_state.display_data["file1"] != st.session_state.file1 or st.session_state.display_data["file2"] != st.session_state.file2:
                st.session_state.display_data = {"file1": st.session_state.file1, "file2": st.session_state.file2, "i": 0}
                st.session_state.extracted_texts = {"file1": st.session_state.display_data["file1"], "file2": st.session_state.display_data["file2"], "extracted_text_1": extract_text_with_pypdf(st.session_state.file1), "extracted_text_2": extract_text_with_pypdf(st.session_state.file2)}
        
        try:
            extracted_text1 = st.session_state.extracted_texts["extracted_text_1"]
            extracted_text2 = st.session_state.extracted_texts["extracted_text_2"]
            with col1.expander(filename1):
                st.write("\n\n".join("\n\n------------------------\n\n".join(extracted_text1).splitlines()))
            
            with col2.expander(filename2):
                st.write("\n\n".join("\n\n------------------------\n\n".join(extracted_text2).splitlines()))
            
            st.success(f"PDF text extraction complete")
        except Exception as e:
            st.error(f"Error saving files: {str(e)}")

        with st.spinner("Processing Pages within the PDFS"):
            try:
                # display_text, parsed_data = process_concurrently(extracted_text1, extracted_text2)
                # display_text, parsed_data = get_llm_response(extracted_text1, extracted_text2)
                
                for i,(etext1, etext2) in enumerate(zip(extracted_text1, extracted_text2)):
                    if i >= st.session_state.display_data["i"]: break
                    data = next((d for d in st.session_state.differences_data if d['etext1'] == etext1 and d['etext2'] == etext2), None)
                    if data:
                        pdata = data['pdata']
                        dext1 = data['dext1']
                        dext2 = data['dext2']
                    else:
                        pdata = get_llm_response(etext1, etext2)
                        dext1 = etext1
                        dext2 = etext2
    
                        for diff in pdata:
                            diff_text1 = diff['text1'].strip()
                            diff_text2 = diff['text2'].strip()          
                            if diff_text1 == "" or diff_text2 == "": continue              
                            diff_text1_phrase = "\n".join([f"<span style='background-color: grey;'>{t}</span>" for t in diff_text1.splitlines()])
                            diff_text2_phrase = "\n".join([f"<span style='background-color: grey;'>{t}</span>" for t in diff_text2.splitlines()])
                            dext1 = diff_text1_phrase.join(dext1.split(diff_text1)) if diff_text1 in dext1 else dext1
                            dext2 = diff_text2_phrase.join(dext2.split(diff_text2)) if diff_text2 in dext2 else dext2
    
                        st.session_state.differences_data.append({"etext1": etext1, "etext2": etext2, "pdata": pdata, "dext1": dext1, "dext2": dext2})
                        reverse_pdata = [{'text1': d['text2'], 'text2': d['text1'], 'explanation': d['explanation']} for d in pdata]
                        st.session_state.differences_data.append({"etext1": etext2, "etext2": etext1, "pdata": reverse_pdata, "dext1": dext2, "dext2": dext1})
                    
                    display_text = "\n\n\n".join([f"**PDF 1:**\n\n{d['text1']}\n\n**PDF 2:**\n\n{d['text2']}\n\n**Explanation:**\n\n{d['explanation']}\n\n----------------------\n" for d in pdata])
    
                    with st.expander(f"**Page {i+1}** - {filename1}"):
                        st.markdown("\n\n".join(dext1.splitlines()), unsafe_allow_html=True)
                    with st.expander(f"**Page {i+1}** - {filename2}"):
                        st.markdown("\n\n".join(dext2.splitlines()), unsafe_allow_html=True)
                    st.markdown(display_text)
            except Exception as e:
                st.error(f"Error finding differences: {str(e)}")


        # Add button at the bottom to run Find Differences function
        if st.button("Find Differences"):
            st.session_state.display_data["i"] = st.session_state.display_data["i"] + 5
            st.rerun()


if __name__ == "__main__":
    main()