app.py

import streamlit as st
import pandas as pd
import numpy as np
import fitz  # PyMuPDF
import re
import json

def extract_pdf_text(pdf_path):
    """Extract text from a PDF file."""
    with fitz.open(pdf_path) as pdf_document:
        content_text = ""
        for page_num in range(len(pdf_document)):
            page = pdf_document.load_page(page_num)
            content_text += page.get_text() + "\n"
    return content_text

# Streamlit Application
st.title("PDF Data Extractor")

uploaded_file = st.file_uploader("Upload a PDF File", type="pdf")

if uploaded_file is not None:
    with open("temp.pdf", "wb") as f:
        f.write(uploaded_file.getbuffer())

    pdf_text = extract_pdf_text("temp.pdf")

    # Step 2: Extract relevant information from the text using regex
    pattern = r"2\s*3\s*4\s*5\s*\n-1,5\s*0([\s\S]*?)\n\nTrainer & Berater-Feedback"
    matches = re.findall(pattern, pdf_text)

    json_chunks = []
    for match in matches:
        match = match.replace(",", ".")
        values = [value.strip() for value in match.split("\n") if value.strip()]
        if len(values) == 22:
            json_chunks.append({"current": values})
        else:
            current = values[1::2]
            json_chunks.append({"current": current})

    json_output = json.dumps(json_chunks, indent=2)
    json_data = json.loads(json_output)

    # Define the original data structure
    original_data = {
        'Title': [
            "Personal Competence", "Personal Competence", "Personal Competence", "Personal Competence", "Personal Competence", "Personal Competence",
            "Personal Competence", "Personal Competence", "Personal Competence", "Personal Competence", "Personal Competence",
            "Personal Competence", "Personal Competence", "Business Competence", "Business Competence", "Business Competence", "Business Competence",
            "Business Competence", "Management Competence", "Management Competence", "Management Competence", "Management Competence",
        ],
        'Code': ["P1", "P2", "P3", "P4", "P5", "P6", "P7", "P8", "P9", "P10", "P11", "P12",
                 "P13", "B1", "B2", "B3", "B4", "B5", "M1", "M2", "M3", "M4"],
        'Dimensions': [
            "Integrity/ Reliability", "Appearance", "Enthusiasm/Passion", "Learning Motivation/ Self-Development", "Ability to Adapt/Flexibility",
            "Communication/Information", "Cooperation/ Team spirit", "Handling of Complex Situations", "Coolness/Handling of Unclear Situations",
            "Self-reliance/Initiative", "Conflict Management", "Ability to Assert Oneself/ Negotiation Skills", "Tact and Sensitivity",
            "Quality Orientation", "Client Orientation", "Specialized Knowledge", "Methodology/ Didactics/ Language", "Creativity/ Conceptional Skills",
            "Project Management", "Result Orientation", "Leadership Skills", "Coach and Advisor"
        ]
    }

    df = pd.DataFrame(original_data)

    # Add extracted scores to the DataFrame
    score_columns = ['Boss_score', 'Colleague_score', 'Colleague_other_score', 'Report_score', 'Customer_score']
    for idx, col in enumerate(score_columns):
        df[col] = json_data[idx]['current'] + [None] * (len(df) - len(json_data[idx]['current']))

    score_pattern = r"\d{1,2},\d{2}"
    code_pattern = r"[A-Z]\.[0-9]{1,2}"

    all_scores = re.findall(score_pattern, pdf_text)
    all_codes = re.findall(code_pattern, pdf_text)

    scores = [float(score.replace(",", ".")) for score in all_scores]
    codes = [code.strip() for code in all_codes]

    if len(codes) >= 44:
        codes = codes[22:44]
    if len(scores) >= 22:
        scores = scores[0:22]

    df1 = pd.DataFrame({'Code': [code.replace('.', '') for code in codes], 'All_raters_Score': scores})
    df_combined = pd.merge(df, df1, on="Code", how="inner")

    feature_cols = ['Boss_score', 'Colleague_score', 'Report_score', 'Customer_score', 'Colleague_other_score']
    df_combined[feature_cols] = df_combined[feature_cols].astype(float)

    def calculate_self_score(row):
        valid_features = [val for val in row[feature_cols] if pd.notna(val)]
        num_features = len(valid_features)
        if num_features > 1:
            sum_features = sum(valid_features) - row['All_raters_Score']
            return (row['All_raters_Score'] * num_features) - sum_features
        return np.nan

    df_combined['Self_score'] = df_combined.apply(calculate_self_score, axis=1)

     #Step 7 : Picking strengths and weaknesses
    # List of keywords/phrases to capture
    keywords = [
        'Integrity/ Reliability', 'Appearance', 'Enthusiasm/Passion',
        'Learning Motivation/ Self-Development', 'Ability to Adapt/Flexibility',
        'Communication/Information', 'Cooperation/ Team spirit',
        'Handling of Complex Situations', 'Coolness/Handling of Unclear Situations', 'Self-reliance/Initiative',
        'Conflict Management', 'Ability to Assert Oneself/ Negotiation Skills',
        'Tact and Sensitivity', 'Quality Orientation', 'Client Orientation',
        'Specialized Knowledge', 'Methodology/ Didactics/ Language',
        'Creativity/ Conceptional Skills', 'Project Management',
        'Result Orientation', 'Leadership Skills', 'Coach and Advisor'
    ]
    
    # Extract phrases between "Topics I would like to discuss... " and "Schedule for the follow-up meeting"
    phrases_pattern = r"Please use the form at the end of the section to finalize your development planning\.\s*(.*?)\s*Schedule for the follow-up meeting"
    phrases_matches = re.findall(phrases_pattern, pdf_text, re.DOTALL)
    
    # Extract the word after "The biggest strengths and room for improvements perceived by:"
    label_pattern = r"The biggest strengths and room for improvements perceived by:\s*(\w+)"
    labels = re.findall(label_pattern, pdf_text)
    
    # Process each match and extract only the required keywords
    json_output = []
    for i, phrases_text in enumerate(phrases_matches):
        extracted_phrases = [
            phrase for phrase in keywords if phrase in phrases_text
        ]
        if extracted_phrases:
            label = labels[i] if i < len(labels) else f"Phrases_{i+1}"
            json_output.append({label: extracted_phrases})
    
    # Convert to JSON format
    json_output_str = json.dumps(json_output, indent=2)
    
    # Print the JSON result
    #print(json_output_str)
    
    json_data = df.to_json(orient='records')
    
    data = []
    for item in json_output:
        for label, phrases in item.items():
            for phrase in phrases:
                data.append({'Label': label, 'Dimensions': phrase})
    
    df4 = pd.DataFrame(data)

    #Step 9: Converting Streangths and Weaknesses with scores into json

    # Filter dataframes based on 'Label' value
    boss, direct, colleague, other_colleague = [df4[df4['Label'] == label].copy() for label in ['Boss', 'Direct', 'Colleagues', 'Colleague (o']]
    
    # Create mapping dictionaries from df3
    mappings = {
        'Boss_score': df_combined.set_index('Dimensions')['Boss_score'].to_dict(),
        'Report_score': df_combined.set_index('Dimensions')['Report_score'].to_dict(),
        'Colleague_score': df_combined.set_index('Dimensions')['Colleague_score'].to_dict(),
        'Other_colleague_score': df_combined.set_index('Dimensions')['Colleague_other_score'].to_dict()
    }
    
    # Map the values from df3 to the appropriate DataFrames
    boss['Boss_score'] = boss['Dimensions'].map(mappings['Boss_score'])
    direct['Report_score'] = direct['Dimensions'].map(mappings['Report_score'])
    colleague['Colleague_score'] = colleague['Dimensions'].map(mappings['Colleague_score'])
    other_colleague['Other_colleague_score'] = other_colleague['Dimensions'].map(mappings['Other_colleague_score'])

    boss.sort_values(by = 'Boss_score', ascending = False)
    boss_json = boss.iloc[3:,1:].to_dict(orient='records')
    prompt = f"You are a corporate trainer who guides me how to behave in corporate settings. You will analyze the top 3 strengths and scores{boss_json} rated by my boss out of 6. You will generate a nudge for each dimension to improveupon these strengths." 
        
    st.write("### Output:")
    st.write("#### 1. Extracted data: Dimensions Assessment of the leader by: Boss, Colleagues, Colleagues (other b.), Direct reports, Customers and All Raters")
    st.write("#### 2. Derived data: Self score")
    st.write("### Dataset Table")
    st.dataframe(df_combined)
    st.write(f"#### {print(prompt)}")