app.py

import gradio as gr
import sqlite3
import fitz # PyMuPDF
import re
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
import requests
import os
from collections import Counter
from datetime import datetime

# ==============================
# DATABASE SETUP
# ==============================

conn = sqlite3.connect("exam_trends.db", check_same_thread=False)
cursor = conn.cursor()

cursor.execute("""
CREATE TABLE IF NOT EXISTS past_questions (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    year INTEGER,
    question TEXT,
    topic TEXT
)
""")

cursor.execute("""
CREATE TABLE IF NOT EXISTS current_questions (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    question TEXT,
    topic TEXT
)
""")

conn.commit()


# ==============================
# PDF TEXT EXTRACTION
# ==============================

def extract_text_from_pdf(file):
    text = ""
    pdf = fitz.open(stream=file.read(), filetype="pdf")
    for page in pdf:
        text += page.get_text()
    return text


# ==============================
# SIMPLE QUESTION SEGMENTATION
# ==============================

def extract_questions(text):
    questions = re.split(r'\n\d+\.|\nQ\d+', text)
    return [q.strip() for q in questions if len(q.strip()) > 20]


# ==============================
# SIMPLE TOPIC CLASSIFIER
# (Basic keyword approach)
# ==============================

def classify_topic(question):
    question = question.lower()

    if "integral" in question or "derivative" in question:
        return "Calculus"
    elif "triangle" in question or "angle" in question:
        return "Geometry"
    elif "probability" in question:
        return "Probability"
    elif "matrix" in question:
        return "Linear Algebra"
    else:
        return "Algebra"


# ==============================
# STORE PAST PAPERS
# ==============================

def upload_past_papers(files):
    for file in files:
        text = extract_text_from_pdf(file)
        questions = extract_questions(text)

        year = datetime.now().year # Simplified

        for q in questions:
            topic = classify_topic(q)
            cursor.execute(
                "INSERT INTO past_questions (year, question, topic) VALUES (?, ?, ?)",
                (year, q, topic),
            )

    conn.commit()
    return "Past papers uploaded and processed successfully."


# ==============================
# STORE CURRENT PAPER
# ==============================

def upload_current_paper(file):
    text = extract_text_from_pdf(file)
    questions = extract_questions(text)

    for q in questions:
        topic = classify_topic(q)
        cursor.execute(
            "INSERT INTO current_questions (question, topic) VALUES (?, ?)",
            (q, topic),
        )

    conn.commit()
    return "Current paper uploaded successfully."


# ==============================
# TREND ANALYSIS ENGINE
# ==============================

def analyze_trends():
    cursor.execute("SELECT topic FROM past_questions")
    past_topics = [row[0] for row in cursor.fetchall()]

    cursor.execute("SELECT topic FROM current_questions")
    current_topics = [row[0] for row in cursor.fetchall()]

    past_count = Counter(past_topics)
    current_count = Counter(current_topics)

    results = []

    for topic, freq in past_count.items():
        frequency_weight = freq
        recency_weight = freq * 0.2
        gap_weight = 1 if topic not in current_count else 0

        prediction_score = (
            frequency_weight * 0.5
            + recency_weight * 0.3
            + gap_weight * 0.2
        )

        results.append((topic, prediction_score))

    results.sort(key=lambda x: x[1], reverse=True)

    df = pd.DataFrame(results, columns=["Topic", "Prediction Score"])
    return df


# ==============================
# HEATMAP VISUALIZATION
# ==============================

def generate_heatmap():
    cursor.execute("SELECT topic FROM past_questions")
    topics = [row[0] for row in cursor.fetchall()]
    count = Counter(topics)

    df = pd.DataFrame.from_dict(count, orient='index', columns=['Frequency'])
    df = df.sort_values(by='Frequency', ascending=False)

    plt.figure()
    plt.imshow(np.array(df['Frequency']).reshape(-1,1))
    plt.yticks(range(len(df.index)), df.index)
    plt.xticks([])
    plt.title("Topic Heatmap")
    plt.colorbar()
    plt.tight_layout()

    return plt


# ==============================
# AI QUESTION GENERATOR (Simple)
# ==============================

def generate_predicted_questions():
    df = analyze_trends()
    top_topics = df["Topic"].head(3).tolist()

    output = "Predicted High Probability Questions:\n\n"

    for topic in top_topics:
        output += f"• Create an exam-level question from {topic}\n"

    return output


# ==============================
# GRADIO UI
# ==============================

with gr.Blocks() as demo:
    gr.Markdown("# 🎓 AI Predictive Exam Intelligence System")

    with gr.Tab("📂 Upload Past 5 Years Papers"):
        past_files = gr.File(file_count="multiple")
        past_button = gr.Button("Upload & Analyze")
        past_output = gr.Textbox()

        past_button.click(upload_past_papers, inputs=past_files, outputs=past_output)

    with gr.Tab("📄 Upload Current Year Paper"):
        current_file = gr.File()
        current_button = gr.Button("Upload Current Paper")
        current_output = gr.Textbox()

        current_button.click(upload_current_paper, inputs=current_file, outputs=current_output)

    with gr.Tab("📊 Trend Analysis"):
        analyze_button = gr.Button("Run Trend Analysis")
        trend_output = gr.Dataframe()

        analyze_button.click(analyze_trends, outputs=trend_output)

    with gr.Tab("🔥 Topic Heatmap"):
        heatmap_button = gr.Button("Generate Heatmap")
        heatmap_output = gr.Plot()

        heatmap_button.click(generate_heatmap, outputs=heatmap_output)

    with gr.Tab("🔮 Predicted Questions"):
        predict_button = gr.Button("Generate Predictions")
        predict_output = gr.Textbox()

        predict_button.click(generate_predicted_questions, outputs=predict_output)


demo.launch()