generate_samples.py

"""
One-time script to generate sample PDF documents for the RAG chatbot demo.
Run: python generate_samples.py
Requires: reportlab (already in requirements.txt)
"""

import os
from reportlab.lib.pagesizes import LETTER
from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle
from reportlab.lib.units import inch
from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer
from reportlab.lib.enums import TA_LEFT, TA_CENTER


OUTPUT_DIR = os.path.join(os.path.dirname(__file__), "sample_docs")
os.makedirs(OUTPUT_DIR, exist_ok=True)


def _doc(filename: str, title: str, sections: list[tuple[str, str]]) -> None:
    """Build a PDF with a title and a list of (heading, body) sections."""
    path = os.path.join(OUTPUT_DIR, filename)
    doc = SimpleDocTemplate(
        path,
        pagesize=LETTER,
        rightMargin=inch,
        leftMargin=inch,
        topMargin=inch,
        bottomMargin=inch,
    )

    styles = getSampleStyleSheet()
    title_style = ParagraphStyle(
        "Title",
        parent=styles["Title"],
        fontSize=20,
        spaceAfter=20,
        alignment=TA_CENTER,
    )
    heading_style = ParagraphStyle(
        "Heading",
        parent=styles["Heading2"],
        fontSize=13,
        spaceAfter=6,
        spaceBefore=14,
        textColor="#1a1a2e",
    )
    body_style = ParagraphStyle(
        "Body",
        parent=styles["Normal"],
        fontSize=11,
        leading=16,
        spaceAfter=8,
        alignment=TA_LEFT,
    )

    story = [
        Paragraph(title, title_style),
        Spacer(1, 0.2 * inch),
    ]
    for heading, body in sections:
        story.append(Paragraph(heading, heading_style))
        story.append(Paragraph(body, body_style))
        story.append(Spacer(1, 0.1 * inch))

    doc.build(story)
    print(f"Created: {path}")


# ── ai_overview.pdf ──────────────────────────────────────────────────────────
_doc(
    "ai_overview.pdf",
    "Artificial Intelligence: An Overview",
    [
        (
            "What is Artificial Intelligence?",
            "Artificial Intelligence (AI) is the simulation of human intelligence processes by "
            "computer systems. These processes include learning (acquiring information and rules "
            "for using it), reasoning (using the rules to reach approximate or definite "
            "conclusions), and self-correction. AI enables machines to perform tasks that "
            "traditionally required human cognitive abilities such as visual perception, speech "
            "recognition, decision-making, and language translation.",
        ),
        (
            "History of AI: 1950s to Present",
            "The concept of AI was formally introduced by Alan Turing in 1950 with his "
            "landmark paper 'Computing Machinery and Intelligence,' where he proposed the "
            "Turing Test. The term 'Artificial Intelligence' was coined by John McCarthy at "
            "the 1956 Dartmouth Conference. The field experienced cycles of optimism and "
            "'AI winters' (periods of reduced funding and interest) through the 1970s and "
            "1980s. The 1990s brought expert systems and IBM Deep Blue defeating chess "
            "champion Garry Kasparov in 1997. The 2010s saw a renaissance driven by deep "
            "learning, massive datasets, and GPU computing power, culminating in breakthroughs "
            "like AlphaGo (2016) and large language models such as GPT (2018–present).",
        ),
        (
            "Types of AI: Narrow, General, and Super",
            "Narrow AI (Weak AI) is designed to perform a specific task — such as facial "
            "recognition, spam filtering, or playing chess — and cannot generalise beyond its "
            "training domain. This is the only type of AI that exists today. "
            "Artificial General Intelligence (AGI) refers to a hypothetical system that can "
            "perform any intellectual task a human can, with the ability to transfer knowledge "
            "across domains. AGI remains an open research challenge. "
            "Artificial Superintelligence (ASI) describes a theoretical AI that surpasses "
            "human cognitive performance in every domain. ASI is purely speculative and raises "
            "significant ethical and existential questions among researchers.",
        ),
        (
            "Real-World Applications of AI",
            "AI applications span virtually every industry. In healthcare, AI assists in "
            "early disease detection, drug discovery, and personalised treatment plans. In "
            "finance, AI powers algorithmic trading, fraud detection, and credit scoring. "
            "Autonomous vehicles use AI for perception and decision-making. Natural Language "
            "Processing (NLP) enables virtual assistants like Siri, Alexa, and ChatGPT. "
            "Recommendation engines at Netflix, Spotify, and Amazon leverage AI to personalise "
            "content. In manufacturing, AI-powered robots perform precision assembly and "
            "quality control. The global AI market was valued at over $200 billion in 2023 "
            "and is projected to exceed $1 trillion by 2030.",
        ),
    ],
)

# ── ml_concepts.pdf ──────────────────────────────────────────────────────────
_doc(
    "ml_concepts.pdf",
    "Machine Learning: Core Concepts",
    [
        (
            "What is Machine Learning?",
            "Machine Learning (ML) is a subset of AI that enables systems to learn and "
            "improve from experience without being explicitly programmed. Instead of writing "
            "rules by hand, a practitioner feeds data into an algorithm, which finds patterns "
            "and builds a model that can make predictions or decisions on new, unseen data. "
            "The key insight is that the system improves automatically as it is exposed to "
            "more data over time.",
        ),
        (
            "Supervised, Unsupervised, and Reinforcement Learning",
            "Supervised Learning trains a model on labelled examples — pairs of inputs and "
            "correct outputs — so it can predict the label for new inputs. Classic tasks "
            "include email spam detection (classification) and house price prediction "
            "(regression). "
            "Unsupervised Learning finds hidden structure in unlabelled data. Clustering "
            "algorithms group similar data points, while dimensionality reduction techniques "
            "like PCA compress data while retaining important features. "
            "Reinforcement Learning trains an agent to make sequential decisions by rewarding "
            "desirable actions and penalising undesirable ones. It has achieved superhuman "
            "performance in games like Go and Atari, and is used in robotics and "
            "recommendation systems.",
        ),
        (
            "Common Algorithms: Linear Regression, Decision Trees, Neural Networks",
            "Linear Regression is the simplest ML model: it fits a straight line through "
            "data points to predict a continuous output. Despite its simplicity, it remains "
            "highly interpretable and effective for many real-world problems. "
            "Decision Trees partition the feature space into regions by asking a series of "
            "yes/no questions at each node. They are easy to visualise and interpret. "
            "Ensemble methods such as Random Forests and Gradient Boosting combine hundreds "
            "of trees for higher accuracy. "
            "Neural Networks are composed of layers of interconnected nodes (neurons) that "
            "transform input data into progressively more abstract representations. Deep "
            "Neural Networks with many layers excel at image recognition, speech synthesis, "
            "and natural language understanding.",
        ),
        (
            "How Models Are Trained and Evaluated",
            "Training involves feeding labelled data to an algorithm and using an "
            "optimisation procedure — typically gradient descent — to minimise a loss function "
            "that measures prediction error. The dataset is split into training, validation, "
            "and test sets to detect overfitting. "
            "Key evaluation metrics include accuracy, precision, recall, F1-score "
            "(for classification), and mean squared error or R² (for regression). "
            "Cross-validation techniques such as k-fold CV provide robust performance "
            "estimates. Hyperparameter tuning — adjusting learning rate, depth, regularisation "
            "strength, etc. — is performed on the validation set. A model generalises well "
            "when it performs consistently on both validation and held-out test data.",
        ),
    ],
)

print("\nSample PDFs generated successfully in ./sample_docs/")