QuizGenerator 0.25.0

Generate randomized quiz questions for Canvas LMS and PDF exams

QuizGenerator

Generate randomized quiz questions for Canvas LMS and PDF exams with support for multiple question types, automatic variation generation, and QR code-based answer keys.

Features

• Multiple Output Formats: Generate PDFs (LaTeX or Typst) and Canvas LMS quizzes
• Automatic Variations: Create unique versions for each student
• Extensible: Plugin system for custom question types
• Built-in Question Library: Memory management, process scheduling, calculus, linear algebra, and more
• QR Code Answer Keys: Regenerate exact exam versions from QR codes
• Canvas Integration: Direct upload to Canvas with variation support

Installation

pip install QuizGenerator

Reproducible installs (recommended)

If you want a fully pinned environment for a semester, use the lockfile:

uv sync --locked

We keep dependency ranges in pyproject.toml for flexibility and rely on uv.lock to pin exact versions when you need reproducible builds.

System Requirements

• Python 3.12+
• Typst (default PDF renderer)
• Optional: LaTeX distribution with latexmk (if using --latex)
• Recommended: Pandoc (for markdown conversion)
• Optional (LaTeX + QR codes): Inkscape for SVG conversion

Optional Dependencies

# For QR code grading support
pip install "QuizGenerator[grading]"

# For CST463 machine learning questions
pip install "QuizGenerator[cst463]"

Quick Start

Need a 2‑minute setup? See documentation/getting_started.md.

1. Create a quiz configuration (YAML)

# my_quiz.yaml
name: "Midterm Exam"

questions:
  10:  # 10-point questions
    "Process Scheduling":
      class: FIFOScheduling

  5:   # 5-point questions
    "Memory Paging":
      class: PagingQuestion

    "Vector Math":
      class: VectorAddition

You can also provide an ordered list of questions:

name: "Midterm Exam"
question_order: yaml
questions:
  - name: "Process Scheduling"
    points: 10
    class: FIFOScheduling
  - name: "Memory Paging"
    points: 5
    class: PagingQuestion

2. Generate PDFs

quizgen generate --yaml my_quiz.yaml --num-pdfs 3

PDFs will be created in the out/ directory.

3. Upload to Canvas

# Set up Canvas credentials in ~/.env first:
# CANVAS_API_URL=https://canvas.instructure.com
# CANVAS_API_KEY=your_api_key_here

quizgen \
  generate \
  --yaml my_quiz.yaml \
  --num-canvas 5 \
  --course-id 12345

4. Generate Tag-Filtered Practice Quizzes

Create one practice quiz assignment per matching registered question type:

quizgen \
  practice \
  cst334 memory \
  --practice-match any \
  --practice-tag-source merged \
  --practice-question-groups 5 \
  --practice-variations 5 \
  --course-id 12345

These are uploaded as regular graded quiz assignments into the practice assignment group, which is configured with 0.0 group weight. Tag filters accept either namespaced tags (for example course:cst334, topic:memory) or bare forms (cst334, memory). Use --practice-tag-source explicit if you want strict explicit-only tag matching.

5. Audit Tags

# Tag summary
quizgen tags list

# Show only question types missing explicit tags
quizgen tags list --only-missing-explicit --include-questions

# Explain tags for matching question types
quizgen tags explain sched

CLI Completion

quizgen --help
quizgen --install-completion
quizgen test 3 --test-question MLFQQuestion

The CLI supports shell completion (bash, zsh, fish, PowerShell) through Typer.

Creating Custom Questions

QuizGenerator supports two approaches for adding custom question types:

Option 1: Entry Points (Recommended for Distribution)

Create a pip-installable package:

# pyproject.toml
[project.entry-points."quizgenerator.questions"]
my_question = "my_package.questions:MyCustomQuestion"

After pip install, your questions are automatically available!

Option 2: Direct Import (Quick & Easy)

Add to your quiz YAML:

custom_modules:
  - my_questions  # Import my_questions.py

questions:
  10:
    "My Question":
      class: MyCustomQuestion

See documentation/custom_questions.md for complete guide.

Question Authoring Pattern (New)

All questions follow the same three‑method flow:

class MyQuestion(Question):
    @classmethod
    def _build_context(cls, *, rng_seed=None, **kwargs):
        context = super()._build_context(rng_seed=rng_seed, **kwargs)
        rng = context.rng
        context["value"] = rng.randint(1, 10)
        return context

    @classmethod
    def _build_body(cls, context):
        body = ca.Section()
        body.add_element(ca.Paragraph([f"Value: {context['value']}"]))
        body.add_element(ca.AnswerTypes.Int(context["value"], label="Value"))
        return body

    @classmethod
    def _build_explanation(cls, context):
        explanation = ca.Section()
        explanation.add_element(ca.Paragraph([f"Answer: {context['value']}"]))
        return explanation

Notes:

• Always use context.rng (or context["rng"]) for deterministic randomness.
• Avoid refresh(); it is no longer part of the API.

Built-in Question Types

Operating Systems (CST334)

• FIFOScheduling, SJFScheduling, RoundRobinScheduling
• PagingQuestion, TLBQuestion
• SemaphoreQuestion, MutexQuestion

Machine Learning / Math (CST463)

• VectorAddition, VectorDotProduct, VectorMagnitude
• MatrixAddition, MatrixMultiplication, MatrixTranspose
• DerivativeBasic, DerivativeChain
• GradientDescentStep

General

• FromText - Custom text questions
• FromGenerator - Programmatically generated questions (requires --allow-generator or QUIZGEN_ALLOW_GENERATOR=1)

Documentation

• Getting Started Guide
• First 5 Minutes
• Custom Questions Guide
• YAML Configuration Reference

Canvas Setup

1. Create a ~/.env file with your Canvas credentials:

# For testing/development
CANVAS_API_URL=https://canvas.test.instructure.com
CANVAS_API_KEY=your_test_api_key

# For production
CANVAS_API_URL_prod=https://canvas.instructure.com
CANVAS_API_KEY_prod=your_prod_api_key

2. Use --prod flag for production Canvas instance:

quizgen generate --prod --num-canvas 5 --course-id 12345 --yaml my_quiz.yaml

Advanced Features

Typst Support

Typst is the default for faster compilation. Use --latex to force LaTeX:

quizgen generate --latex --num-pdfs 3 --yaml my_quiz.yaml

Experimental: --typst-measurement uses Typst to measure question height for tighter layout. It can change pagination and ordering, so use with care on finalized exams.

Layout Optimization

By default, questions keep their YAML order (or point-value ordering for mapping format). Use --optimize-space to reorder questions to reduce PDF page count. This also affects Canvas order.

Deterministic Generation

Use seeds for reproducible quizzes:

quizgen generate --seed 42 --num-pdfs 3 --yaml my_quiz.yaml

Generation Controls

Limit backoff attempts for questions that retry until they are "interesting":

quizgen generate --yaml my_quiz.yaml --num-pdfs 1 --max-backoff-attempts 50

Set a default numeric tolerance for float answers (overridable per question):

quizgen generate --yaml my_quiz.yaml --num-pdfs 1 --float-tolerance 0.01

Per-answer override in custom questions:

ca.AnswerTypes.Float(value, label="Result", tolerance=0.005)

QR Code Regeneration

Each generated exam includes a QR code that stores:

• Question types and parameters
• Random seed
• Version information

Use the grading tools to scan QR codes and regenerate exact exam versions.

Security Considerations

FromGenerator Warning

The FromGenerator question type executes arbitrary Python code from your YAML configuration files. This is a powerful feature for creating dynamic questions, but it carries security risks:

• Only use FromGenerator with YAML files you completely trust
• Never run --allow-generator on YAML files from untrusted sources
• Be cautious when sharing question banks that contain generator code

FromGenerator is disabled by default. To enable it, use one of:

quizgen generate --allow-generator --yaml my_quiz.yaml
# or
QUIZGEN_ALLOW_GENERATOR=1 quizgen generate --yaml my_quiz.yaml

If you need dynamic question generation with untrusted inputs, consider writing a proper Question subclass instead, which provides better control and validation.

LaTeX -shell-escape Warning

When using --latex, QuizGenerator invokes latexmk -shell-escape to compile PDFs. This allows LaTeX to execute external commands (for example, via \write18). If your question content includes raw LaTeX (e.g., from custom question types or untrusted YAML sources), this can be a command‑execution vector.

Guidance:

• Only use --latex with trusted question sources.
• Prefer Typst (default) when possible.
• If you need LaTeX but want to reduce risk, avoid raw LaTeX content and keep custom questions constrained to ContentAST elements.

Local Release Helper (Recommended)

Install repository-managed git hooks and alias:

bash scripts/install_git_hooks.sh

This installs a pre-commit hook for repository hygiene checks and a local alias:

git bump patch

The hook reports Ruff issues but does not block commits on lint failures.

git bump bumps pyproject.toml via uv version, stages pyproject.toml and uv.lock, commits, tags, and pushes by default. Use git bump patch --no-tag --no-push for a local-only bump.

Project Structure

QuizGenerator/
├── QuizGenerator/           # Main package
│   ├── question.py         # Question base classes and registry
│   ├── quiz.py            # Quiz generation logic
│   ├── contentast.py      # Content AST for cross-format rendering
│   ├── premade_questions/ # Built-in question library
│   └── ...               # Question types and rendering utilities
├── example_files/        # Example quiz configurations
├── documentation/        # User guides
└── quizgen             # CLI entry point

Contributing

Contributions welcome! Areas of interest:

• New question types
• Additional LMS integrations
• Documentation improvements
• Bug fixes

License

GNU General Public License v3.0 or later (GPLv3+) - see LICENSE file for details

Citation

If you use QuizGenerator in academic work, please cite:

@software{quizgenerator,
  author = {Ogden, Sam},
  title = {QuizGenerator: Automated Quiz Generation for Education},
  year = {2024},
  url = {https://github.com/OtterDen-Lab/QuizGenerator}
}

Support

• Issues: https://github.com/OtterDen-Lab/QuizGenerator/issues
• Documentation: https://github.com/OtterDen-Lab/QuizGenerator/tree/main/documentation

---

Note: This tool is designed for educational use. Ensure compliance with your institution's academic integrity policies when using automated quiz generation.