Magical blueprints for procedural generation of content. Based roughly on a series of articles by Sean Howard. Overview here.

• Introduction
• Fields and Generators
• Tags
• Mods
• Factories
• TODO
• HELP
• DEVELOPMENT
• Changelog

Blueprints are data objects. The essential idea is that you write subclasses of blueprint.Blueprint with fields that define the general parameters of their values (e.g. an integer between 0 and 10). When you instantiate a blueprint, you get a "mastered" blueprint with well-defined values for each field. Mastered blueprints may define special "generator" instance methods that build final objects from the master.

Think of it as prototypal inheritance for Python! (Yeah, I probably don't know what I'm talking about.)

Most of the big moving parts have their documentation, often with examples, in the docstring. Blueprint is best played with at the command line, trying out how things work. For the impatient, an example:

import blueprint


class Item(blueprint.Blueprint):
    value = 1
    tags = 'foo bar'

    class Meta:
        abstract = True


class Weapon(Item):
    name = 'Some Weapon'
    tags = 'dangerous equippable'
    damage = blueprint.RandomInt(1, 5)

    class Meta:
        abstract = True


class Spear(Weapon):
    tags = 'primitive piercing'
    name = 'Worn Spear'
    damage = blueprint.RandomInt(10, 15)
    value = blueprint.RandomInt(4, 6)


class PointedStick(Weapon):
    tags = 'primitive piercing'
    name = 'Pointed Stick'
    damage = 6
    value = 2


class Club(Weapon):
    tags = 'primitive crushing'
    name = 'Big Club'
    damage = blueprint.RandomInt(10, 15)
    value = 2


class Actor(blueprint.Blueprint):
    tags = 'active'


class CaveMan(Actor):
    name = 'Cave Man'
    weapon = blueprint.PickOne(
        Club, Spear, PointedStick
        )

And then:

>>> actor = CaveMan()
>>> actor
<CaveMan:
    name -- 'Cave Man'
    weapon -- <Spear:
        damage -- 5
        name -- 'Spear'
        value -- 6
        >
    >
>>> actor.weapon.name
'Spear'

Now, we can take our reified master data object and do something with it--use it as-is, or build another entity using the generated data.

Blueprints are data objects. By default, every member of a blueprint is treated as a field, either static or dynamic. Static fields are simple data attributes. Dynamic fields are callable objects that take one positional argument, the blueprint on which they are being called.

Dynamic fields make blueprints quite useful. A few basic fields are provided to get you started, and Blueprints themselves can be used as fields. Fields are designed to be nestable. They can rely upon each other too--use the blueprint.depends_on decorator to declare these dependencies.

If you really must have a callable method on your mastered blueprint, use the blueprint.generator decorator (or mark your callable object with the is_generator flag). These are called "generators" ("contractors" in squidi's terminology) because they're intended to be used to generate your final entity, whether it be a dict or a WAD file.

Blueprints automatically organize themselves using tags (domains in squidi's parlance). A direct descendant of Blueprint has its own tag repository (blueprint.taggables.TagRepository), which all its subclasses will share. So, in the above example, you can query Weapon.tag_repo.query(with_tags=('piercing')) and receive set([Spear, PointedStick]).

Blueprints are also automatically tagged by their class name (and their ancestor superclass names!), with camel-cased words separated out. So CaveMan will automatically get the tags set(['cave', 'man', 'actor']).

This makes the following possible:

class MammothHunter(CaveMan):
    weapon = blueprint.PickFrom(
        blueprint.WithTags('pointed weapon')
        )

Sometimes, you'll want to dynamically modify a blueprint. To do this, create a subclass of Mod. Mods are just special blueprints:

class OfDoom(blueprint.Mod):
    name = blueprint.FormatTemplate('{meta.source.name} of DOOM')
    value = lambda _: _.meta.source.value * 5

Then, apply it to another blueprint:

>>> club = OfDoom(Club)
>>> club.name
'Big Club of DOOM'

Mods always produce mastered blueprints.

Factories put all the pieces together--they're rather a blueprint factory. Say that you want an item drop that selects from a few common Weapon blueprints and adds a couple magical Mods to make it cooler. Here's our second mod:

class MagicalItemPrefix(blueprint.Mod):
    prefix = blueprint.PickOne(
        'Gnarled',
        'Inscribed',
        'Magnificent',
        )
    name = blueprint.depends_on('prefix')(
        blueprint.FormatTemplate('{parent.prefix} {meta.source.name}'))

Now, here's our Magical Item factory:

class MagicalItemFactory(blueprint.Factory):
    product = blueprint.PickFrom(
        blueprint.WithTags('weapon'))
    mods = [MagicalItemPrefix, OfDoom]

Now, when we call the factory, we get a random Weapon with magical properties:

>>> weapon = MagicalItemFactory()
>>> weapon.name
'Gnarled Worn Spear of DOOM'

Factories always produce mastered blueprints.

• Better documentation. :)
• Support all operators on blueprint.Field

If you run into trouble, or find a bug, file an issue in the tracker on github.

Itching to hack on blueprint? Fork the repository on on github and submit a pull request. If you're not sure what you're doing, follow these guidelines.

On github, bleeding-edge development works should be done on feature branches. master should always remain stable.

Blueprint uses uv for dependency management. To set up your development environment:

# Install dependencies (dev + test groups)
uv sync --group dev --group test

# Install pre-commit hooks
uv run pre-commit install

Tests are written using pytest and are located in the tests/ folder. Blueprint maintains 100% test coverage (including branch coverage) and validates test independence with randomized execution order.

To run the test suite:

# Run all tests with coverage
uv run pytest

# Run tests with randomized order (validates test independence)
uv run pytest --random-order

# Run the comprehensive test script (format, lint, type-check, test)
./runtests.sh

Blueprint enforces strict code quality standards:

• Type checking: All code is fully type-annotated and checked with mypy in strict mode
• Linting: Comprehensive linting with ruff (50+ rule groups enabled)
• Formatting: Consistent code formatting with ruff
• Documentation: Google-style docstrings with executable doctests

To check code quality:

# Run type checker
uv run mypy src tests

# Run linter
uv run ruff check .

# Auto-fix linting issues
uv run ruff check --fix .

# Run formatter
uv run ruff format .

# Run all pre-commit hooks manually
uv run pre-commit run --all-files

• 0.7: Major modernization release with comprehensive quality improvements:

  • Breaking changes:
    • Dropped Python 2.7 support: Minimum Python version is now 3.11+.
    • Method naming convention: Changed from camelCase to snake_case for internal methods (e.g., _getMaster() → _get_master()). Public API field classes (PickOne, PickFrom, etc.) remain unchanged.
    • Comparison methods: Replaced __cmp__ with total ordering (__lt__, __le__, __eq__, __ge__, __gt__) for Python 3 compatibility.
  • Quality improvements:
    • 100% test coverage: Achieved complete branch coverage with comprehensive test suite.
    • Full type annotations: All code is now fully type-annotated and validated with mypy in strict mode. Package includes py.typed marker for downstream type checking.
    • Comprehensive linting: Enforced via ruff with 50+ rule groups enabled, ensuring consistent code style and catching potential issues.
    • Doctest integration: All code examples in docstrings are now executable and validated during testing.
    • Test independence: Validated with pytest-random-order to ensure tests can run in any order without dependencies.
  • Infrastructure improvements:
    • Modern tooling: Migrated from setuptools to uv + hatchling for faster, more reliable builds.
    • Src layout: Adopted modern src/ layout for better packaging practices.
    • Pre-commit hooks: Added automated formatting, linting, and type checking before commits.
    • GitHub Actions CI: Automated testing on Python 3.11, 3.12, and 3.13.
    • Dynamic versioning: Improved version handling with uv-dynamic-versioning.
  • Bug fixes:
    • Fixed abstract blueprints being incorrectly included in tag repositories.
    • Removed test ordering dependencies that could cause flaky test failures.
    • Improved dice compilation and random number generation consistency.
  • Testing framework: Migrated from behave (BDD) to pytest with modern fixtures and parametrization.

• 0.6.1: Fixed Python 3 compatibility in dice roller.

• 0.6: Experimental Python 3 compatibility, and bug-fixes:

  • Feature: Experimental Python 3 compatibility, thanks to 0ion9.
  • Major bug fix: Fixed bug in dice compilation.

• 0.5: A couple new features, some interfaces and many bug-fixes:

  • Feature: Added Property descriptor which acts like a field. May not actually be useful.
  • Feature: Dice rolls now return a results list, which auto-sums when doing integer or floating point arithmetic. No more mandatory sum() in your dice expressions.
  • Major bug fix: Fixed bug where Dice fields did not use the correct random object, with nondeterministic results.
  • Bug fix/Interface change: Improved (though not yet perfect) field resolution mechanics. Fields that depend on other, deferred fields now have a fighting chance at resolving.
  • Bug fix/Interface change: DiceTable no longer accepts - or arbitrary numbers of . or : as a range separator. Only .. or : work now.
  • Interface change: Operators are now Fields in their own right, with all resulting rights and privileges.

• 0.4: Added a dice roller through blueprint.dice.roll, and a corresponding Dice and DiceTable fields. Blueprint subclasses now have a better __repr__ through the metaclass. METACLASSES ROCK.

  Modified the behavior of field resolution. All fields now use fields.resolve to consistently handle nested callables.

• 0.3.4: Learned how to read. Corrected Sean Howard's name in the intro copy. Three micro-releases in 1 hour!

• 0.3.3: Learned how to use distutils. :P (Fixed a unicode string in setup([packages=[...]]).)

• 0.3.2: Added the LICENSE file to the source distribution, so pip won't fail.

• 0.3.1: Radically improved docstrings, with relevant examples. Added a changelog!

• 0.3: Added Factories. Bugfixes.

• 0.2: Added Mods. Bugfixes.

• 0.1: Initial release.