pydanclick

Use Pydantic models as Click options.

Getting started

Install:

pip install pydanclick

Let's assume you have a Pydantic model:

class TrainingConfig(BaseModel):
    epochs: int
    lr: Annotated[float, Field(gt=0)] = 1e-4
    early_stopping: bool = False

Add all its fields as options in your Click command:

from pydanclick import from_pydantic

@click.command()
@from_pydantic(TrainingConfig)
def cli(training_config: TrainingConfig):
    # Here, we receive an already validated Pydantic object.
    click.echo(training_config.model_dump_json(indent=2))

~ python my_app.py --help
Usage: my_app.py [OPTIONS]

Options:
  --early-stopping / --no-early-stopping
  --lr FLOAT RANGE                [x>0]
  --epochs INTEGER                [required]
  --help                          Show this message and exit.

Keep reading:

• Take a tour of the Features below
• Read the Examples
• See the API Reference

Features

Use native Click types

The following types are converted to native Click types:

Pydantic type	Converted to
bool	click.BOOL
str	click.STRING
int	click.INT
float	click.FLOAT
Annotated[int, Field(lt=..., ge=...)	click.IntRange()
Annotated[float, Field(lt=..., ge=...)	click.FloatRange()
pathlib.Path	click.Path()
uuid.UUID	click.UUID
datetime.datetime, datetime.date	click.DateTime()
Literal	click.Choice

Complex container types such as lists or dicts are also supported: they must be passed as JSON strings, and will be validated through Pydantic TypeAdapter.validate_json method:

--arg1 '[1, 2, 3]' --arg2 '{"a": bool, "b": false}'

In any case, Pydantic validation will run during model instantiation.

Add multiple models

pydanclick.from_pydantic can be called several times with different models.

Use the prefix parameter to namespace the options from different models:

class Foo(BaseModel):
    a: str = ""
    b: str = ""

class Bar(BaseModel):
    x: int = 0
    y: int = 0

@click.command()
@from_pydantic(Foo, prefix="foo")
@from_pydantic(Bar, prefix="bar")
def cli(foo: Foo, bar: Bar):
    pass

will give:

~ python cli.py
Usage: cli.py [OPTIONS]

Options:
  --foo-a TEXT
  --foo-b TEXT
  --bar-x INTEGER
  --bar-y INTEGER
  --help           Show this message and exit.

Add regular options and arguments

pydanclick can be used alongside regular options and arguments:

@click.command()
@click.argument("arg")
@click.option("--option")
@from_pydantic(Foo)
def cli(arg, option, foo: Foo):
    pass

will give:

~ python cli.py
Usage: cli.py [OPTIONS] ARG

Options:
  --option TEXT
  --a TEXT
  --b TEXT
  --help         Show this message and exit.

Specify a custom variable name for the instantiated model with the same syntax as a regular Click option:

@click.command()
@from_pydantic("some_name", Foo)
def cli(some_name: Foo):
    pass

Document options

Options added with pydanclick.from_pydantic will appear in the command help page.

From docstrings: if griffe is installed, model docstring will be parsed and the Attributes section will be used to document options automatically (you can use pip install pydanclick[griffe] to install it). Use docstring_tyle to choose between google, numpy and sphinx coding style. Disable docstring parsing by passing parse_docstring=False.

From field description: pydanclick supports the Field(description=...) syntax from Pydantic. If specified, it will take precedence over the docstring description.

Explicitly: you can always specify a custom help string for a given field by using extra_options={"my_field": {"help": "my help string"}} where my_field is the name of your field.

Here are these three methods in action:

class Baz(BaseModel):
    """Some demo model.

    Attributes:
        a: this comes from the docstring (requires griffe)
    """

    a: int = 0
    b: Annotated[int, Field(description="this comes from the field description")] = 0
    c: int = 0


@click.command()
@from_pydantic(Baz, extra_options={"c": {"help": "this comes from the `extra_options`"}})
def cli(baz: Baz):
    pass

will give:

~ python cli.py --help
Usage: cli.py [OPTIONS]

Options:
  --a INTEGER  this comes from the docstring (requires griffe)
  --b INTEGER  this comes from the field description
  --c INTEGER  this comes from the `extra_options`
  --help       Show this message and exit.

Customize option names

Specify option names with rename and short option names with shorten:

@click.command()
@from_pydantic(Foo, rename={"a": "--alpha", "b": "--beta"}, shorten={"a": "-A", "b": "-B"})
def cli(foo: Foo):
    pass

will give:

~ python cli.py --help
Usage: cli.py [OPTIONS]

Options:
  -A, --alpha TEXT
  -B, --beta TEXT
  --help            Show this message and exit.

Note that prefix won't be prepended to option names passed with rename or shorten.

Pass extra parameters

Use extra_options to pass extra parameters to click.option for a given field.

For example, in the following code, the user will be prompted for the value of a:

@click.command()
@from_pydantic(Foo, extra_options={"a": {"prompt": True}})
def cli(foo: Foo):
    pass

Add nested models

Nested Pydantic models are supported, with arbitrary nesting level. Option names will be built by joining all parent names and the field names itself with dashes.

class Left(BaseModel):
    x: int

class Right(BaseModel):
    x: int

class Root(BaseModel):
    left: Left
    right: Right
    x: int

@click.command()
@from_pydantic(Root)
def cli(root: Root):
    pass

will give:

~ python cli.py --help
Usage: cli.py [OPTIONS]

Options:
  --left-x INTEGER   [required]
  --right-x INTEGER  [required]
  --x INTEGER        [required]
  --help             Show this message and exit.

To use rename, shorten, exclude, extra_options with a nested field, use its dotted name, e.g. left.x or right.x. Note that the alias of a field will apply to all its sub-fields:

@click.command()
@from_pydantic(Root, rename={"right": "--the-other-left"})
def cli(root: Root):
    pass

will give:

~ python cli.py --help
Usage: cli.py [OPTIONS]

Options:
  --left-x INTEGER            [required]
  --the-other-left-x INTEGER  [required]
  --x INTEGER                 [required]
  --help                      Show this message and exit.

Limitations

pydanclick doesn't support (yet!):

• Pydantic v1
• converting fields to arguments, instead of options
• fields annotated with union of Pydantic models can only be used with JSON inputs, instead of properly merging all sub-fields
• custom argument validators

Other missing features:

• Reading model from file
• Specifying all field-specific options directly in the Pydantic model (would allow easier re-use)
• Most Click features should be supported out-of-the-box through the extra_options parameter. However, most of them aren't tested
• Click and Pydantic both include validation logic. In particular, Click support custom ParamType, validation callbacks and BadParameter errors: it's not clear if we want to fully rely on Pydantic or on Click or on a mixture of both
• populating Pydantic fields from existing options or arguments (combined with exclude, it will provide a complete escape hatch to bypass Pydantclick when needed)
• attaching Pydanclick arguments directly to the model class, to avoid duplication when re-using a model in multiple commands