General options¤
allow_inspection¤
- Type
boolTrue
Whether to allow inspecting modules (importing them) when it is not possible to visit them (parse their source code).
When loading data for a given package, Griffe discovers every Python module, compiled or not, and inspects or visits them accordingly.
If you have compiled modules but also provide stubs for them, you might want to disable the inspection of these modules, because inspection picks up many more members, and sometimes the collected data is inaccurate (depending on the tool that was used to compile the module) or too low-level/technical for API documentation.
See also force_inspection.
Packages are loaded only once.
When mkdocstrings-python collects data from a Python package (thanks to Griffe), it collects the entire package and caches it. Next time an object from the same package is rendered, the package is retrieved from the cache and not collected again. The allow_inspection option will therefore only have an effect the first time a package is collected, and will do nothing for objects rendered afterwards.
plugins:
- mkdocstrings:
handlers:
python:
options:
allow_inspection: true
::: path.to.object
options:
allow_inspection: false
Preview
SomeClass
Docstring of the class.
__eq__
Method docstring.
__weakref__
Method docstring.
documented_method
Method docstring.
SomeClass
Docstring of the class.
documented_method
Method docstring.
backlinks¤
Sponsors only — Insiders 1.10.0
- Type
Literal["flat", "tree", False]False
The backlinks option enables rendering of backlinks within your API documentation.
When an arbitrary section of your documentation links to an API symbol, this link will be collected as a backlink, and rendered below your API symbol. In short, the API symbol will link back to the section that links to it. Such backlinks will help your users navigate the documentation, as they will immediately which functions return a specific symbol, or where a specific symbol is accepted as parameter, etc..
Each backlink is a list of breadcrumbs that represent the navigation, from the root page down to the given section.
The available styles for rendering backlinks are flat and tree.
flatwill render backlinks as a single-layer list. This can lead to repetition of breadcrumbs.treewill combine backlinks into a tree, to remove repetition of breadcrumbs.
Global-only option.
For now, the option only works when set globally in mkdocs.yml.
plugins:
- mkdocstrings:
handlers:
python:
options:
backlinks: tree
Preview
extensions¤
The extensions option lets you enable Griffe extensions, which enhance or modify the data collected from Python sources (or compiled modules).
Elements in the list can be strings or dictionaries.
Strings denote the path to an extension module, like griffe_typingdoc, or to an extension class directly, like griffe_typingdoc.TypingDocExtension. When using a module path, all extensions within that module will be loaded and enabled. Strings can also be the path to a Python module, and a class name separated with :, like scripts/griffe_extensions.py or scripts/griffe_extensions.py:MyExtension.
Dictionaries have a single key, which is the module/class path (as a dot-separated qualifier or file path and colon-separated class name, like above), and its value is another dictionary specifying options that will be passed when to class constructors when instantiating extensions.
Packages are loaded only once.
When mkdocstrings-python collects data from a Python package (thanks to Griffe), it collects the entire package and caches it. Next time an object from the same package is rendered, the package is retrieved from the cache and not collected again. Only the extensions specified the first time the package is loaded will be used. You cannot use a different set of extensions for specific objects rendered afterwards, and you cannot deactivate extensions for objects rendered afterwards either.
plugins:
- mkdocstrings:
handlers:
python:
options:
extensions:
- griffe_sphinx
- griffe_pydantic: {schema: true}
- scripts/exts.py:DynamicDocstrings:
paths: [mypkg.mymod.myobj]
::: your_package.your_module.your_func
options:
extensions:
- griffe_typingdoc
extra¤
- Type
dict{}
The extra option lets you inject additional variables into the Jinja context used when rendering templates. You can then use this extra context in your overridden templates.
Local extra options will be merged into the global extra option:
plugins:
- mkdocstrings:
handlers:
python:
options:
extra:
hello: world
::: your_package.your_module.your_func
options:
extra:
foo: bar
...will inject both hello and foo into the Jinja context when rendering your_package.your_module.your_func.
Warning
Previously, extra options were supported directly under the options key.
plugins:
- mkdocstrings:
handlers:
python:
options:
hello: world
Now that we introduced optional validation of options and automatic JSON schema generation thanks to Pydantic, we require extra options to be put under options.extra. Extra options directly under options are still supported, but deprecated, and will emit deprecation warnings. Support will be removed in a future version of mkdocstrings-python.
find_stubs_package¤
- Type
boolFalse
When looking for documentation specified in autodoc instructions (::: identifier), also look for the stubs package as defined in PEP 561 if it exists. This is useful when most of your documentation is separately provided by such a package and not inline in your main package.
Packages are loaded only once.
When mkdocstrings-python collects data from a Python package (thanks to Griffe), it collects the entire package and caches it. Next time an object from the same package is rendered, the package is retrieved from the cache and not collected again. The find_stubs_package option will therefore only have an effect the first time a package is collected, and will do nothing for objects rendered afterwards.
plugins:
- mkdocstrings:
handlers:
python:
options:
find_stubs_package: true
::: your_package.your_module.your_func
options:
find_stubs_package: true
def your_func(a, b):
# Function code
...
# rest of your code
def your_func(a: int, b: str):
"""
<Function docstring>
"""
...
# rest of your code
Preview
your_func
Function docstring
your_func
force_inspection¤
- Type
boolFalse
Whether to force inspecting modules (importing them) even if their source code is available.
This option is useful when you know that dynamic analysis (inspection) yields better results than static analysis. Do not use this blindly: the recommended approach is to write a Griffe extension that will improve extracted API data. See How to selectively inspect objects.
See also allow_inspection.
plugins:
- mkdocstrings:
handlers:
python:
options:
force_inspection: false
::: path.to.object
options:
force_inspection: true
Packages are loaded only once.
When mkdocstrings-python collects data from a Python package (thanks to Griffe), it collects the entire package and caches it. Next time an object from the same package is rendered, the package is retrieved from the cache and not collected again. The force_inspection option will therefore only have an effect the first time a package is collected, and will do nothing for objects rendered afterwards.
inheritance_diagram_direction¤
The direction of the Mermaid chart presenting the inheritance diagram of a class, TD by default.
extra_javascript:
- https://unpkg.com/mermaid@10.9.0/dist/mermaid.min.js
plugins:
- mkdocstrings:
handlers:
python:
options:
inheritance_diagram_direction: TD
::: path.to.object
options:
inheritance_diagram_direction: TD
Preview
With the following classes:
class SuperAbstract:
"""Super abstract class."""
class Mixin1:
"""Mixin 1."""
class Abstract(SuperAbstract, Mixin1):
"""Abstract class."""
class Mixin2A:
"""Mixin 2A."""
class Mixin2B(Mixin2A):
"""Mixin 2B."""
class Concrete(Abstract, Mixin2B):
"""Concrete class."""
class SuperConcrete(Concrete):
"""Super concrete class."""
flowchart TD
SuperConcrete[SuperConcrete]
Concrete[Concrete]
Abstract[Abstract]
SuperAbstract[SuperAbstract]
Mixin1[Mixin1]
Mixin2B[Mixin2B]
Mixin2A[Mixin2A]
Concrete --> SuperConcrete
Abstract --> Concrete
SuperAbstract --> Abstract
Mixin1 --> Abstract
Mixin2B --> Concrete
Mixin2A --> Mixin2B flowchart BT
SuperConcrete[SuperConcrete]
Concrete[Concrete]
Abstract[Abstract]
SuperAbstract[SuperAbstract]
Mixin1[Mixin1]
Mixin2B[Mixin2B]
Mixin2A[Mixin2A]
Concrete --> SuperConcrete
Abstract --> Concrete
SuperAbstract --> Abstract
Mixin1 --> Abstract
Mixin2B --> Concrete
Mixin2A --> Mixin2B flowchart RL
SuperConcrete[SuperConcrete]
Concrete[Concrete]
Abstract[Abstract]
SuperAbstract[SuperAbstract]
Mixin1[Mixin1]
Mixin2B[Mixin2B]
Mixin2A[Mixin2A]
Concrete --> SuperConcrete
Abstract --> Concrete
SuperAbstract --> Abstract
Mixin1 --> Abstract
Mixin2B --> Concrete
Mixin2A --> Mixin2B flowchart LR
SuperConcrete[SuperConcrete]
Concrete[Concrete]
Abstract[Abstract]
SuperAbstract[SuperAbstract]
Mixin1[Mixin1]
Mixin2B[Mixin2B]
Mixin2A[Mixin2A]
Concrete --> SuperConcrete
Abstract --> Concrete
SuperAbstract --> Abstract
Mixin1 --> Abstract
Mixin2B --> Concrete
Mixin2A --> Mixin2B preload_modules¤
Pre-load modules that are not specified directly in autodoc instructions (::: identifier). It is useful when you want to render documentation for a particular member of an object, and this member is imported from another package than its parent.
For an imported member to be rendered, you need to add it to the __all__ attribute of the importing module. The package from which the imported object originates must be accessible to the handler (see Finding modules).
plugins:
- mkdocstrings:
handlers:
python:
options:
preload_modules:
- their_package
::: your_package.your_module
options:
preload_modules:
- their_package
from their_package.their_module import their_object
__all__ = ["their_object"]
# rest of your code
Preview
your_module
Docstring of your module.
their_object
Docstring of their object.
your_module
Docstring of your module.
show_bases¤
- Type
boolTrue
Show the base classes of a class.
plugins:
- mkdocstrings:
handlers:
python:
options:
show_bases: true
::: path.to.object
options:
show_bases: false
Preview
SomeClass()
Docstring of the class.
show_inheritance_diagram¤
Sponsors only — Insiders 1.7.0
- Type
boolFalse
Show the inheritance diagram of a class using Mermaid.
With this option enabled, an inheritance diagram (as a flowchart) will be displayed after a class signature. Each node will act as a cross-reference and will bring you to the relevant class' documentation when clicking on it.
It should work out of the box with Material for MkDocs. For other themes, you must include Mermaid's Javascript code manually:
extra_javascript:
- https://unpkg.com/mermaid@10.9.0/dist/mermaid.min.js
plugins:
- mkdocstrings:
handlers:
python:
options:
show_inheritance_diagram: true
::: path.to.object
options:
show_inheritance_diagram: false
Preview
With the following classes:
class SuperAbstract:
"""Super abstract class."""
class Mixin1:
"""Mixin 1."""
class Abstract(SuperAbstract, Mixin1):
"""Abstract class."""
class Mixin2A:
"""Mixin 2A."""
class Mixin2B(Mixin2A):
"""Mixin 2B."""
class Concrete(Abstract, Mixin2B):
"""Concrete class."""
class SuperConcrete(Concrete):
"""Super concrete class."""
The diagram for SuperConcrete will look like this:
flowchart TD
SuperConcrete[SuperConcrete]
Concrete[Concrete]
Abstract[Abstract]
SuperAbstract[SuperAbstract]
Mixin1[Mixin1]
Mixin2B[Mixin2B]
Mixin2A[Mixin2A]
Concrete --> SuperConcrete
Abstract --> Concrete
SuperAbstract --> Abstract
Mixin1 --> Abstract
Mixin2B --> Concrete
Mixin2A --> Mixin2B Nodes are not clickable in this example because these classes do not exist in our documentation.
show_source¤
- Type
boolTrue
Show the source code of this object.
plugins:
- mkdocstrings:
handlers:
python:
options:
show_source: true
::: path.to.object
options:
show_source: false
Preview
some_function()
Docstring of the function.
Source code in package/module.py
1 2 | |
some_function()
Docstring of the function.
skip_local_inventory¤
- Type
boolFalse
Whether to skip registering symbols in the objects inventory.
With this option enabled, re-rendering docstrings for objects from external inventories is possible with their cross-references pointing to the original external inventory, not local. Similarly, it becomes possible to render the same symbol several times in the same documentation, with only one canonical location being used for cross-references (preventing confusion in mkdocs-autorefs).
plugins:
- mkdocstrings:
handlers:
python:
options:
skip_local_inventory: false
::: path.to.module
options:
skip_local_inventory: true
Preview
bisect_left builtin ¤
bisect_left(a, x, lo=0, hi=None, *, key=None)
- Usage Configuration options General options skip_local_inventory
Notice how bisect.bisect_left now points to the section above.
Notice how bisect.bisect_right points to the original Python documentation.