MkDocs Plugins
A Guide to installing, using and creating MkDocs Plugins
Installing Plugins
Before a plugin can be used, it must be installed on the system. If you are
using a plugin which comes with MkDocs, then it was installed when you installed
MkDocs. However, to install third party plugins, you need to determine the
appropriate package name and install it using pip
:
pip install mkdocs-foo-plugin
Warning
Installing an MkDocs plugin means installing a Python package and executing any code that the author has put in there. So, exercise the usual caution; there's no attempt at sandboxing.
Once a plugin has been successfully installed, it is ready to use. It just needs to be enabled in the configuration file. The Catalog repository has a large ranked list of plugins that you can install and use.
Using Plugins
The plugins
configuration option should contain a list of plugins to
use when building the site. Each "plugin" must be a string name assigned to the
plugin (see the documentation for a given plugin to determine its "name"). A
plugin listed here must already be installed.
plugins:
- search
Some plugins may provide configuration options of their own. If you would like
to set any configuration options, then you can nest a key/value mapping
(option_name: option value
) of any options that a given plugin supports. Note
that a colon (:
) must follow the plugin name and then on a new line the option
name and value must be indented and separated by a colon. If you would like to
define multiple options for a single plugin, each option must be defined on a
separate line.
plugins:
- search:
lang: en
foo: bar
For information regarding the configuration options available for a given plugin, see that plugin's documentation.
For a list of default plugins and how to override them, see the configuration documentation.
Developing Plugins
Like MkDocs, plugins must be written in Python. It is generally expected that each plugin would be distributed as a separate Python module, although it is possible to define multiple plugins in the same module. At a minimum, a MkDocs Plugin must consist of a BasePlugin subclass and an entry point which points to it.
BasePlugin
A subclass of mkdocs.plugins.BasePlugin
should define the behavior of the plugin.
The class generally consists of actions to perform on specific events in the build
process as well as a configuration scheme for the plugin.
All BasePlugin
subclasses contain the following attributes:
config_scheme
A tuple of configuration validation instances. Each item must consist of a
two item tuple in which the first item is the string name of the
configuration option and the second item is an instance of
mkdocs.config.config_options.BaseConfigOption
or any of its subclasses.
For example, the following config_scheme
defines three configuration options: foo
, which accepts a string; bar
, which accepts an integer; and baz
, which accepts a boolean value.
class MyPlugin(mkdocs.plugins.BasePlugin):
config_scheme = (
('foo', mkdocs.config.config_options.Type(str, default='a default value')),
('bar', mkdocs.config.config_options.Type(int, default=0)),
('baz', mkdocs.config.config_options.Type(bool, default=True))
)
New in version 1.4
Subclassing Config
to specify the config schema
To get type safety benefits, if you're targeting only MkDocs 1.4+, define the config schema as a class instead:
class MyPluginConfig(mkdocs.config.base.Config):
foo = mkdocs.config.config_options.Type(str, default='a default value')
bar = mkdocs.config.config_options.Type(int, default=0)
baz = mkdocs.config.config_options.Type(bool, default=True)
class MyPlugin(mkdocs.plugins.BasePlugin[MyPluginConfig]):
...
Examples of config definitions
! EXAMPLE:
from mkdocs.config import base, config_options as c class _ValidationOptions(base.Config): enabled = c.Type(bool, default=True) verbose = c.Type(bool, default=False) skip_checks = c.ListOfItems(c.Choice(('foo', 'bar', 'baz')), default=[]) class MyPluginConfig(base.Config): definition_file = c.File(exists=True) # required checksum_file = c.Optional(c.File(exists=True)) # can be None but must exist if specified validation = c.SubConfig(_ValidationOptions)
From the user's point of view
SubConfig
is similar toType(dict)
, it's just that it also retains full ability for validation: you define all valid keys and what each value should adhere to.And
ListOfItems
is similar toType(list)
, but again, we define the constraint that each value must adhere to.This accepts a config as follows:
my_plugin: definition_file: configs/test.ini # relative to mkdocs.yml validation: enabled: !ENV [CI, false] verbose: true skip_checks: - foo - baz
? EXAMPLE:
import numbers from mkdocs.config import base, config_options as c class _Rectangle(base.Config): width = c.Type(numbers.Real) # required height = c.Type(numbers.Real) # required class MyPluginConfig(base.Config): add_rectangles = c.ListOfItems(c.SubConfig(_Rectangle)) # required
In this example we define a list of complex items, and that's achieved by passing a concrete
SubConfig
toListOfItems
.This accepts a config as follows:
my_plugin: add_rectangles: - width: 5 height: 7 - width: 12 height: 2
When the user's configuration is loaded, the above scheme will be used to
validate the configuration and fill in any defaults for settings not
provided by the user. The validation classes may be any of the classes
provided in mkdocs.config.config_options
or a third party subclass defined
in the plugin.
Any settings provided by the user which fail validation or are not defined
in the config_scheme
will raise a mkdocs.config.base.ValidationError
.
config
A dictionary of configuration options for the plugin, which is populated by
the load_config
method after configuration validation has completed. Use
this attribute to access options provided by the user.
def on_pre_build(self, config, **kwargs):
if self.config['baz']:
# implement "baz" functionality here...
New in version 1.4
Safe attribute-based access
To get type safety benefits, if you're targeting only MkDocs 1.4+, access options as attributes instead:
def on_pre_build(self, config, **kwargs):
if self.config.baz:
print(self.config.bar ** 2) # OK, `int ** 2` is valid.
All BasePlugin
subclasses contain the following method(s):
load_config(options)
Loads configuration from a dictionary of options. Returns a tuple of
(errors, warnings)
. This method is called by MkDocs during configuration
validation and should not need to be called by the plugin.
on_<event_name>()
Optional methods which define the behavior for specific events. The plugin
should define its behavior within these methods. Replace <event_name>
with
the actual name of the event. For example, the pre_build
event would be
defined in the on_pre_build
method.
Most events accept one positional argument and various keyword arguments. It
is generally expected that the positional argument would be modified (or
replaced) by the plugin and returned. If nothing is returned (the method
returns None
), then the original, unmodified object is used. The keyword
arguments are simply provided to give context and/or supply data which may
be used to determine how the positional argument should be modified. It is
good practice to accept keyword arguments as **kwargs
. In the event that
additional keywords are provided to an event in a future version of MkDocs,
there will be no need to alter your plugin.
For example, the following event would add an additional static_template to the theme config:
class MyPlugin(BasePlugin):
def on_config(self, config, **kwargs):
config['theme'].static_templates.add('my_template.html')
return config
New in version 1.4
To get type safety benefits, if you're targeting only MkDocs 1.4+, access config options as attributes instead:
def on_config(self, config: MkDocsConfig):
config.theme.static_templates.add('my_template.html')
return config
Events
There are three kinds of events: Global Events, Page Events and Template Events.
See a diagram with relations between all the plugin events
- The events themselves are shown in yellow, with their parameters.
- Arrows show the flow of arguments and outputs of each event. Sometimes they're omitted.
- The events are chronologically ordered from top to bottom.
- Dotted lines appear at splits from global events to per-page events.
- Click the events' titles to jump to their description.
One-time Events
One-time events run once per mkdocs
invocation. The only case where these tangibly differ from global events is for mkdocs serve
: global events, unlike these, will run multiple times -- once per build.
on_startup
The startup
event runs once at the very beginning of an mkdocs
invocation.
New in MkDocs 1.4.
The presence of an on_startup
method (even if empty) migrates the plugin to the new
system where the plugin object is kept across builds within one mkdocs serve
.
Note that for initializing variables, the __init__
method is still preferred.
For initializing per-build variables (and whenever in doubt), use the on_config
event.
Parameters:
-
command
(Literal['build', 'gh-deploy', 'serve']
) –the command that MkDocs was invoked with, e.g. "serve" for
mkdocs serve
. -
dirty
(bool
) –whether
--dirty
flag was passed.
on_shutdown
The shutdown
event runs once at the very end of an mkdocs
invocation, before exiting.
This event is relevant only for support of mkdocs serve
, otherwise within a
single build it's undistinguishable from on_post_build
.
New in MkDocs 1.4.
The presence of an on_shutdown
method (even if empty) migrates the plugin to the new
system where the plugin object is kept across builds within one mkdocs serve
.
Note the on_post_build
method is still preferred for cleanups, when possible, as it has
a much higher chance of actually triggering. on_shutdown
is "best effort" because it
relies on detecting a graceful shutdown of MkDocs.
on_serve
The serve
event is only called when the serve
command is used during
development. It runs only once, after the first build finishes.
It is passed the Server
instance which can be modified before
it is activated. For example, additional files or directories could be added
to the list of "watched" files for auto-reloading.
Parameters:
-
server
(LiveReloadServer
) –livereload.Server
instance -
config
(MkDocsConfig
) –global configuration object
-
builder
(Callable
) –a callable which gets passed to each call to
server.watch
Returns:
-
LiveReloadServer | None
–livereload.Server
instance
Global Events
Global events are called once per build at either the beginning or end of the build process. Any changes made in these events will have a global effect on the entire site.
on_config
The config
event is the first event called on build and is run immediately
after the user configuration is loaded and validated. Any alterations to the
config should be made here.
Parameters:
-
config
(MkDocsConfig
) –global configuration object
Returns:
-
MkDocsConfig | None
–global configuration object
on_pre_build
The pre_build
event does not alter any variables. Use this event to call
pre-build scripts.
Parameters:
-
config
(MkDocsConfig
) –global configuration object
on_files
The files
event is called after the files collection is populated from the
docs_dir
. Use this event to add, remove, or alter files in the
collection. Note that Page objects have not yet been associated with the
file objects in the collection. Use Page Events to manipulate page
specific data.
Parameters:
-
files
(Files
) –global files collection
-
config
(MkDocsConfig
) –global configuration object
Returns:
-
Files | None
–global files collection
on_nav
The nav
event is called after the site navigation is created and can
be used to alter the site navigation.
Parameters:
-
nav
(Navigation
) –global navigation object
-
config
(MkDocsConfig
) –global configuration object
-
files
(Files
) –global files collection
Returns:
-
Navigation | None
–global navigation object
on_env
The env
event is called after the Jinja template environment is created
and can be used to alter the
Jinja environment.
Parameters:
-
env
(Environment
) –global Jinja environment
-
config
(MkDocsConfig
) –global configuration object
-
files
(Files
) –global files collection
Returns:
-
Environment | None
–global Jinja Environment
on_post_build
The post_build
event does not alter any variables. Use this event to call
post-build scripts.
Parameters:
-
config
(MkDocsConfig
) –global configuration object
on_build_error
The build_error
event is called after an exception of any kind
is caught by MkDocs during the build process.
Use this event to clean things up before MkDocs terminates. Note that any other
events which were scheduled to run after the error will have been skipped. See
Handling Errors for more details.
Parameters:
-
error
(Exception
) –exception raised
Template Events
Template events are called once for each non-page template. Each template event will be called for each template defined in the extra_templates config setting as well as any static_templates defined in the theme. All template events are called after the env event and before any page events.
on_pre_template
The pre_template
event is called immediately after the subject template is
loaded and can be used to alter the template.
Parameters:
-
template
(Template
) –a Jinja2 Template object
-
template_name
(str
) –string filename of template
-
config
(MkDocsConfig
) –global configuration object
Returns:
-
Template | None
–a Jinja2 Template object
on_template_context
The template_context
event is called immediately after the context is created
for the subject template and can be used to alter the context for that specific
template only.
Parameters:
-
context
(TemplateContext
) –dict of template context variables
-
template_name
(str
) –string filename of template
-
config
(MkDocsConfig
) –global configuration object
Returns:
-
TemplateContext | None
–dict of template context variables
on_post_template
The post_template
event is called after the template is rendered, but before
it is written to disc and can be used to alter the output of the template.
If an empty string is returned, the template is skipped and nothing is is
written to disc.
Parameters:
-
output_content
(str
) –output of rendered template as string
-
template_name
(str
) –string filename of template
-
config
(MkDocsConfig
) –global configuration object
Returns:
-
str | None
–output of rendered template as string
Page Events
Page events are called once for each Markdown page included in the site. All page events are called after the post_template event and before the post_build event.
on_pre_page
The pre_page
event is called before any actions are taken on the subject
page and can be used to alter the Page
instance.
Parameters:
-
page
(Page
) –mkdocs.structure.pages.Page
instance -
config
(MkDocsConfig
) –global configuration object
-
files
(Files
) –global files collection
Returns:
-
Page | None
–mkdocs.structure.pages.Page
instance
on_page_read_source
Deprecated
Instead of this event, prefer one of these alternatives:
- Since MkDocs 1.6, instead set
content_bytes
/content_string
of aFile
insideon_files
. - Usually (although it's not an exact alternative),
on_page_markdown
can serve the same purpose.
The on_page_read_source
event can replace the default mechanism to read
the contents of a page's source from the filesystem.
Parameters:
-
page
(Page
) –mkdocs.structure.pages.Page
instance -
config
(MkDocsConfig
) –global configuration object
Returns:
-
str | None
–The raw source for a page as unicode string. If
None
is returned, the default loading from a file will be performed.
on_page_markdown
The page_markdown
event is called after the page's markdown is loaded
from file and can be used to alter the Markdown source text. The meta-
data has been stripped off and is available as page.meta
at this point.
Parameters:
-
markdown
(str
) –Markdown source text of page as string
-
page
(Page
) –mkdocs.structure.pages.Page
instance -
config
(MkDocsConfig
) –global configuration object
-
files
(Files
) –global files collection
Returns:
-
str | None
–Markdown source text of page as string
on_page_content
The page_content
event is called after the Markdown text is rendered to
HTML (but before being passed to a template) and can be used to alter the
HTML body of the page.
Parameters:
-
html
(str
) –HTML rendered from Markdown source as string
-
page
(Page
) –mkdocs.structure.pages.Page
instance -
config
(MkDocsConfig
) –global configuration object
-
files
(Files
) –global files collection
Returns:
-
str | None
–HTML rendered from Markdown source as string
on_page_context
The page_context
event is called after the context for a page is created
and can be used to alter the context for that specific page only.
Parameters:
-
context
(TemplateContext
) –dict of template context variables
-
page
(Page
) –mkdocs.structure.pages.Page
instance -
config
(MkDocsConfig
) –global configuration object
-
nav
(Navigation
) –global navigation object
Returns:
-
TemplateContext | None
–dict of template context variables
on_post_page
The post_page
event is called after the template is rendered, but
before it is written to disc and can be used to alter the output of the
page. If an empty string is returned, the page is skipped and nothing is
written to disc.
Parameters:
-
output
(str
) –output of rendered template as string
-
page
(Page
) –mkdocs.structure.pages.Page
instance -
config
(MkDocsConfig
) –global configuration object
Returns:
-
str | None
–output of rendered template as string
Event Priorities
For each event type, corresponding methods of plugins are called in the order that the plugins appear in the plugins
config.
Since MkDocs 1.4, plugins can choose to set a priority value for their events. Events with higher priority are called first. Events without a chosen priority get a default of 0. Events that have the same priority are ordered as they appear in the config.
mkdocs.plugins.event_priority(priority: float) -> Callable[[T], T]
A decorator to set an event priority for an event handler method.
Recommended priority values:
100
"first", 50
"early", 0
"default", -50
"late", -100
"last".
As different plugins discover more precise relations to each other, the values should be further tweaked.
Usage example:
@plugins.event_priority(-100) # Wishing to run this after all other plugins' `on_files` events.
def on_files(self, files, config, **kwargs):
...
New in MkDocs 1.4. Recommended shim for backwards compatibility:
try:
from mkdocs.plugins import event_priority
except ImportError:
event_priority = lambda priority: lambda f: f # No-op fallback
New in version 1.6
There may also arise a need to register a handler for the same event at multiple different priorities.
CombinedEvent
makes this possible.
mkdocs.plugins.CombinedEvent
Bases: Generic[P, T]
A descriptor that allows defining multiple event handlers and declaring them under one event's name.
Usage example:
@plugins.event_priority(100)
def _on_page_markdown_1(self, markdown: str, **kwargs):
...
@plugins.event_priority(-50)
def _on_page_markdown_2(self, markdown: str, **kwargs):
...
on_page_markdown = plugins.CombinedEvent(_on_page_markdown_1, _on_page_markdown_2)
Note
The names of the sub-methods can't start with on_
;
instead they can start with _on_
like in the the above example, or anything else.
Handling Errors
MkDocs defines four error types:
mkdocs.exceptions.MkDocsException
Bases: ClickException
The base class which all MkDocs exceptions inherit from. This should not be raised directly. One of the subclasses should be raised instead.
mkdocs.exceptions.ConfigurationError
Bases: MkDocsException
This error is raised by configuration validation when a validation error is encountered. This error should be raised by any configuration options defined in a plugin's config_scheme.
mkdocs.exceptions.BuildError
Bases: MkDocsException
This error may be raised by MkDocs during the build process. Plugins should not raise this error.
mkdocs.exceptions.PluginError
Bases: BuildError
A subclass of mkdocs.exceptions.BuildError
which can be raised by plugin
events.
Unexpected and uncaught exceptions will interrupt the build process and produce typical Python tracebacks, which are useful for debugging your code. However, users generally find tracebacks overwhelming and often miss the helpful error message. Therefore, MkDocs will catch any of the errors listed above, retrieve the error message, and exit immediately with only the helpful message displayed to the user.
Therefore, you might want to catch any exceptions within your plugin and raise a
PluginError
, passing in your own custom-crafted message, so that the build
process is aborted with a helpful message.
The on_build_error event will be triggered for any exception.
For example:
from mkdocs.exceptions import PluginError
from mkdocs.plugins import BasePlugin
class MyPlugin(BasePlugin):
def on_post_page(self, output, page, config, **kwargs):
try:
# some code that could throw a KeyError
...
except KeyError as error:
raise PluginError(f"Failed to find the item by key: '{error}'")
def on_build_error(self, error, **kwargs):
# some code to clean things up
...
Logging in plugins
To ensure that your plugins' log messages adhere with MkDocs' formatting and --verbose
/--debug
flags, please write the logs to a logger under the mkdocs.plugins.
namespace.
Example
import logging
log = logging.getLogger(f"mkdocs.plugins.{__name__}")
log.warning("File '%s' not found. Breaks the build if --strict is passed", my_file_name)
log.info("Shown normally")
log.debug("Shown only with `--verbose`")
if log.getEffectiveLevel() <= logging.DEBUG:
log.debug("Very expensive calculation only for debugging: %s", get_my_diagnostics())
log.error()
is another logging level that is differentiated by its look, but in all other ways it functions the same as warning
, so it's strange to use it. If your plugin encounters an actual error, it is best to just interrupt the build by raising mkdocs.exceptions.PluginError
(which will also log an ERROR message).
New in version 1.5
MkDocs now provides a get_plugin_logger()
convenience function that returns a logger like the above that is also prefixed with the plugin's name.
mkdocs.plugins.get_plugin_logger(name: str) -> PrefixedLogger
Return a logger for plugins.
Parameters:
-
name
(str
) –The name to use with
logging.getLogger
.
Returns:
-
PrefixedLogger
–A logger configured to work well in MkDocs, prefixing each message with the plugin package name.
Example
from mkdocs.plugins import get_plugin_logger
log = get_plugin_logger(__name__)
log.info("My plugin message")
Entry Point
Plugins need to be packaged as Python libraries (distributed on PyPI separate
from MkDocs) and each must register as a Plugin via a setuptools entry_points
.
Add the following to your setup.py
script:
entry_points={
'mkdocs.plugins': [
'pluginname = path.to.some_plugin:SomePluginClass',
]
}
The pluginname
would be the name used by users (in the config file) and
path.to.some_plugin:SomePluginClass
would be the importable plugin itself
(from path.to.some_plugin import SomePluginClass
) where SomePluginClass
is a
subclass of BasePlugin which defines the plugin behavior. Naturally, multiple
Plugin classes could exist in the same module. Simply define each as a separate
entry point.
entry_points={
'mkdocs.plugins': [
'featureA = path.to.my_plugins:PluginA',
'featureB = path.to.my_plugins:PluginB'
]
}
Note that registering a plugin does not activate it. The user still needs to tell MkDocs to use it via the config.
Publishing a Plugin
You should publish a package on PyPI, then add it to the Catalog for discoverability. Plugins are strongly recommended to have a unique plugin name (entry point name) according to the catalog.