Open source software is awesome. It's a shared resource that we all work on to benefit ourselves and others all at once. CircuitPython is built on the open source code of MicroPython for example.

One challenge of open source software is making your code and documentation available in a way that makes it accessible to those that use it. It's not just a matter of writing code and posting it somewhere on the internet. For the core of CircuitPython we have existing source control on GitHub, low-level documentation on ReadTheDocs, continuous testing on GitHub Actions, tutorials on Adafruit Learn System, community support on the Adafruit Forums and live chat on Discord.

However, there is much more to CircuitPython than just the core code. There are also additional libraries that expand what can be done on CircuitPython. They may add additional support for a particular FeatherWing or making simple tasks easier.

To make CircuitPython libraries easy to find and understand we've come up with a number of best practices when creating a new library. This guide is an overview of those practices with links to more information on specific topics.

Getting coding

Usually when creating a library to share with the world, the biggest issue isn't the code itself. Its common for a library to evolve naturally out of an existing project. It may have been a clock project where you switched the real-time clock partway through or a simple project which has grown in complexity demanding some code refactoring. The first challenge is identifying what can make up a library.

In CircuitPython (and Python in general) libraries are known as packages and modules. A module is a single python file that can be imported by another file by using the import statement. A package is a folder filled with modules and is used to group common modules together. The adafruit_rgb_display package is an example of this. Both packages and modules can be called libraries.

By importing libraries, code can gain functionality without the clutter of additional code within the same file. Instead, one only needs to understand the Application Programming Interface or API to use the power of the imported library. Its not important to read and understand the code that implements the API. Outside of software, you can think of the steering wheel in a car as the API to the turning functionality of the car. You don't need to understand how the car's wheels turn, you just know that when you turn the steering wheel, the car turns.

Hopefully, as your project has evolved, you've begun to use functions and classes to create APIs. Creating powerful, easy to understand APIs is a difficult thing to do and something a Google search can provide many resources on. Here is one Python specific talk from PyCon 2017:

Once you've got an API you are happy with, its time to make it available to the world.

Cookie cutter

One of the first helpful tools for creating a CircuitPython library is called CookieCutter. It is an open source project that was made to help manage common content between projects. This common content is also called a boiler plate. For CircuitPython libraries, this boiler plate can include more than twenty files. Whoa! Thats a lot for a simple library. However, they are all useful. Only one will hold your code. Below is an overview. Don't worry too much about understanding it all because most relate to topics we'll discuss later.

  • .gitignore - Prevents autogenerated files from being shared publicly by git.
  • .github/workflows/ - A set of workflows run by GitHub Actions when something is committed, a PR is opened or updated, or the library is released.
  • - Outlines expectations for behavior of contributors to ensure a friendly, welcoming environment. Source from Contributor Convenant.
  • {something}.license - These contain license information for the file specified by the name. e.g. examples.rst.license is the license for examples.rst.
  • LICENSE - Contains the legal text for the license which allows others to use and modify the code you've written. We default to MIT but others can be used. We recommend by GitHub for an easy way to decide what is right for you. Typically a comment will be placed at the top of source code files to clarify their license if it gets copied away from the rest of the files.
  • README.rst - This is your most important documentation! It is the first thing people read when they unzip your source or pull it up on GitHub. It can be plain text, Markdown (.md) or reStructuredText (.rST). The latter two will be nicely rendered on GitHub. We prefer .rST for CircuitPython libraries because it makes it easy to integrate with Sphinx, ReadTheDocs and autogenerated documentation.
  • adafruit_{name}.py - Your code! This is the module you can use to organize your code. It automatically includes example license text and comments. It should start with a different name for Community Bundle projects.
  • docs/api.rst - Triggers the autogeneration of documentation for the above module.
  • docs/ - Configuration settings for Sphinx.
  • docs/examples.rst - Examples section, gets inclueded in documentation pages.
  • docs/index.rst - Table of contents and main docs page section.
  • readthedocs.yml - Configuration settings for ReadTheDocs
  • requirements.txt - Lists other libraries that your depends on to work correctly. Typically things you import in your code that aren't provided by CircuitPython automatically.
  • .pre-commit-config.yaml - Configuration settings for the pre-commit tool. This sets the versions for black and reuse.
  • .pylintrc - Configuration settings for Pylint.
  • - Required for PyPI deployment. File will be present even if library is not deployed to PyPI.
  • examples/{name} - You should put a basic example script in this file, to show how the library is used.

If your making a library for the Adafruit Bundle the filename of your library is adafruit_{name}.py. The {name} portion will be replaced when the cookiecutter is actually run. For Community Bundle projects you'll enter different values for some prompts so your code file will come out with a different name. Only libraries supported by Adafruit should keep the adafruit_ prefix.

Choose the right instructions from below depending on your operating system.

Installing cookiecutter (Mac & Linux)

cookiecutter is itself a Python library so you'll need Python and pip installed first. Then run:

pip install cookiecutter

Running cookiecutter (Mac & Linux)

Running cookiecutter is equally easy. All it needs is a location of where to pull the template from. Everything else is done via prompt.

cookiecutter gh:adafruit/cookiecutter-adafruit-circuitpython

Installing cookiecutter (Windows)

At the moment, cookiecutter is incompatible with Windows systems. You'll need to install an alternate version of cookiecutter:

pip install ansys-cookiecutter

Running cookiecutter (Windows)

Running cookiecutter is equally easy. All it needs is a location of where to pull the template from. Everything else is done via prompt.



  • target_bundle - Which bundle is the library for? 1 (default) for Adafruit Bundle. 2 for the Community Bundle.
  • github_user - GitHub user or organization which will host this repo. For example, Adafruit funded libraries should say adafruit here.
  • author_name - Who you are! Sets the copyright to you.
  • company - Used to give Copyright credit to the company funding the library. For example, Adafruit funded libraries should say "Adafruit Industries" here.
  • library_name - Shortest name for the library. Usually a chip name such as LIS3DH. THIS MUST BE ENTERED EXACTLY THE SAME WAY IT WILL LOOK IN THE GITHUB REPO URL (e.g. for, you must enter CharLCD). Use all caps or camel case as necessary. If you enter this differently than the GitHub URL, you will need to fix a number of things in your library later.
  • library_description - Write a sentence describing the purpose of this library (e.g. CircuitPython helper library for the DC & Stepper Motor FeatherWing, Shield and Pi Hat kits.).
  • library_keywords - Used to populate keywords for searching for the library on PyPi. Enter a string of lowercase keywords (e.g dht temp humidity) Please be thorough! The more search keywords you enter, the easier it is to find. This step should be completed even if you don't think the library will end up deployed on PyPI. NOTE: The following are included by default: adafruit, blinka, circuitpython, micropython, and library_name.
  • library_prefix - Used to prefix the code to the organization creating the library. For example, Adafruit supported libraries should say Adafruit here. Do not add a - or _.
  • adafruit_pid - Numeric Adafruit product ID associated with this library. Leave blank if none.
  • requires_bus_device - Determines whether to add comments about a dependency on Leave empty if the library won't use BusDevice.
  • requires_register - Determines whether to add comments about a dependency on Leave empty if the library won't use Register.
  • other_requirements - Add any other module dependencies. Enter a comma separated string of modules, using the lowercase full name of the module, using - instead of _  (e.g. adafruit-circuitpython-pca9685, adafruit-circuitpython-motor, pyserial). This is used to for PyPI. This step should be completed even if you don't think the library will be deployed to PyPI.. NOTE: Adafruit-Blinka is always included, so no need to include it here.
  • pypi_release - Will the library be released on PyPi? y/n.
  • sphinx_docs - Will the library have sphinx documentation files? y/n.
  • default_branch - Should the repo use master or main as the default branch name.

At this point all the files you need should be in place. Now you'll need to migrate your code into the generated .py file. If you have a lot of code you can make a directory of the same name and have multiple modules within it. For now, we'll keep it simply a module.


Once your code is there, test it in CircuitPython by copying it over to your board. If you get a MemoryError on import you can create a smaller .mpy file that's easier to load in CircuitPython.

Executables for mpy-cross for Windows, MacOS, Linux x64, and Raspbian are available at the link below. The 6.x versions will also work for 5.x. The .mpy format changed in CircuitPython 7, so check the version number in the mpy-cross filename.

Alternatively you can build mpy-cross yourself. You'll need to download (or clone) the CircuitPython source, make mpy-cross and run it on your source file. See the Building CircuitPython guide for detailed setup instructions.

git clone
cd circuitpython
make fetch-all-submodules
cd mpy-cross
./mpy-cross example/

Make sure example/ is changed to your file. It will make a example/adafruit_example.mpy file that you can copy to CircuitPython just like a .py file.

CircuitPython 3.x and 4.x: This zip contains mpy-cross binaries for MacOS X, Raspbian, Ubuntu x86 and Windows that will work with old versions of CircuitPython 3.x and 4.x:

Saving your work

At this point you've done a ton of work getting your library code going, initally tested and settings in place. Its time to commit your code and prep to publish this first version to the world.

Committing code is a way to keep track of how code changes over time. It also makes it possible to view what the files looked like previously. The simplest way to do this is to copy the whole folder to a new location. However, source control software is a much better way to organize it. Source control software makes it easy to view changes between versions (called diffs), back up the code to another computer and collaborate with others.

git is our and many others' preferred source control software. It was created by the same person as the Linux kernel and quickly grew in popularity due to its use there and its distributed nature. GitHub, a website for project hosting, grew up around git and is now the de facto place on the web to share your projects. Open source projects are hosted for free and private projects are hosted for a small fee.

Anyways, lets get git going to commit our code locally. See this guide for details on installing git. Consider this a quick start to initializing a repository (make sure you are in the library's directory).

The first command initializes a .git folder that stores the repository metadata. The second stages all of the files in the directory to be committed. Being more precise with individual files is recommended when changing your code for multiple reasons at once. Making two commits will make it easier to understand what changed and why. For now, we'll do one initial commit.

The final step will actually commit your code after prompting your for a commit message. In the message you should give a short one line summary followed by a detailed paragraph about the changes. The video below gives good tips on commit messages and code reviews.

Temporarily unable to load embedded content:
If you've used other source control theres an important distinction between commits in the subversion, cvs and perforce sense and the git sense. In those others, committing is the same as sharing with others. In git, a commit is for you alone until its shared or "pushed". That means you can change commits and revise the history of your code to better reflect your changes and motivations after committing it. This is known as "rewriting history" in git terminology and is super powerful. Its only when a commit is shared with others on a shared branch in a shared repo that it shouldn't be done because others may have built on top of it already.

Now that we've got everything squared away on our own computer, its time to share it with the world. The main avenue we'll use to share the code is GitHub, a code hosting site that is very popular with open source projects and developers. Once that is setup, we'll set up ReadTheDocs integration so that our documentation is hosted and automatically updated. We'll set up GitHub Actions so that our code is automatically checked by mpy-cross and that releases have mpy files on them.


GitHub is the bees knees. They offer free project hosting including code, issues, wiki and website for open source projects. It is extremely popular for good reason. They integrate with git to host your code.

So, the first step to sharing your project is creating the GitHub git repository.

The repository name can be anything but we're sticking to Adafruit_CircuitPython_{name} for official Adafruit libraries and suggest CircuitPython_{name} for community libraries. Fill in the description with a short, informative sentence about your library. DO NOT check the Initialize box. We'll do that with our existing files in the next step.

On the next page click SSH and copy the text to the right of it starting with git@. We're going to follow the second set of instructions because we've already created our local git repository (repo for short). The first step makes our local repo aware of the GitHub repo and names it tannewt rather than origin because I find it clearer to match the GitHub repository name. The second step pushes the commit we made before to the GitHub repository and makes the code public.

git remote add  
git push -u origin main
For others who have used another version control system, this is the point where you shouldn't change your commits. Before this they were private to you. At this point they are public and others may build on them.

At this point, you may have been prompted for your GitHub password. If you entered it correctly it should work. I use GitHub Desktop to manage my login and recommend it.

Now we can verify our code is live by refresh our browser tab.

Woohoo! We're live! While we're here. Lets add the circuitpython topic to our repo by clicking Add topics, typing in circuitpython and clicking Done.


Now if we scroll down, we'll see our README. This is the single most important document for the whole project. It should be concise but contain basic description of the library, what it depends on, how to use it, how to contribute and where to find more detailed documentation.

Since our README doesn't have that information we'll do a quick edit to add some more information. I use Atom (created by GitHub) for file editing. It has good git integration including highlighting edited lines and files in orange.

Now that the file is up to date, I do a quick commit and push to make it live.

And if we refresh the repo page we see our new and improved README.

Ok! We're up and rolling. The code we've written is public, we've learned how to update it further and have a good README. We can click Commits on the front page to see a list of all our public commits on the front page. Next up is getting our docs on ReadTheDocs.

Great documentation is a key component of a successful open source library. Luckily, after a little setup, its super easy to write and maintain great docs.


First, you should install Sphinx. Sphinx allows you to build the docs locally to ensure there aren't any documentation errors when you push to your repository.

pip install Sphinx sphinx-rtd-theme

Now, once you have Sphinx installed:

cd docs
sphinx-build -E -W -b html . _build/html

This will output the documentation to docs/_build/html. Open the index.html in your browser to view them. It will also (due to -W) error out on any warning like GitHub Actions will. This is a good way to locally verify it will pass.


ReadTheDocs is another free hosting service used by many open source projects. It integrates with GitHub so that every time new code or docs are pushed to your GitHub, the nicely formatted versions hosted by ReadTheDocs are also updated.

Lets get it setup! First, sign up and login if you don't already have an account. After signing in, your dashboard will be displayed.

On the dashboard click Import a Project and select Import Manually.

For the next page we'll need the https version of our GitHub repository so we go grab it by clicking Clone or Download on the repo front page and selecting Use HTTPS.

Then press the copy icon.

And paste it into the Repository URL field in ReadTheDocs. The name should be Adafruit CircuitPython Example. Then click next.

If you created your ReadTheDocs account and linked your GitHub (like I have below) then it should be able to automatically connect GitHub to itself by creating a Webhook. This Webhook is how ReadTheDocs gets updates from GitHub.

On the right hand side you should also see links to your documentation that you can use to link others to it.

Now refresh. We see that its been built successfully on the right and at the top we can view the docs.

Now, as a last step, lets make sure our docs look as expected by clicking the View Docs button.

Awesome! By default it shows our README.rst (didn't I say it was important?). Furthermore, the lefthand side has a link with our module's name. Lets click it to see the module's documentation.

Yay! Everything looks good. This is exactly as we've defined in our Python file. For now, docs are done. :-)

Adding Your ReadTheDocs Project as a CircuitPython Subproject

If you're creating documentation for an Adafruit repository, you'll want to add your project as a subproject to the CircuitPython project. Note that you will need access to the CircuitPython project to do this.

Before you do that though, you'll need to add Adabot as a maintainer. 

First, click on the Admin tab. Then, scroll down and click on the tab labeled Maintainers and at the bottom of that tab, enter the name of the user to add, adabot in this case, and press Add.

Now, navigate over to the CircuitPython project, click on Admin, click on Subprojects, click Add Subproject, and select your project. Set the alias to the project name with Adafruit CircuitPython removed.

Maintaining code is quite a challenge over the long-term. Over the lifetime of a library, many people read and edit it. Each person has a different background in coding and also a different goal in mind. These variations can lead to inconsistencies throughout the code that makes it a bit harder for the next person to understand. In CircuitPython libraries, we use tools like Pylint and Black to ensure consistency in new code.

As we've added more automated checks, we've changed to a system called pre-commit to manage the checks overall. Once installed properly, you can run pre-commit locally, before committing new code into Git. It also runs remotely on GitHub when a pull request has been proposed. pre-commit is set up to remotely run on all existing libraries. It will automatically run remotely on a new library thanks to cookiecutter.

However, it won't run locally unless you install it into your local directory. We highly recommend doing this because it will both check and fix your code locally.

One-time initial install of pre-commit

If you've never used pre-commit on your computer before, you'll need to install it globally (there is a second "install" for each repository.) The easiest way to install it is with pip.

pip install pre-commit

Workaround for pre-commit issues on Ubuntu 22.04 and Debian

In ubuntu 22.04 or the analogous Debian release, you may see the error "expected environment for python to be healthy immediately after install" when trying to use pre-commit. To fix this, add this line to your .bashrc or .bash_aliases file, or other shell startup file. Restart your shell as necessary to pick up this setting.


This export must be present before pre-commit sets up its virtualenv environment, which happens the first time you do pre-commit run or you try to push a commit. If the virtualenv is already set up, do pre-commit clean, which removes the existing virtualenv.

See the instructions from the pre-commit project for installation for alternative ways of installing pre-commit.

Do not use the export SETUPTOOLS_USE_DISTUTILS=stdlib on Ubuntu 24.04 or Debian Bookworm. It will cause some uses of pip install to fail.

Per-repository installation

For every new repository, you'll need to perform an pre-commit installation. This installs the specific versions of checks that the repository specifies. From within the repository do:

pre-commit install

After running this command, pre-commit will automatically run when you do git commit.

However, if you don't do this, you can still run pre-commit manually.

Running pre-commit

pre-commit will run each check every commit for all of the modified files and either pass or fail. Most checks that fail will also modify the source file to make it pass (like removing extra spaces). Once that happens, you'll see newly modified files in git status. git add them and then try the commit again.


You can run the pre-commit checks on every file whenever you like with:

pre-commit run --all-files

More Info

For more info on pre-commit see

If you are using pre-commit, you do not need to follow these steps. pre-commit runs Pylint for you. We highly recommend using pre-commit. For more information, check out this page:

Now that you've installed Pylint and downloaded the .pylintrc configuration file, you're ready to start linting. First thing we need is an example to check.

Download using the "" link below. Then, place the file in the same location as your recently downloaded .pylintrc file.

Temporarily unable to load content:

Pylint looks in a series of locations in order to find the configuration file. The first place it looks is the current working directory. This is the easiest way to ensure you're using the right configuration file.

Return to your terminal program or command line. From the command line, navigate to the folder containing and .pylintrc. Then, run the following command:


Alright! Your first error! Consider it a badge of honor. And don't worry! The next section will walk through how to read the Pylint output with a series of common errors. Time to start linting!

If you are using pre-commit, you do not need to follow these steps. pre-commit runs Black for you. We highly recommend using pre-commit. For more information, check out this page:

Black is a really useful code formatting tool maintained by the Python Software Foundation. It reformats files to improve readability.

It can be run in two ways, the first just checks to see if there is any code it would reformat. This is the way we use on all of our CircuitPython repositories. The second way actually reformats the files. This is the way you'll be wanting to use locally.

This page explains how to install and run black, the different ways to run it, and some cases when you may not want to adhere to black's suggestions.

Installing Black

Installing black is super easy. Simply run:

pip install black

Note: if you also have a version of python2 installed, you may have to run:

pip3 install black

Using Black

As I mentioned above, there are two ways to run black. The first way is just checking the code. This is what GitHub actions runs to test commits and pull requests.

This is run by typing in Linux:

black --check --target-version=py35 .

And for Windows command line (Python 3 must be installed):

python -m black .

You can replace the . with whatever files you want it to check if you don't want it to check every .py file in your current directory.

Here's what that output looks like:

However, most of the time, you're going to want Black to actually reformat the code. This is accomplished by running:

black --target-version=py35 .

Here's what running that looks like:

Black isn't always right

Sometimes, Black will make a change that just looks really bad. We've mostly encountered this with longer lists of numbers or short strings. 

For example: Black would make each element of this list have it's own line.

Here's what the list looked like originally:

heatmap_gp = bytes([
    0, 255, 255, 255, # White
    64, 255, 255, 0,  # Yellow
    128, 255, 0, 0,   # Red
    255, 0, 0, 0])    # Black

Here's what the same list looked like after being reformatted by Black:

heatmap_gp = bytes([
	255, # White
	0,  # Yellow
	0,   # Red
])    # Black

You can disable black In that section by adding # fmt: off at the start of the section you don't want Black to reformat and # fmt: on at the end of said section to re-enable it.

Here's how we disabled Black for the list above:

# fmt: off
heatmap_gp = bytes([
    0, 255, 255, 255, # White
    64, 255, 255, 0,  # Yellow
    128, 255, 0, 0,   # Red
    255, 0, 0, 0])    # Black
# fmt: on

Lint Black's changes

Make sure that after you run Black, you re-run Pylint. They don't always agree, and their major point of disagreement (Pylint's bad-continuation check) has been dealt with for all of our CircuitPython repositories.

Documentation and Contributing

Check out Black's ReadTheDocs page.

Also, Black's source code is hosted on a public GitHub repository.

Navigating GitHub Actions

To navigate to the actions tab, click on the 'Actions' tab circled near the top of the page.

  1. This is the box that contains all of the Actions workflows that have been run, will be run, or are running for a particular repository. There are four different workflows that Adafruit CircuitPython libraries use. Workflows are a series of tests run to determine if a PR, commit, or release is up to the desired standards and is correctly working. Unless you're releasing libraries, you'll only need to worry about the 'Build CI' workflow.
  2. This icon tells you the status of the workflow. A green check means it has run successfully. A red X means it has failed, and an orange circle means it is either waiting to run or currently running.
  3. This menu allows you to filter by Workflow. The first three workflows are run upon release, and mostly deal with packaging and deploying the newest version of the library. The fourth one is run when a commit or PR is made and checks formatting and documentation.
  4. You can click on the bolded title to the left of number 4 to see the readout from the workflow.

When your PR fails

As soon as you open a PR, you should see a box with an orange border. It will tell you the status of you PR. While the workflow is running, it will say "Some checks haven't completed yet."

The Actions CI often takes around a minute to run, but depending on the amount of workflows currently being run, and the size of the repository, it could be longer or shorter. Below, you can see that the PR is currently failing actions. To find out what specific test it is failing, you can click on one of four places. First, you can click on the red X to the left of the commit hash. Second, you can click on the red X to the left of the GitHub logo in the orange box. You can also click on details, on the same line as the previous but all the way to the right. Finally, you can click on the 'Actions' tab at the top of the page. This works but it isn't ideal as you'll have to navigate to the test for your PR.

Now that you're on the Actions page for that specific commit, you can find out what test is causing it to fail. There are three tests a PR will fail on: black formatting (check formatting), PyLint, and Build Docs. In this case, the Black formatting check is failing.

Note: If one test is failing, tests that run after it will not run

Fixing a Black formatting failure is actually very easy. Check out the Black page in this guide for info on running Black. Then commit and push your changes.

Now that we've ran Black, and committed and pushed those changes, we can check back in with the PR. It's still failing, so we can navigate to the Actions page again.

It looks like the Black formatting check is now passing, but Pylint isn't. You can find out how to use Pylint in the Pylint page in this guide.

Now, run Pylint in the your repository's base directory. Make the changes it says to make and re-run Pylint until it says that your code is rated a 10/10. Then, commit and push those changes.

The PR is still failing, so we should check the Actions workflow output again.

The good news is that the Pylint check is no longer failing. However, it appears that sphinx is failing. Click on the 'Build Docs' dropdown to see that test's output. It looks like I inadvertently removed some necessary colons. 

By going into the docs directory, and building sphinx documentation like in the Sharing Docs on ReadTheDocs page in this guide, we can mirror the output of the 'Build Docs' test. In this case, the issue is in index.rst. After fixing those issues, run the command again to verify that the docs are building correctly. In this case, they are, so we can commit and push those changes.

Now that everything is passing and the box has a green border, the PR is ready for review.

Phew, what a whirlwind. Now that we've got our code on GitHub, our docs on ReadTheDocs and GitHub Actions going we're ready to release!

A release is simply a point in the lifecycle of a software project where the maintainers believe the software is worth trying and using. Pre-release releases are used for cases where the code is in early preview and may still contain bugs. Stable releases typically aren't released with known bugs and are geared towards wide adoption.

Creating or "cutting" a new release is also the point where the maintainer typically provides pre-built or binary releases. In our case, we've setup Actions to automatically provide mpy binary files for every release.

To get started, pull up your repo and click the releases button.

Once there, click Create a new release.

Now, decide on a version number for this release. CircuitPython's tools require Semantic Versioning (shortened to SemVer) to work. SemVer is three numbers separated by periods. This is not the same as a decimal number since there are two periods. Anything less than 1.0.0 is typically pre-release and anything after marked alpha, beta or rc for release candidate is also. For example, 1.0.0-rc.1 would be the first release candidate for 1.0.0 and be pre-release. Normal releases would be something like 0.10.0. Do not start the version with a v.

The first number should be incremented and others reset to 0 when the library was changed in a way that may require old code to be rewritten. Changing public function names for example would cause 1.3.0 to become 2.0.0.

The second number should be incremented, leaving the first and setting the third to zero if new functionality was added but the existing code is compatible. For example add a new function would lead from 1.3.3 to 1.4.0.

Lastly, the third number should be incremented if existing functionality was fixed. For example a function released in 1.4.0 didn't always work as expected and it was fixe in 1.4.1.

Once you decide on a version number enter it into the top box. This will create a git tag that will mark when in the git history the release occurred. You can also do this from git and then refer to an existing tag here.

Next, fill in the title and give the release a description including pointers to other related resources.

The release description uses Markdown for formatting and you can preview what it would look like after the save using the Preview tab. Here is a copy of the release description so you can start with it!

Adds basic `hello` functionality.

To use in CircuitPython, download the .mpy file and copy it to the `lib` folder on the `CIRCUITPY` drive. Or, simply install the [Community bundle](

Read the [docs]( for info on how to use it.
If you chose not to set up Actions to build your release binaries, this is when you can upload them yourself.

Now that everything looks good, lets publish the release. Don't forget to tick the pre-release box (under the arrow) if you want this to be marked as such.

After you hit publish you'll be taken to the release page on GitHub. This is a great place to link folks to so they can get all of the latest information on the release. GitHub includes zip files of the full source code by default.

The release also triggers a few release workflows from Actions. Navigate to the Actions page to check on them. Everything should be green.

Note: for a release to be deployed to PyPi, it requires the PyPi credentials to be added to the repository secrets.

Once Actions is green, the release should be updated with the .mpy file for the release. This file can then be copied to the CIRCUITPY drive to be used.

The other way folks can use the library is by downloading it as a part of a larger bundle. We have the Adafruit bundle for officially supported libraries and the Community bundle for those created and maintained by the larger community.

Common Release Failures

Almost all release failures happen in the 'Build and Publish' section of the 'upload-pypi' workflow. Here are the two most common ones.

HTTPError: 400 Client Error: File already exists.

This happens when there is no change in the driver since the last release. This usually happens when something needs to be re-released with no changes for some reason, often the incorrect semver version being used. The easiest way to fix this is to make a tiny change in a docstring or comment in the driver file and re-release.

HTTPError: 403 Client Error: Invalid or non-existent authentication information.

This happens when PyPi credentials haven't been added. It's possible that library shouldn't be on PyPi, so be sure of that before adding credentials. To get that fixed, create an issue on GitHub and assign it to CircuitPythonLibrarians, and briefly explain the issue, linking to the Release Action.

Error: unexpected status code: 403 {"message":"Resource not accessible by integration" while uploading a release

This happens when the workflow on github is not allowed write access to the repository. To fix this, go to "repository settings". On the left, open the "actions" section and click "general". Finally, under "workflow permissions" allow "read and write permissions". This is necessary in order to allow uploading files to a GitHub release from the release action.

We've created a couple library bundles for CircuitPython. The idea is that one can download all available libraries once and then copy them onto the CIRCUITPY drive as needed.

Another advantage of the bundle is that it makes it easy to find a list of all libraries. Even if you can't store all of the libraries on the drive, for example the Gemma M0 is only 64kb, the bundle makes it easy to copy over what you need when you need it.

It also helps CircuitPython devs keep track of libraries and update them as things change.

So, lets get our example library into the community bundle. To do so, we'll make a pull request on GitHub. First, lets fork the community bundle.

After the fork is done, the repo will look the same except the repo name will be different at the top left.

Next we'll clone the repo to our computer. See this guide for detailed instructions on cloning and submitting a pull request. Consider this a quickstart.

When cloning, make sure that you add --recursive to get all of the submodules too.

Adding the submodule to the Library Bundle

Now to add our library to the bundle we'll do:

git submodule add URL_TO_LIBRARY_INCLUDING.git libraries/SUBFOLDER/lowercase_library_name

You'll change the following:

  • URL_TO_LIBRARY_INCLUDING.git  - This will be the full URL to the library including the .git on the end. To get this URL, click on "Clone" in the GitHub repo, choose HTTPS, and copy that URL. Or you can copy the URL from the repo and add the .git.
    • Example:
  • SUBFOLDER - The subfolder will be helpers or drivers. Choose which based on your library. Is it a driver? Choose drivers. Is it a helper? Choose helpers.
    • Example:  libraries/helpers/ or libraries/drivers/
  • lowercase_library_name - This is the name of your library, minus the Adafruit_CircuitPython_, in lowercase letters.
    • Example: pyportal


  • git submodule add libraries/drivers/pyportal

And then we'll verify it worked by looking at the build log and versions file.

Updating the Library List

Before we submit a pull request with our new library, lets update the bundle documentation to include a link to the documentation for our library. Use whichever method you wish to open /docs/drivers.rst for the Adafruit bundle or for the community bundle.

In library listing, choose one of the categories to place your documentation link. The current categories are: Board-specific Helpers, Helper, Blinky, Displays, Real-time Clocks, Motion Sensors, Environmental Sensors, Light Sensors, Distance Sensors, Radio, IO Expansion, and Miscellaneous. Or, you can create a new category if that best describes your library.

With your category chosen, simply add a new line to that section using the file's hyperlink format. The easiest way, is to copy an existing line and edit it to change both the link text and the URL.

The reST markup language is indent-sensitive. Make sure you are properly aligned under the '.. toctree::' directive.

Next, lets do our normal commit and push process for a new branch.

Going to GitHub will show an easy prompt to create the pull request.

Follow that to see a differences. Make sure that at the top you are comparing against adafruit/CircuitPython_Community_Bundle as the base fork and your branch as the head fork. Make sure at the bottom left that maintainers can edit it (for simple fixes before they submit). Then click Create pull request at the bottom right.

This is the last step you need to do. One of the maintainers will come and respond on the pull request. Once they are happy with the request, it will be merged into the bundle and go out with the next release. It'll show up as purple once accepted.

After a library has been merged into the bundle, the version included will be automatically updated once a day. You don't need to make any other changes to the bundle repo. Simply release new versions in the library repository, and within 24 hours the bundle will be updated and released with the new version.

This is my checklist since there's so many things to do - maybe useful for others but its specifically for people deploying 'official' adafruit libraries

  1. Create Adafruit_CircuitPython_sensor in
  2. Go to git repo settings, manage access, add CircuitPython Librarians team, then set to Write
  3. Fork to @ladyada
  4. git clone to M4 board and to personal library folder
  5. cd to the personal library directory in MSYS2, run cookiecutter gh:adafruit/cookiecutter-adafruit-circuitpython
  1. Move everything from the new directory up one to git repo
  2. Remove

Open up and fix capitalization of first naming after license

Add short description for first .. todo::

Add URL or remove second todo

Remove third todo, uncomment deps

Like so!

  • Open up requirements.txt - verify Adafruit-Blinka is there, possibly adafruit-circuitpython-busdevice. For UART devices, add pyserial
  • Open up README.rst and update the description .. todo::
  • Open up docs/index.rst and remove the todo's
  • Do first commit & PR
  • Write code!
  • Add example to examples/ empty file
  • Run pylint on and examples/
  • Once those pass, do another commit
  • Put example in README.rst like so:

Python is considered a dynamically-typed language, meaning that the type of a variable can change during runtime if its value is changed to something else. For example:

x = 7
print(type(x)) # prints: <class 'int'>
x = "Hello World"
print(type(x)) # prints: <class 'str'>

In the code above, the variable x contains an integer type value when it is originally set. Afterward, the value is changed to "Hello World", which is a String type value. Ordinarily when we write Python code, we do not need to declare the specific types of the variables; instead the Python interpreter will keep track of the current type internally, based on the values that we assign to the variable.

The opposite of this would be a statically-typed language such as C or Java. In a statically-typed language, the developer must explicitly declare the type when a variable is created, and the type is not able to change simply by changing the value. If the user wants the type to change, they must use a specific conversion or casting function to create a new variable of the desired type.

What is Typing Information?

Type hints are an optional extra bits of code that declare the intended types for function arguments and returns.  Adding type hints to Python code is basically sharing the developers intentions in the code. It is stating something like "the person who wrote this code says that this variable is a String and should always remain a String. If you are writing code that interacts with this it is safe for you to assume this variable will be a String." Since Python is a dynamically-typed language, it is not required for us to declare types like this, but there are some benefits of doing so even though it isn't required.

Benefits of Typing

The Python and CircuitPython interpreters ignore typing information. So it will make no difference when your code is running whether or not you've included types. The real benefits of typing are felt during the development phase, rather than at runtime. Types allow you, the developer, to be more certain about what kinds of values are expected to be contained by variables. This can make it easier to spot bugs in the code before it's executed. For instance:

x = "Hello World"
print(x.lower()) # prints: hello world
x = 7
print(x.lower()) # raises AttributeError

In the code above, the x variable is initially assigned the value "Hello World", which is a String type. In Python, Strings have a function called lower() that returns the lower-cased version of the String. The first print() statement executes successfully. Afterward the value is changed to 7, an integer type which does not contain the lower() function. So when the second print() statement tries to execute, it causes an AttributeError to be raised, which will crash the program if it's not caught. Type hints try to help us spot this type of error earlier, ideally before we've even attempted to run the program at all. IDEs and other programs can parse the code with type hints and attempt to warn us ahead of time when situations like this arise.

In the image above, PyCharm IDE has highlighted part of the code and shows a warning  to the programmer that they are attempting to access the lower() function, which does not exist for the integer type; it's likely to cause an error if it's executed before resolving this problem. PyCharm and other IDEs use the typing information provided to double check our work and warn us when we accidentally try to use a variable in an invalid way.

The Syntax of Type Hints

So now you know what type hints are and why they are beneficial, but how do you actually add type hints to your code? For CircuitPython libraries, we are primarily concerned with types for function arguments and returns. CPython (desktop computer Python) supports other types beyond them. But these are the ones we are focused on right now.

Function Argument Types

To add types to function arguments, we code a colon : after the argument name and then put the type after that. For example:

def unhexlify(hexstr: str):
    # code that would be here isn't relevant to the example

In the example above, the argument hexstr has its type declared as str, which is short for String. So the type hint is indicating that when you call unhexlify(), you should always pass it a String type variable as the hexstr argument. If there is more than one argument, then each argument should get its own colon and type declaration; if you are providing default values for the arguments, they are coded with a equals sign after the type. Here is another example showing a more complex function with its arguments typed:

def __init__(
        i2c_bus: busio.I2C,
        address: int = 0x0C,
        gain: int = GAIN_1X,
        resolution: int = RESOLUTION_16,
        filt: int = FILTER_7,
        oversampling: int = OSR_3,
        debug: bool = False,

In this example, there are 5 integer type arguments, 1 boolean type argument, and 1 busio.I2C object argument. Many of the arguments are given default values of constant variables declared in the class that this comes from. Note that the first argument self does not receive a type.

Function Return Types

To code function return types, we use a hyphen and an angle bracket to make an arrow that points to the type that will be returned by the function. This is coded after the parentheses containing the arguments, and before the colon that indicates the beginning of the functions code definition.

def last_status(self) -> int:
	# code that would be here isn't relevant to the example

The above function has been declared to return an integer type value. So any code using this function should be expecting to receive an integer value and should not try to treat it as a String or anything else.

One thing to keep in mind while coding Python classes is that the __init__() function always returns the None type. Here is an example of an __init__() function with its return type coded:

def __init__(self, spi: busio.SPI, cs: digitalio.DigitalInOut) -> None:
	# code that would be here isn't relevant to the example

CircuitPython-Specific Considerations

In CPython ("normal" desktop Python), there is a built-in module called typing which contains some helper classes sometimes used when declaring types. typing was introduced with Python version 3.5. CircuitPython does not currently have this module, so any code that attempts to import it will raise an ImportError if it runs on a CircuitPython microcontroller. This may sound problematic, but it actually ends up working out in our favor. We can catch this exception and ignore it, and since types make no difference at runtime, our code will still work properly. The benefit is that we can use this exception as an indicator that the code is running on a microcontroller and choose to ignore the error without importing any of the other classes used only for typing, which will save some precious RAM on the microcontroller.

	# imports used only for typing
    from typing import Tuple # <- this line causes the error
    # any imports below this won't happen if the error gets raised
    from circuitpython_typing import ReadableBuffer
    from busio import I2C
except ImportError:
    pass # ignore the error

In the CircuitPython libraries, we code the imports as shown. The first import should always be from the typing module. The remaining imports after the first one are other classes that may actually exist in CircuitPython or in CircuitPython-compatible libraries such as circuitpython_typing

By putting the typing import first, it will cause the ImportError to be raised before any of the classes that might actually exist get imported, preventing them from being imported and consuming RAM that won't be used when the program executes.

Finding the Correct Types

Now you know why type hints are helpful, and you know the syntax used to code the type hints. But you may be wondering how do you figure out which type an argument or return is supposed to be so that you can code the hint? If you are authoring a brand new library from scratch, you probably will know what types are expected since you are the one writing the code and choosing which things get passed in as arguments and which things get returned from functions.

If you're adding type hints to an existing library, you'll have to put on your detective's hat and look for context clues and other evidence that suggests which type any given argument or return type is supposed to be. 

Common Places to Look

No two libraries or functions are the exact same, so there is no "one size fits all" approach to determining types. However there are a few common places to look around in the code that are likely to provide good clues about the types.


If the code contains docstring comments, this might already have the type listed in them and you can use the type specified in the docstring for the type hint.

def _get_pixel(self, xpos, ypos):
        Get value of a matrix pixel
        :param int xpos: x position
        :param int ypos: y position
        :return: value of pixel in matrix
        :rtype: int
        # code that would be here isn't relevant to the example

In this function, the docstring comment does document all of the arguments and their types as well as the return type. In this case, the detective work is mostly done for us, we just need to put the types specified by the docstring into our type hints using the syntax noted above.

Here's how it would look after adding the type hints:

def _get_pixel(self, xpos: int, ypos: int) -> int:
        Get value of a matrix pixel
        :param int xpos: x position
        :param int ypos: y position
        :return: value of pixel in matrix
        :rtype: int
        # code that would be here isn't relevant to the example

Function Definition

If there is no docstring comment, or if it does not contain the argument and return types, then you'll have to dive a bit deeper. The next place to look is inside of the function definition. You're looking for the lines of code that use the argument variables, and then from those lines of code you may be able to determine what the type is meant to be.

def scroll(self, delta_x, delta_y):
        if delta_x < 0:
            shift_x = 0
            xend = self.width + delta_x
            dt_x = 1
            shift_x = self.width - 1
            xend = delta_x - 1
            dt_x = -1
        if delta_y < 0:
            y = 0
            yend = self.height + delta_y
            dt_y = 1
            y = self.height - 1
            yend = delta_y - 1
            dt_y = -1

In this function, we see that there are two arguments that we want to find the type for, delta_x and delta_y. When we look in to the definition code for the function, we see a couple of if statements that are comparing delta_x and delta_y to 0 using the less than operator <. We also find some lines of code performing mathematical operations with delta_x and delta_y such as:

xend = delta_x - 1

In this code, it appears that delta_x and delta_y are being treated as numbers. Numbers can be compared to other numbers with greater than or less than, and they can also be added and subtracted to other numbers. None of the numbers shown contain decimals, and there is no division operator being used, so it's a good guess that these numbers are specifically integers which use the abbreviation int for the type. If we found decimals, or if the variable we are trying to type came from the division operator such as:

some_var = 100/3

then it may be more likely that our numbers could be of the float type rather than int. You'll have to look for context clues in the code like this to try to make the most educated guess possible based on the information you find.

Function Usage

If you don't find concrete evidence in the function definition, another place to look is in the code that is calling the function. This might be inside of an example file, or perhaps in another source code file within the library if it is a multi-file package; sometimes it could even be a usage in the same file where it's defined, but just in a different area of code within that file. Try using the find utility with the function in your code editor to search for usages. Some of the more advanced IDEs even provide a specific "Function Usages" menu item when you right click a function name that will parse all of the project files and show you a list of every found usage. Here is an excerpt from an example script:

text_to_show = "Hello world"
text_area = bitmap_label.Label(terminalio.FONT, text=text_to_show)

In this piece of code, we see a variable named text_to_show which has been given the value of "Hello World". We know that "Hello World" is a String because it has quotes around it which is exactly how strings are coded.

Next we see that text_to_show has been passed to the Label() initializer function as an argument named text. So from this, we can conclude that the argument named text has the type of string.

In this case, the variable names also contain a clue: since they both mention the word "text", we know that Strings in Python store text values, so this is further evidence that suggests string is the correct type.

These pages discuss processes no longer used. We are maintaining them for posterity.

This process is no longer used in Adafruit CircuitPython repositories. If you want to know how to test your contribution to a CircuitPython repository, please see the GitHub Actions page:

One of the biggest challenges for software projects of all kinds is ensuring the software works as expected. This isn't as simple as it seems, particularly when multiple people are working on a project at once. Each person has different expectations about what the code should do and different testing setups.

Testing is used to make sure that each person's expectations are compatible and that the software still meets those expectations. Often testing is overlooked out of an urge for expediency. However, its well understood that, in the long term, good testing can help increase the development pace of a library. This is due to the confidence gained from tests that validate all expected functionality. It allows a maintainer to know immediately whether a requested change breaks any existing expectations or not. If it doesn't break anything, then its much easier to accept into the existing code.

Rust uses thorough testing as part of their "Not Rocket Science Rule" of software engineering. More info here:

Unfortunately, testing can be complicated and is worth covering in another full guide. This is especially true with CircuitPython where code is running on a microcontroller rather than a full operating system. For now, we'll just cover the basics of setting up Travis.

Travis CI

A cornerstone of good testing practice (and the Not Rocket Science Rule) is continuous integration (CI for short). CI is the process of automatically running tests on every change to a source code repository. In our case, we'll hook up Travis CI to our GitHub repo so that it tests each commit including proposed pull requests. Furthermore, it will automatically create an mpy file for each of our releases.

To get started with Travis, visit the website and sign up. Once signed in, you'll see a list of currently active repositories.

Once there, click the plus to activate your new repository.

Now scroll down the page until you see the new repo and click the switch on the left to activate it.

Now that its activated, click the repo name to see its status.

Now from that page click Build History. Nothing should appear yet, so do a small commit (Add more to your README for example) to test that its set up correctly. Refresh the build history tab and you should see a yellow pending build. Once it completes, it should be green.

Releasing mpys

mpy files are convenient, binary Python files that are easier to load when memory is tight. Travis can automatically build them and attach them to a release. The .travis.yml file already has everything configured except for permissions to attach files to releases. :-) (Shout out to TonyD for adding this support.)

This involves granting a third-party many rights to your public repositories. If that makes you uncomfortable, skip this step and manually attach mpy files to releases instead.

To give Travis rights to attach files to our releases we'll need to generate a personal access token for GitHub. Never ever share these tokens with anyone! They allow anyone to act as you on GitHub in limited scopes.

To generate a new one, head to your GitHub settings and select Personal Access Tokens.

This process is no longer used in Adafruit CircuitPython repositories. If you want to know how to test your contribution to a CircuitPython repository, please see the GitHub Actions page:

Once there you will see everything that you've already generated an access token for. The tokens themselves aren't ever visible after they are generated. Instead, you can always generate a new token. Delete any tokens you aren't using currently or those that may have accidently been shared.

For now, we'll generate a new token.

Now fill out the description so that you know what the key was used for and can tell in the future if you still need it.

Also tick the box next to public_repo this gives anyone with the token access to all of your public repos so be very careful with the token. We trust Travis CI so its ok to give it to them.

Lastly, scroll down and click Generate Token.

This will take you back to the token list with a new entry in green with the token visible. Click the copy button to copy it to keyboard and switch back to the Travis tab. If you lose the token, you can regenerate a new one by clicking edit.

I've intentionally covered the code with the arrow because tokens should only be shared to those you trust. Never ever commit them to git for example because anyone can see that. If for any reason you believe someone you do not trust has it, either regenerate a new key through Edit or Delete it altogether. I'm regenerating this one after writing this guide even though I covered it. Its ok to be paranoid with these tokens.

Now that we have a code, lets set it up in Travis. Navigate to the build's settings by clicking Settings under More Options. Scroll down to the Environment Variables section and enter it with the name GITHUB_TOKEN and the copied token as the value. Leave Display value in build log OFF otherwise you'll share the token with the world. Lastly, click Add.

After clicking Add we should see a new entry for GITHUB_TOKEN with our token obscured. Don't worry, it saved ok. They are just doing their best to keep it secret too. Time to release!

Also, don't forget to close or refresh the GitHub tab with your token. You don't want to accidently share it in later screenshots. Refreshing will show our description rather than the token.

This process is no longer used in Adafruit CircuitPython repositories. If you want to know how to test your contribution to a CircuitPython repository, please see the GitHub Actions page:

This guide was first published on Jul 31, 2017. It was last updated on Jul 09, 2024.