This is an example of using AppImageBuilder to bundle a Python module as an AppImage (only supported on AArch64 for now), and then running it on a robot.

The built AppImage is a binary that runs the Viam module on any Linux AArch64 robot without installing anything else or worrying about the system Python runtime and package dependencies. It is executed via run.sh, and packaged to the Viam Modular Registry as a tar.gz file (see Makefile for details). Note that the AppImage bundling isn't always the best solution for packaging Viam Python modules, but likely the best choice for some use cases.

Use this repository's build process as boilerplate to deploy your own Python Viam module as an AppImage.

Quick tour of this repo

packaging/: the directory containing the Dockerfile that configures the Docker image to run the builder in, the AppImageBuilder.yml config file, and the icon. You must have Docker installed before running the build.
src/: Python module code for the Viam module. Replaceable boilerplate for whatever robot resource your module implements.
Makefile: make commands to build the module in different ways
meta.json: the metadata config file used when uploading/updating module info on the registry
requirements.txt: list of Python packages the module depends on
run.sh: the entrypoint execution script that starts up the module process

Using this repo as boilerplate

If you fork this and want to deploy to a robot, you'll need to change all placeholder names and references to the 'viam' namespace. Follow the fork instructions from the Python example repo.

If you aren't making a camera module, you may want to use a different module repo as the actual Python code boilerplate, such as replacing the contents of src/ Python code. When doing this, make sure to maintain proper Python module structure (i.e. preserving src/__init__.py) to ensure the build doesn't break. Moreover, in this case, be sure to update your meta.json with the updated models and other nomenclature.