Releasing to PyPi¶

Usually the simulation is run by cloning the code locally, which works well if you're changing code within the repo itself. However, if you're looking to use Stretch Mujoco as a library within your own codebase, it can be convenient to be able to pip install the project. This guide covers how you can release new versions of the project. Users can install the library by adding "hello-robot-stretch-mujoco" to their project's dependencies or using:

pip install hello-robot-stretch-mujoco

Steps¶

Merge your code changes. Update version in pyproject.toml. We're using semantic versioning.
Update the changelog to describe what's new in this version.
Merge the above changes in a release commit. Tag this commit as v<version> (e.g. v0.5.0). Your tag should show up here: https://github.com/hello-robot/stretch_mujoco/tags
```
git add .
git commit -m "Release commit v0.5.0"
git tag -a v0.5.0 -m "Short description of release"
git push
```
Run uv build. It generates .whl and .tar.gz files in the dist/ folder. Open the tar file in an file explorer to confirm it contains the modules and assets you want included.
Run uv publish. You will need an token for PyPi to accept the release.
For larger releases, consider creating a new Github Release and posting to the forum.

Using the library¶

Getting a specific version¶

You can get a specific version e.g. "v0.x.y" by adding "hello-robot-stretch-mujoco==0.x.y" to your dependencies (e.g. using uv add hello-robot-stretch-mujoco==0.x.y if using UV) or using:

pip install hello-robot-stretch-mujoco==0.x.y

Importing the modules¶

You can import modules from Stretch Mujoco using:

from stretch_mujoco.mujoco_server import MujocoServer, MujocoServerProxies

Accessing the XMLs¶

The XML models ship with the PyPi package. You can import them using:

import importlib.resources

scene_xml = importlib.resources.files("stretch_mujoco") / "models" / "scene.xml"
server = MujocoServer(scene_xml_path=str(scene_xml), ...)

What won't work¶

Robocasa won't work since it's not included with the PyPi release. Using the Robocasa assets or the stretch_mujoco.robocasa_gen module currently requires cloning the repo locally and pulling in the submodules.