Installation

Install the Python package:

pip install "rlmesh[gymnasium,numpy]"

RLMesh supports Python 3.10 and newer. Start with Gymnasium for environments and the NumPy backend for examples and notebooks, then follow the Quickstart.

Optional Extras

Install only the extras you need:

pip install rlmesh
pip install "rlmesh[numpy]"
pip install "rlmesh[gymnasium]"
pip install "rlmesh[torch]"
pip install "rlmesh[jax]"
pip install "rlmesh[hf]"

Pick gymnasium when serving a Gymnasium environment, or gym for a classic Gym environment. torch decodes client-side values as Torch tensors; jax decodes them as JAX arrays; numpy decodes them as arrays. hf adds host-side, container-less resolution of hf:// model weights and EnvHub sources; in a sandbox the container fetches them for you. See Framework Backends for how the backend extras change value decoding.

Repository Examples

The runnable examples are in the RLMesh source repository, not in the installed package. Clone it and run the example commands from its root:

git clone https://github.com/ArenaX-Labs/rlmesh.git
cd rlmesh

The examples use uv; install it with python -m pip install uv if needed. The CartPole walkthrough gives you the first uv run commands, which create the project environment when you run them. You can also browse the examples for another setup. Sandbox examples also need Docker. The in-repository examples share this environment; heavier demos with their own lockfiles use separate setup steps.