RLMesh documentation

RLMesh connects models to Gymnasium-style environments, even when they run in different processes or on different machines. The Python package is the main way in, and Rust, C, and C++ APIs are in early access for simulators and inference stacks written in those languages. The managed platform runs evaluations on RLMesh-hosted GPUs for invited organizations.

Start locally

1. Install RLMesh.
2. Serve CartPole and run an evaluation in two local processes.
3. Bring your own environment or your own model.

The quickstart introduces EnvFactory and Model subclasses. These are the recommended starting point for code you will run locally and package for the managed catalog. The repository walkthrough has runnable files.

Run on managed

If your organization has been invited, start with Your First Evaluation. It goes from sign-in to a finished evaluation in the dashboard. To bring your own model or environment from a local container, follow Upload a Model or Environment.

For other tasks, see Connect a Client, Run an Evaluation, or Adapters. The examples and Python API cover runnable recipes and signatures. For supported versions, see Compatibility.