> For the complete documentation index, see [llms.txt](https://roar.gitbook.io/roar_py_rl-documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://roar.gitbook.io/roar_py_rl-documentation/code-structure.md).

# Code Structure

The `ROAR_PY_RL` repository contains two Python packages, `roar_py_rl`, and `roar_py_rl_carla`. 

It also includes a `training` folder that resides with training scripts, environment utilities, and utilities to debug the environment dynamics.

The `roar_py_rl` package contains base abstract `gymnasium.Env` classes that can be extended, providing a basic environment skeleton that can be implemented with a few lines of environment logic code.

The `roar_py_rl_carla` package contains an implemented `RoarRLCarlaSimEnv` that uses the Carla simulator as the underlying simulator. You can use it directly

{% hint style="info" %}
Note that due to gymnasium interface not supporting async traits, we have to `await` for `RoarPyActor`'s `receive_observation()` and `apply_action()` inside the gymnasium's synchronous `step`, `reset` functions. \
Therefore every Python code that uses `ROAR_PY_RL`'s environments should import `next_asyncio` and call `next_asyncio.apply()`
{% endhint %}