Thanks for skrl + quick question on future updates

[xiaoxinyezaiya]

Hi @Toni-SM ,
First off, thank you so much for creating and maintaining "skrl"—it’s been an absolute game-changer for my graduate research! As someone just starting with reinforcement learning, this library has made the learning curve so much more manageable. The code is clean, the documentation is clear, and honestly, it’s one of the best RL open-source tools I’ve come across. What I love most is how straightforward and practical it is—even for beginners like me, it never feels overwhelming, and the training results are consistently impressive. You clearly put so much thought and care into it, and it shows.

I noticed the project hasn’t had updates in a few months lately, and I figured you might be swamped with other things—totally get it! I’ve been following its progress closely, though, so I wanted to ask (very politely) if you have any plans for future updates? Specifically, I’ve been curious if there might be expanded support for more multi-agent algorithms, or if RNN capabilities for multi-agent setups might be on the horizon. No rush at all, of course—just curious!

And if you’ve had less bandwidth to maintain it lately, please know that’s completely understandable. Open source projects take an incredible amount of time and energy, and you’ve already built something that helps so many people.

Thank you again for all the work you’ve poured into this. It’s truly a valuable resource, and I’d be thrilled to see it keep growing.

Best regards,

X1aoXin

[Toni-SM]

Hi @xiaoxinyezaiya

First, thank you very much for these kind words.

skrl it is still under active development/maintenance.

In fact there is an new (major) release (branching flow: toni/develop_observation_states -> develop -> main) in process where the main feature is to differentiate between environment observations and states (a concept that is currently mixed in the current implementation). Here, you can see the changelog:

skrl/CHANGELOG.md

Lines 5 to 13 in 4361bdd

	## [2.0.0] - unreleased
	### Added
	- Add `compute_space_limits` space utility to get Gymnasium spaces' limits
	### Changed (breaking changes)
	- Refactor the library to differentiate between environment observations and states (also known as privileged observation)
	### Removed
	- Remove checkpoints/models migration support from other RL libraries

Given the enormous amount of time that has been invested in the new modification, adding new multi-agent algorithms will probably be deferred until version 2.1.

[xiaoxinyezaiya]

Hi @Toni-SM ,​
Thank you so much for taking the time to reply and for sharing such detailed updates on the project's development progress! Seeing that skrl is still under active maintenance and iteration really puts my mind at ease — after all, it’s been an enormous help for my graduate research. I’d had a slight worry about the update pace before, but now all those concerns have been completely laid to rest!​
I also fully understand that the new modifications have required a huge amount of time and effort, so it’s totally reasonable to postpone the multi-agent algorithms to version 2.1. I’m in no rush at all and will wait patiently for future updates. After all, good design needs ample time to refine, and the work you’ve put into optimizing the core concepts will actually build a more solid foundation for future feature expansions.​
Lastly, I want to express my gratitude once more! Maintaining such an easy-to-use yet powerful RL library consistently is truly no small feat. Going forward, I’ll keep following the developments in the development branch, and I’ll try out version 2.0.0 as soon as it’s released. If I have any feedback during use, I’ll communicate it with you promptly.​
Thank you again for all the dedication and hard work you’ve invested in this project!​
Best regards,​
X1aoXin