@ruegerth No worries Tom, you didn’t step on anyone’s toes
And if my message seemed abrupt, it’s probably my fault (and maybe a little bit the fault of machine translation) — that wasn’t the intention at all.
On the subject, I agree with you more than my first message might have suggested. The README itself says it: the robot part had never run on a real Roborock before. Your question is therefore legitimate: should we publish an integration in this state?
I think yes, but not just any way:
- No publication, no real tests. No developer has all the models of a brand, and Roborock releases several per year. If we wait for someone to test everything, the integration will never be released. Your message shows this well: it’s because it’s published that you were able to try your two robots and find all that.
- However, you’re right on an important point: the user must know what to expect. An integration that hasn’t yet run on real hardware should clearly announce this from its presentation (« beta, testers welcome »), and not just in the README that few people read. The store doesn’t yet allow this to be indicated: it would be a good idea to suggest for Gladys.
- The risk remains limited: at worst, the robot cleans the wrong room or doesn’t react. It’s not a smoke detector. You can uninstall at any time, and each correction comes quickly with a new version.
And that’s exactly why feedback like yours counts so much. Your issue #13 is super clear, with each test, what was expected, and what actually happened. For a developer, that’s the ideal feedback ![]()
What I take from your message:
- the room and mode chosen do not match what the robot does (the most annoying);
- sub-settings (suction power, washing intensity, number of passes, path) are missing, especially on recent models;
- several rooms at once, rather than launching them one after the other;
- mixed translations;
- the widget, which doesn’t always respond.
I have a Saros 20x: if I have time, I’ll retest your tests at home, no promises. But with a second recent model, it could help @Lokkye find where the offset comes from.
Thanks again for the time spent testing, and keep reporting all that! ![]()










