Hey all oh waveq For the upcoming beta release, here is the process we are going to follow: • We will use a release candidate branch for the beta, which @fierce-account-55183 will post once he has it cut •

main

will remain the primary branch • Merge bug fixes into

main

after testing/validation, then cherry-pick the approved mvp ones into the release candidate branch. • Test fixes on release branch • Please refrain from merging large, non-MVP changes to

main

, that will help will cherry picking bug fixes. Please reach out with questions!

Sorry, aren’t we using this? https://openhands-ai.slack.com/archives/C06QT0AGY4W/p1780330774264549?thread_ts=1780328572.151489&channel=C06QT0AGY4W&message_ts=1780330774.264549

We were tagging off main until now - we are moving to explicit release branches

Hello @hundreds-twilight-32833, @ancient-car-4311, @fierce-account-55183, I am planning to merge @breezy-kilobyte-39430's pull request: Design updates: navigation polish, modal tokens, and git control bar fixes The pull request includes changes across 81 files, the majority of which are styling-related updates. Before proceeding, I wanted to check whether anyone has any concerns or objections to the merge. Thank you very much! thankyou

What level of confidence do you have in the fix?

I have run the pull request locally and verified the changes, and everything looks good on my end. If you have the time, I would appreciate a second review to help validate the updates. That said, if the team would prefer to prioritize stability at the moment, I am also comfortable postponing the merge. Since most of the changes are styling-related, there is no immediate urgency. I am happy to follow the team's decision thankyou

Given it's just style updates, I'm okay with it

One thing I see is some styling for plan mode in there. Does the conversation-store have both code and plan mode still in them @millions-afternoon-47582

@hundreds-twilight-32833 We support the planning agent when connecting to the OpenHands Cloud backend thankyou

Good point! So not an issue then 🙂

@millions-afternoon-47582 sorry, just for clarity, do you mean the planning agent is not supported in local/remote backends? Any reason why not?

Hello @calm-artist-7206, 1. At the moment, we do not support the planning agent when connecting to a local backend. We can refer to this Slack thread for additional context and information. 2. We do support the planning agent when connecting to the OpenHands Cloud backend. Thank you very much! thankyou

Ah, please correct me if wrong, you mean that the app-server was doing quite a bit of plumbing here 🤔 Do we have an issue to look into alternatives for wiring in the planning agent? I guess for purely local backend that… seems easy enough, not sure for docker

Ah, please correct me if wrong, you mean that the app-server was doing quite a bit of plumbing here 🤔

Yes, that's correct. The current support for the planning agent is closely tied to the sandbox architecture and the app-server implementation.

Do we have an issue to look into alternatives for wiring in the planning agent?

I haven't checked this yet, but we may not currently have a GitHub issue tracking this feature. If one doesn't already exist, I'll create it today so we can track it properly.

I guess for purely local backend that… seems easy enough, not sure for docker

Yes, I don't think it would be particularly difficult to support the planning agent when connecting to a local backend. However, our current focus is on delivering the MVP, so we've prioritized other work for now. Support for this feature is on our roadmap thankyou

Sorry, is it? I don’t see it on the roadmap? https://github.com/orgs/OpenHands/projects/4

My fault! I thought I'd made an issue and added it there

Thank you! And please don’t worry about that. Just for clarity, what I worry about is a bit different actually, it’s that this kind of thing seems to me a bit risky perhaps: Let's say we say:

“in canvas, the planner agent is only available on the cloud backend”

It seems to me that we mean like

“it wasn’t totally obvious how to design it in the chaos and hurry to get a stable canvas, so we left it a bit in the backburner, but everyone is welcome to pick it up and integrate it”

But idk, people may hear

“we stripped down the OSS GUI and made the planner Saas-only”

which, you know, is inaccurate, but it’s dynamite I would really, really love to avoid.