The platform team has been working on some release and deployment process improvements. To make our applications more serviceable, we would like to create a shared release process that: 1. Automatically generates a changelog, so that it is easy to understand what changes are included in a release 2. Creates SemVer tags on each release, so that it is easy to locate the code included in a release and compare two different versions. 3. Is as consistent as is reasonably possible across our projects 4. Has no key person dependencies a. In other words, it should be easy enough to follow that any maintainer could cut a release, and releases shouldn't be blocked by the absence of a single person. I've put together this example repository which demonstrates a simple release process built with `release-please`: https://github.com/OpenHands/release-please-demo. The readme has a detailed walkthrough. It's very similar to the pipeline already implemented in software-agent-sdk, but with some additional automation. I'd like to get some feedback on this process before generalizing it so that we can share it across many repos, and then begin working on propagating it everywhere. Something the platform team already discussed is that for some of our projects like OpenHands Enterprise, we create release branches before every release so that we have some time to perform manual validation without picking up additional changes from

main

. I've added a section on how that would work to the readme, but would like to encourage releasing from

main

as the default. I think the extra complexity it introduces will discourage individuals from cutting a release, which ultimately harms goal #4 from above.

nod Yes, Xingyao has been working a lot to integrate and automate the necessary tests for a release. The regular agent knows details of the process, the reviewer agent knows details of the process too.

All of this basically sounds good to me! Thanks for the work on it!

LGTM as well!

LGTM!