GitLab 17.9 allows self-hosting for AI assistants GitLab Duo Chat
With GitLab Duo Self-Hosted, LLMs can now be hosted on-premises or in a private cloud – for higher data protection.
(Image: Tommy Lee Walker / Shutterstock.com)
The GitLab development platform has reached version 17.9. It brings innovations for the use of its AI offerings GitLab Duo Chat and Code Suggestions in terms of hosting and interaction with Visual Studio Code and JetBrains development environments. In addition, older pipelines can now be deleted automatically.
Updates for AI service GitLab Duo
GitLab regularly provides its AI service GitLab Duo with new functions. In the current release, users who use the self-managed Ultimate Edition or GitLab Duo Enterprise benefit from the new GitLab Duo Self-Hosted feature. This allows them to host selected large language models (LLMs) on-premises or in a private cloud to use GitLab Duo Chat and Code Suggestions with more data control and data protection.
Another update concerns the interaction with Visual Studio Code and the JetBrains development environments: Project files can be inserted from these directly into GitLab Duo Chat. This gives the AI assistant a more comprehensive understanding of the code base, which should ensure more powerful AI support while taking the context into account. This innovation is reserved for all paying GitLab users.
Videos by heise
Automatically delete older pipelines
Users of all GitLab editions can now have older pipelines deleted automatically, which was previously only possible via API. The new setting allows you to specify a value that can range from one day to one year. Once this period has expired, GitLab deletes older pipelines. If the field remains empty, no automatic deletion takes place.
(Image:Â GitLab)
In addition to other new features, all GitLab users can now restrict access to GitLab Pages at the group level: group owners can set that all Pages in a group and its subgroups are only visible to project members.
All further information on the new release can be found on the GitLab blog.
(mai)