Dear (potential) #zlink users: if there are any breaking changes you think are needed, now would be a very good time to report them. We recently bumped the semver for a few essential fixes and I'd really like to avoid doing this too often.
Please RT.
Here is a screenshot of ai-playground where an AI agent is running on the left in a #systemd container, but cargo is running outside of the container on the host through the gatekeeper. I use the #varlink streaming protocol to stream stdout/stderr of the process executed on the host. It was pretty easy turning on streaming using #zlink
Learn more at:
https://gitlab.com/cryptomilk/ai-playground
I've created yet another AI agent sandbox and just released 0.1.0. I did it differently than the others. It is called AI Playground.
Features:
* Lightweight containers (systemd-nspawn)
* Git worktree support
* Easy to set up and keep up to date (it is a chroot)
* Host command gatekeeper - lets you approve or deny host commands from inside the container (varlink)
* Multiple AI tools supported
* Contributions welcome
Just rolled out the first release of zlink, a new async-first Varlink Rust crate:
https://crates.io/crates/zlink
https://docs.rs/zlink/latest/zlink/
This release focuses on the client-side of things more but you can already use it to write services as well. A high-level service API is planned for the near future.
Also provided is a codegen utility, zlink-codegen (duh!) that generates client-side proxy trait and auxiliary types for you:
I gave a lightning talk at @guadec on the #zlink work, I've been doing:
Zeeshan Ali Khan 
๐บ๐ฆ
Andreas Schneider
