Sorry I know you probably don’t want another tip from me, but the post did include the agent directly using the docker daemon, which runs as root typically. Because you didn’t mention running rootless docker or podman, your sudoers file probably allows the agent full access to root instead of preventing it.
It seems the main use case is restarting docker containers, why not use the built-in healthcheck feature of docker? The automatic backup and upgrade are also confusing to me, operating systems come with that built in. I just don’t quite understand the point of replacing existing deterministic systems with a natural language interface, I would have trouble believing the logs at face value.
Edit: also your handling of current-alert.txt is a perfect example of a race condition, another potential source of indeterminism. An alert could be missed if the file is overwritten before being handled.
LLMs are inextricably tied to nvidia gpus. Local or cloud, the technology exists to help the shovel salespeople. The gold diggers, everyone this tag is supposed to segregate, have been misled by corporations. Without their lies, and a pliant media, this tag would be unnecessary, and llms would be rolled out in a more limited and responsible way. To promote different uses of gold during a gold rush is going to inflate the bubble and enrich the rich, unless it is properly contextualized. Technology does not exist in a void, pretending it does digs us a deeper hole.