For all their faults, LLMs are pretty damn good at basic trouble shooting of Linux. Ideally prepare context for them with installation details. Use CLI client, recommend opencode CLI, plan agent is good to inspect the commands it will plan to run and let’s you inspect and think through what it is doing. Can also ask for clarifications along the way.
Honestly, if you go that route I’d probably recommend using an LLM only for information and asking questions, but run commands yourself and actually read it’s responses. If you don’t understanding something, ask for elaboration. Otherwise, you risk too much automation, letting it set up configurations you don’t have actual understanding of and making a spaghettified mess.
For all their faults, LLMs are pretty damn good at basic trouble shooting of Linux. Ideally prepare context for them with installation details. Use CLI client, recommend opencode CLI, plan agent is good to inspect the commands it will plan to run and let’s you inspect and think through what it is doing. Can also ask for clarifications along the way.
It’s not perfect but very good.
Honestly, if you go that route I’d probably recommend using an LLM only for information and asking questions, but run commands yourself and actually read it’s responses. If you don’t understanding something, ask for elaboration. Otherwise, you risk too much automation, letting it set up configurations you don’t have actual understanding of and making a spaghettified mess.