Nice project. One pattern I've seen work well with natural language infra tools is pairing them with persistent terminal sessions so you can resume context mid-troubleshoot without re-explaining your network state. The LLM context gets expensive to rebuild. Also worth considering multi-pane layouts when you're running diagnostics alongside the CLI itself - seeing live output next to the agent conversation catches hallucinated commands before they propagate.
No comments yet.