Managing my home network with a coding assistant
How I use a coding assistant to inspect the setup, make changes, and check the result.
I’ve been using a coding assistant to help manage my home network. It can inspect the machines, work through a problem, and make a requested change. The useful part is being able to carry an investigation through to a checked result, with the context from earlier work available.
Giving it enough context
The assistant needs to know what the setup is supposed to look like. Which machine runs a service? How is it reached? Should it be public, or only available privately?
I keep a description of the setup and the changes we’ve made. That gives the next investigation a starting point. It also means the assistant can check the part relevant to the current problem instead of rediscovering everything.
That description isn’t a replacement for looking at the machine. It records what was observed and when. When a change depends on the current state, that state still needs to be checked.
Working through a change
The process starts with what I want to achieve. The assistant inspects the relevant service or host, works out a concrete change, and checks the result afterwards. Keeping those observations separate from assumptions makes it easier to see what has actually been established.
The scanner is one example. It involved a driver, a button-triggered workflow, and the hand-off to Paperless. Being able to follow the document across those parts was more useful than treating each failure as an isolated error message.
Some checks can be done automatically. Others need me: putting documents in the scanner, confirming that a device behaves correctly, or deciding that an application should be exposed publicly. The assistant can help prepare those checks, but a command completing successfully doesn’t replace the observation.
The setup behind it
Most home services run on a Proxmox laptop, with Docker applications and Home Assistant kept separate. Administration goes through Tailscale. A small VPS provides a public entry point for the services I’ve chosen to expose.
Those details provide the context, but they aren’t the main point of the workflow. What matters is keeping track of the intended setup, investigating the actual state, and leaving enough information that the next change can start from what we already know.