My Home Infra

Hello, good evening, etc. Welcome to the repo where I keep all my home infrastructure.

Yes indeed, everything I have on my home infra is controlled by this repo.

Most of it is Ansible, as it's great for configuring individual machines, but you'll find some Terraform, Kubernetes, and ESPhome configs in there as well.

Credentials are stored in this repo, encrypted. (though I'm slowly moving off onto 1Password credentials)

I access everything though Tailscale, which is an eye-twitchingly annoying dependency. But I keep the integration as light as possible - I can detach my infra from them when they inevitably turn evil.

Installation

Not sure why you'd want to install this, as it's my home infra. But hey, knock yourself out.

Setting up the control plane

# Clone repo
git clone [email protected]:awfulwoman/infra.git

# Install supporting Python packages
cd infra
python -m venv .venv
source .venv/bin/activate
pip3 install -r requirements.txt

# To activate auto loading of resources when entering the directory
apt install direnv -y # Ubuntu
brew install direnv # MacOS
direnv allow .

LLM Disclaimer

So, you've spooted that I'm using Claude Code to write portions of this, eh? (add me to the AI Dirty List! (no, don't - people who make lists of individuals like that are cunts)).

Let's be very clear: the push for "AI" is a wanking contest between Silicon Valley pricks and their techbro worshippers. It rides rampant over creativity, art, and all things that require a soul.

But the technology does have uses, epsecially in the non-creative space of everyday coding. No, not the original stuff that happens at the low level of the development ecosystem. But at the high end, where code is just a tool to achieve tangible goals... yes, I think there it's very useful.

Claude seems better (this is a relative statement) than the other AI shite companies. The session limitations, even on the paid plans, have the (unintended) effect of dampening the addiction cycle that other AIs seem intent on promoting. Obviously they're still bastards - they're a capitalist hypercorp, after all. I'm under no illusions that they have my interests at heart - no corporation under late-stage capitalism possibly can.

But there's the rub, ain't it? I can use a tool and still understand how damn dangerous it is. If I use a chainsaw I can acknowledge that it's useful, while at the same time understanding that it can severely fuck things up if handed out at a kiddies birthday party.

What was my point? Oh, something useful, I'm sure. But nevertheless, Claude has proved useful for writing informative and detailed commit messages, and has forced me to start documenting everything here.

I'm using Claude as progressive enhancement of skills. I will not use Claude unless I understand the underlying codebase and technology first. Claude is useful for doing arduous tasks that I would not otherwise find the executive function to engage in. I assume at all times that CLaude could disappear in a heartbeat, and that I could lose it as a tool (for that is all an LLM is: a useful tool).

And finally, this will be an unpopular view: Claude is an accessibility aid. I could not otherwise do the stuff I've done with it due to ADHD and dyslexia. Go look at the commit log before 20th December 2025 if you don't believe me.

Anyway, laters.