Thanks for this thoughtful write up of your process. I’m increasingly thinking about what context the model has and keeping it as focused as possible - both to reduce token usage, and to ensure it doesn’t have any cruft in it that potentially causes the model to go down an un-useful path. The prompts for this read like what I imagine a conversation with a junior developer would be when handing off a task.

In practice, this is usually clearing the context after quite small changes and the prompting for the next one with just what I think it is going to need. I guess this is ‘context engineering’ although that sounds like too fancy a term for it.

Proxmox on the metal, then every service as a docker container inside an LXC or VM. Proxmox does nice snapshots (to my NAS) making it a breeze to move them from machine to machine or blow away the Proxmox install and reimport them. All the docker compose files are in git, and the things I apply to every LXC/VM (my monitoring endpoint, apt cache setup etc) are all applied with ansible playbooks also in git. All the LXC’s are cloned from a golden image that has my keys, tailscale setup etc.

100% this. And Lenovos and HPs designed for the business market generally are a pleasure to work on (in the hardware sense) if you need, with good manuals and secondhand spare parts.

I’m local first - stuff I’m testing, playing with, or “production” stuff like Jellyfin, Forgeo, AudioBookshelf, Kavita etc etc. Local is faster, more secure, and storage is cheap. But then some of my other stuff that needs 24/7 access from the internet - websites and web apps - they go on the VPS.

I just do one Docker container per LXC. All the convenience of compose, plus those sweet Proxmox snapshots.

Is there a reason not to use Tailscale for this?

When I switched to webdev, I dropped $20 on a system admin Linux course on Udemy. I highly recommended this approach.

Forgejo - actively developed open source. It’s what powers Codeberg. Easy to set up and manage with Docker. I moved to it from Gogs and skipped Gitea after reading about the forks.

It is only resolving for devices in the Tailnet. Kuma is checking they are all up, and this Ansible playbook is checking they have all their updates. I wouldn’t have thought that was an unusual arrangement - and it’s worked perfectly for about a year till about three weeks ago.

Yes, this.

Thanks yes - that’s exactly what I needed.

Thanks - this is exactly what I needed.

Yes - we’re “I’ll let you use my electricity for your computer thing” friends, not “I’m okay with seeing your printer on my home network” friends.

Kavita is for ebooks - it’s not perfect, has some weirdness with series sometimes because of it’s manga heritage.

For me, AudioBookShelf is the clear standout for audio books, and I ended up going with Kavita for ebooks.

I have it in a git repo, broken down by the nodes and vps names. In each of these folders is a mixture of Ansible playbooks, docker compose or just markdown files with the descriptions. Some is random stuff - my VPS allows the export of the cloud firewalls as JSON for instance. All the secrets needed by Ansible are in an Ansible vault, the rest in KeePass.

Yep. Glad he’s got a system that works for him, but as a solo dev I love my Forgejo. I self host it, (so no Trust issues) and if you’ve hosted any other services before, the setup is a simple Docker compose - so I’m not sure I accept the Heavyweight argument either.

Great comprehensive answer. The only thing I might have added (at the risk of confusing things) is that Codeberg is a non-profit, community-led effort that provides Git hosting (with Forgejo), so a sort of open source GitHub

It got enshittified. I went to use it one day and it wouldn’t work without creating an account.

Same - As an Insomnia refugee, I thought “Oh no, not this again” and felt foolish for evangelising it.