Predicting what a compiler outputs is quiet hard, especially if you enable optimization flags or get near any undefined behavior. But I think you’re right to say that a compiler is more predictable than any LLM. I don’t get the point why you should be unable to audit the output of an LLM? If its used as a tool like described the user should have the knowlage to audit any output

Even though this kind of proofs your point as this script is really long. I just wanted to recommend the community script install source as a good starting point for installing more complex software baremetal. immich-install-script

You could run it in LXC directly with [community-scripts] (https://community-scripts.org/scripts/immich). I have it running like this for close to a year and am really happy with it. It provides you with a update script that you would use instead of a docker pull

When you say immich has build in tooling for google takeout, do you mean [immich-go] (https://github.com/simulot/immich-go) or is there another, possibly better, way?