run on linux server

[hollow badge]

I want to run invokeAI on my linux server and publish it on my web page. How can I do it?

[fluid bough]

That’s actually a terrible idea, given that the web ui we have is meant for local use. Because of that, we haven’t done anything to ensure that the web server powering it is secure enough to put out on the Internet

[hollow badge]

I get it, it's sad. So do you know of another open source project that will help me?

[fluid bough]

Specifically for serving over the public Internet? Maybe ask in #🌏invoke-chat - someone in there might have some ideas

[worthy laurel]

I think @hollow badge wants to consume our API to provide a web service

It shouldn't use our web API, only our internal API

[celest egret]

This is doable, the first step is pretty simple: (add --web --host 0.0.0.0 to your invoke.py file):
"...venv/bin/python scripts/invoke.py --free_gpu_mem --web --host 0.0.0.0;;"

but you need a solution for logins, or you would be publicly exposing a very easily abused api.

@hollow badge

Personally, I'm running Nginx Proxy Manager within an Unraid Docker, I added the host and ip to nginx, then restricted the allowed ips to computers in my Tailscale. I already had the infrastructure setup for my NAS, so this was easier than starting from scratch, but theoretically it would work, with or without the Docker (but containers are awesome)