pretty much there with llama running locally, a few tweaks and we are good to go, drop in to my low rent gonzo AI lab twitch.tv/m88t and I can explain how we could do it...
thanks for the local tip, is it this, https://huggingface.co/blog/stable_diffusion ? Right now, aside from loading models etc, once it's running, you can disconnect the internet plus all open source :)
We can use Stable Diffusion. Huggingface's Diffusers library do a good job in running some versions locally really well.