Flatpak Help Needed
from UNY0N@lemmy.world to linux@lemmy.ml on 23 Sep 2024 20:25
https://lemmy.world/post/20106509

Hi all!

I’m trying to use a local LLM to help me write in an Obsidian.md vault. The local LLM is running through a flatpak called GTP4ALL, which states that it can expose the model through an OpenAI server. The obsidian.md plugin can’t reach the LLM on the specified port, so I’m wondering if the flatpak settings need to be changed to allow this.

(This is all on bazzite, so the obsidian program is a flatpak too)

Anyone have an idea where to start?

#linux

threaded - newest

that_leaflet@lemmy.world on 23 Sep 2024 20:35 next collapse

Flatpak doesn’t care about your ports, they can access them if they have network permission.

UNY0N@lemmy.world on 24 Sep 2024 04:35 collapse

Thank you for the info!

asap@lemmy.world on 24 Sep 2024 10:27 collapse

It’ll be easier to run the LLM in Podman on Bazzite.

UNY0N@lemmy.world on 24 Sep 2024 13:52 collapse

Thanks! I’ll try it out.

Toribor@corndog.social on 27 Sep 2024 22:02 collapse

I’ve not tried GPT4ALL but Ollama combined with Open WebUI is really great for selfhosted LLMs and can run with podman. I’m running Bazzite too and this is what I do.

UNY0N@lemmy.world on 27 Sep 2024 22:49 collapse

I was trying so hard to get GTP4ALL to work inside of obsidian because it has this localdocs feature, where you can feed it documents and it integrates them into itself. I want the model to be aware of the whole vault, and also generate notes based on the vault contents, constantly updating.

It looks like podman desktop + chatGPG obsidian plugin is the way to go through, I’m playing around with it and it looks promising.