Enable offline use of embedding model

[InAnYan]

When JabRef with AI PR is run first, djl will download files in background in order to work with embedding models (I guess the PyTorch backend and embedding model).

JabRef already has some issues that it takes too much time to open, now it will open 5 minutes later more 😢

I think this is only one time job, but it should run in background

[koppor]

The user has to be modified and has to agree to that. ("agree" means that the user can make an informed decision. They should know: What is downloaded, from where, and why)

Assume a secret research company using JabRef. No outside communication allowed. Each outside communication immediately reported to the manager - if not on the "allow-list". Think of agreeing to a privacy policy putting something to the "allow-list". -- Maybe, that helps to understand why JabRef should not "just communicate" to the outside world.

Put differently: Companies have cooperate proxies which JabRef does not use out of the box. Thus, when starting JabRef, there will be network exceptions in this case.

[InAnYan]

Fixed with some shit code. Or maybe it's not that bad

[koppor]

Fixed with some shit code. Or maybe it's not that bad

You can link commits here. Otherwise, it is hard to follow.

Just pasting the commit id should work for linking

[koppor]

JabRef downloads immediately when hitting "I agree".

See also: See #105 (comment).

[koppor]

I think, I reopened, because I thought, I as suer was not informed, what's going on. I checked the help button

It lead me to craft the comment JabRef/user-documentation#491 (comment).

I don't know about the proferences for local providers. Will there be a checkbox for enable chatting in general and then the local providers? - If yes, thenn a downlaod should only be made if external providers are enabled.

I think, with a split to local and remote AI #"providers", and the link to the help, it is OK as is.

Advanced: Add text "This will download embedding models to your machine".

[koppor]

Remember, I write somewhere about cooperate proxies. And machines not connected to the internet. How can an offline installation of JabRef work? How can users get these files?

[koppor]

Since the issue provides some interesting discussions, I keep it and just move it to the Week for offline LLMs.

[InAnYan]

DevCall: introduce online/offline mode

[koppor]

Folder which should maybe be copied from one machine to another: C:\Users\vagrant\.djl.ai.

[koppor]

Setting to try: Use Qubes OS as operating system.

Scroll through https://www.qubes-os.org/screenshots/ to get an idea. Each color represents a separate VM.

[koppor]

At https://forum.image.sc/t/trouble-getting-gpu-to-work-with-instanseg-qupath/102042/26 I saw a nice download dialog. Maybe, we can get this, too?

[InAnYan]

Closing, as moved here: JabRef#12240