Draft: Add integration tests

[mutlusun]

This MR is based on #282 and should be merged after it.

The goal of this MR was to add first integration tests. I refactored the code on the way to have a cleaner solution for this.

Changes:

• To have a whisper model in the CI to run the integration tests, I added a feature to download certain models into the package directory. This has the advantage that also for the packaging of new version only the download command needs to be used to get the models. Model origin and version (commit hash) is set in the python code. This should prevent the use of other/newer models without specifying them before. To implement this cleanly, I refactored the code related to detect whisper models as well.
• I refactored the code to follow the model-view-controller paradigm (see here for example for dive-in). This makes it easier to separate the code between UI and background logic. It should also make it easier to implement another GUI in the future for noScribe.

I'm sorry for the rather big MR but I think noScribe profits from the integration tests and the started refactorings.

Best

[kaixxx]

Hey @mutlusun, thank you, this looks very reasonable.
I have started this project from a simple script that was extended step by step into a fully fledged application. A fundamental refactoring and modularization of the code is definitely overdue. I will need a little bit of time to take a proper look.

One thing to consider: I plan to extend noScribe in the direction of a client-server-architecture that can also be run on a powerful server in a local network. Several university research institutes have asked for such a feature. It would be great if you could keep that in mind when working on separating front- and backend logic.
I'm not completely sure what the best approach would be for such a client-server architecture. Maybe TCP for piping the communication between front- and backend, similar to the current communication with background processes in noScribe? I have zero experience with client-server architectures.
If you have any ideas, I would be happy to hear them. We could continue the discussion in a separate issue threat.

[gernophil]

I have some experience using apache2 with a reverse proxy with a shiny for python based app like here: https://bescreen.ostendorflab.org
This is the backend: https://github.com/ostendorflab/bescreen. The frontend is a simple Python script that's not publicly available, but I can share it amongst us.

[kaixxx]

@mutlusun @gernophil : Lets continue here: #289

[mutlusun]

@kaixxx Thanks for your feedback!

Please make sure to look at #282 before looking at this MR as changes of both MR appear here at the moment.

[mutlusun]

I will rebase this after #282 is merged.

[kaixxx]

@mutlusun I've seen you've changed this to a draft again, probably for the rebase. Take your time, no hurry. Thank you for all the work.

[mutlusun]

Thanks! Yes, only for the rebase but I think this will take some time as the previous PR and this are a bit overlapping at the moment. I need to clean this up before.