create a really big + complex app as a demo

[pngwn]

This would be really helpful internally to understand the UX and impact of our changes on large + complex gradio apps.

We probably won't want to deploy this or tests against it due to how expensive it would be to run (in terms of time) but it would be useful as a demo.

This demo should have many components (thousands) and inference functions that do something but nothing complex. It should also have no dependencies or at least no new dependencies. This way it mimics a real app but is easy to run and quick to test against.

[shafiqihtsham]

@abidlabs @pngwn Hello, hope you are doing well. I am very interested in working on this issue. I have read the previous comments left by @abidlabs and just wanted to clarify the complexity of the components we would be adding. Would creating something that uses an external dataset like https://huggingface.co/spaces/gradio/map_airbnb this be too much? Also what about using python libraries such as numpy, pandas etc? Just want to understand what is covered by no dependencies or no new dependencies. Thank you!

[pngwn]

Hello @shafiqihtsham I am very well thank you. I hope you are also doing well.

Using an external Dataset is not an issue as long as it doesn't make starting the app very slow and doesn't require authentication. We will use this app in various testing scenarios so it is important to that it doesn't take too long to start and that it can easily be used locally (as well as in spaces).

Regarding dependencies, we mainly want to avoid any new, complex or heavy dependencies. We are already have quite a few dependencies and list of these are fine to use. Things like numpy, pandas, pillow are all fine. I'd just avoid things like transformers, diffusers, etc.

The main purpose of this app isn't the logic specifically but the components being in lots of different layouts, tabs, lots of components acting as inputs to lots of other components. Different tabs with more of the same etc.

We want it to emulate a complex app (like Stable Diffusion Web UI, etc) but without any of the heavy processing. That is obviously a little vague but outside of that we don't have any strong opinions.