Kushal Kolar

Now possible with juptyer_rfb too: https://github.com/vispy/jupyter_rfb/pull/84

Actually the better way to do this might be to put it on dockerhub with a few variations: 1. Barebones, only with `wgpu` 2. With `fastplotlib` installed using venv 3....

I did the PR for mcorr so it uses the env var: https://github.com/flatironinstitute/CaImAn/pull/978 For CNMF you can already save the hdf file to whereever you want. Longterm we should have...

Just a thought on (1), the shared memory interface is pretty low level and I would rather use zarr. Also, I'm not sure that (1) would solve the memory size...

Remote frame buffer is different from how bokeh shows stuff, they are completely different things. And it was jupyter_rfb that added vscode support recently, not the other way around.

this also does the same thing but needs to be exposed publicly with a similar API as the non-collection features, perhaps `graphic[:].present(), or graphic[:].present[slice]()` ``` [p() for p in temporal_graphic[:].present._feature_instances]...

The mcorr file naming stuff was fixed, I had to update mesmerize-core for it as well a few months ago. pretty sure it was this release: https://github.com/flatironinstitute/CaImAn/releases/tag/v1.9.12

I was going to add this feature too until I saw this, would be great if it could be merged in the next pypi release :)

it's been a while since I've used kshape clustering but this is probably dataset specific, you could try iterating through different seeds for the centroids

You guys are epic and so fast! With pointer events I can already see how it can simplify our [drag PR](https://github.com/fastplotlib/fastplotlib/pull/413) ! I didn't realize we can just do things...