[lmeyerov]

RE:Streaming, Graphistry wrote much of the Apache Arrow JS lib to optimize for this case of streaming GPU DBs -> GPU browser frontends, you can see a pretty striking before/after here: https://www.graphistry.com/blog/experiencing-the-future-of-g... . JS networking devs can handwrite that stuff.. but it stinks, and this way, everyone benefits. I recall Vega folks had some fun experiments with GPU DB -> CPU frontends here (ended up part of an HCI or VLDB paper?), not sure what happened since. Our production focus on that layer is now more on scaling up node js cpu+gpu code <> pydata GPU code (Nvidia RAPIDS).

For OmniSci folks: Cool to see Jupyter embedded! Our more advanced users love this path, but we find versioning & team collab tricky in big enterprise settings, so every additional company and (hopefully) contributor here should help. Getting there :)