Are there any plans to support a Tensorflow JS version, where any inference can happen automatically in the browser instead of on the server via Python?
This is a really interesting idea. We are currently in the requirements generation phase for the Streamlit plug-in architecture, and I can see how we might be able to enable this functionality quite elegantly. I added your comment to that feature request.
Please feel free to followup on that issue with any more thoughts you have about that feature or details of your use-case.
Also, welcome to the community!