panel VS dockerized-jupyter-notebook

panel – data exploration & web app framework for Python

https://panel.holoviz.org/ It's a web app framework for Python similar to what Dash does for plotly. It plays nicely with bokeh visuals and I think the front-end is built using bokeh css elements.

I'm following the Panel FastAPI example here: https://github.com/holoviz/panel/blob/main/examples/apps/fastApi/main.py

If you want to build Python dashboards, look at the solara (react-style lib, https://solara.dev/) and panel (https://panel.holoviz.org/).

panel

My suggestion is https://panel.holoviz.org/

Fully open sourced, makes it easy to make reactive apps with small changes, can even configured as a graphical REPL.

I am doing something like this in a [panel](https://panel.holoviz.org/) dashboard, which I am currently converting to nicegui. Maybe I can provide an example in some days.

Ill have to check it out and see how it compares to voilà and holoviz panel. What I like about Holoviz panel is you can create a data web app from code that resides in a notebook or create a completely standalone app from just plain py scripts, and it supports many different visualization backends. I have found it to be the more flexible and generalizable data web app framework among the others I have come across (like Voilà, Dash, Plotly, and Streamlit).

Like the previous three alternatives, Panel is an open-source Python library for creating interactive dashboard web apps. Panel is extremely flexible, allowing you to use any plotting library you like. Like Gradio but unlike Streamlit, you can use Panel in Jupyter notebooks. Panel dashboards can also be deployed as standalone web apps, but like Plotly Dash, you'll need to set up a server to deploy it yourself.

your task is very very broad

you mention you don't want to deal with AWS, if it's because of ad-hoc installation concerns and nothing else you can just run your notebooks in ready-made solutions like Google Colab, or Jupyter-book in Github ( https://github.com/executablebooks/jupyter-book ))

that would cover a lot of use cases right away without next to no learning curve

If you don't want to deal with AWS or similar, in that case:

- if it's a static notebook then you can obviously render it and serve the web content (might seem obvious but needs to be considered)

- if it's dynamic but has light hardware requirements, you can try jupyterlite which runs in the browser and should do a pyodine (webassembly CPython kernel) can do: https://jupyterlite.readthedocs.io/en/latest/try/lab/

- otherwise, you can try exposing a dockerised jupyter env ( as in https://github.com/MKAbuMattar/dockerized-jupyter-notebook/b... ) or even better a nixified one ( https://github.com/tweag/jupyenv )

there might be other approaches I'm missing, but I think that's pretty much it that doesn't entail some proprietary solution or an ad-hoc installation as you've been doing