Support Panel Webapps in DataIku

MarcSkovMadsen

Panel is a powerful #python #mlapp framework build on top of Bokeh.

A high-level app and dashboarding solution for Python — Panel 0.12.4 documentation (holoviz.org)

Panel can be supported similarly to Bokeh. The Panel server is build on top of Bokeh, therefore it should be very easy to support. For example `bokeh serve` is replaced with `panel serve`.

For some use cases the existing Bokeh Webapp feature of DataIku can be used. For example this one

import numpy as np
import pandas as pd
import panel as pn

pn.extension(sizing_mode='stretch_width')

layout = pn.layout.FlexBox(*(
pn.indicators.Trend(
data={'x': list(range(10)), 'y': np.random.randn(10).cumsum()},
width=150,
height=100,
plot_type=pn.indicators.Trend.param.plot_type.objects[i%4]
) for i in range(28)
))

def stream():
for trend in layout:
trend.stream(
{
'x': [trend.data['x'][-1]+1],
'y': [trend.data['y'][-1]+np.random.randn()]
}, rollover=20)

cb = pn.state.add_periodic_callback(stream, 500)

pn.template.FastListTemplate(
site="Panel",
title="Streaming Trend Indicators",
main=[layout,],
header_background="#428bca"
).servable()

I would expect the panel server to sometimes serve extra panel assets and support other features that the bokeh server does not. Therefore the need for a dedicated Panel option.

CoreyS

Thank you for your feedback @MarcSkovMadsen
! We felt that this would work better as a Product Idea so we moved it here. Thanks again!

MarcSkovMadsen

Thx. Any updates on this feature?

Panels popularity is increasing and now have +800.000 downloads per month across pypi and conda. Would be nice to be able to guide users similarly to other deployment options https://panel.holoviz.org/how_to/deployment/index.html.