[MVP] Dashboard Apps JupyterHub Extension

[aktech]

Value and/or benefit

The CDSDashboards extension is not compatible with current versions of JupyterHub (it requires JupyterHub < 2.0). People using Nebari won't be able to use CDS Dashboard in the future release, hence there is an urgent need to create a solution for the same. The goal here is to create a replacement for CDS Dashboard. This is a relatively larger project and here we'll define the scope for the MVP (4-6 week half-time):

A JupyterHub Extension to create and share dashboards that supports the following:

• Works with single server JupyterHub
• Supports the following dashboard frameworks (ability to create dashboards):
  • Plotly
  • Streamlit
  • Panel
  • Bokeh
  • Voila
• Lists available dashboards
• Shareable: Dashboard running as the user who shared (No Auth)

Long term goals/questions

• Dashboard - Plotly, streamlit, Panel, Bokeh
• Share to group/access management - not as important
• Dashboard running as user who shared
• Long running dashboards / dashboard restart is important
• Environments management, done manually.
• Adding logs & error messages
• Add debug modes for things like plotly
• Launch modes & bringing down costs.
• idle -culler + restart when accessed
• Avoid 1-1 correspondence between pods and dashboards. Shared/oversubscribed pods.
• Can we launch multiple dashboards on the same pod.
• Log max cpu & max memory used by a dashboard so you can decide how to right size the pod.
• Decide where to store state, how will restarts happen, can we upgrade jhub without losing dashboards. Which dashboards are available, which are running etc.
• Listing of Available Dashboards
• Dashboards accessible to the public without authentication

Anything else?

No response

[aktech]

Some wireframes:

[aktech]

• Add Edit Dashboard
• Conda Environment in Create Dashboard
• Ability to see shared dashboards (?)

[tylerpotts]

@aktech This all looks great! Thanks for writing it up.

[aktech]

This is done now.