/
A UKRDC Stats dashboard

A UKRDC Stats dashboard

    It’s fairly easy to embed a link to a dashboard on our website, as is done in the following ticket:

    https://renalregistry.atlassian.net/browse/OPS-8

     

    Then it would be a case of finding the easiest way of serving up a link.

    Realistically if we want to replace tableau the task is one of developing and hosting a simple app. The options (see ai slop in table below for more detailed comparisons), are summed up by the following:

    1. Super simple jupyter type dashboard (with option of using some thing like voila to publish in a form that looks more like an app)

    2. Panel/streamlit/dash etc … Dashboard development tools with tradeoff between flexibility/customisability and difficulty

    3. Python web framework like flask which allows complete flexability but is a bit more advanced

    4. Non python web framework vue or other

    If I (pm) was tasked with development without any other considerations I would probably go for either 1 or 3 depending on the audience and level of polish.

    Tool

    Embedding Support

    Okta/SSO Support

    Multi-User Handling

    Production Readiness

    UI Flexibility

    Notes

    Tool

    Embedding Support

    Okta/SSO Support

    Multi-User Handling

    Production Readiness

    UI Flexibility

    Notes

    JupyterHub

    Not designed for embedding (separate hub URL)

    Yes (via OAuthenticator, Okta supported)

    Per-user isolated kernels

    Mature, scales via k8s/Docker

    Limited (notebooks + widgets)

    Heavyweight; best for multi-user notebook environments

    Dash (Python)

    <iframe> or Django/Flask integration

    No native; add reverse proxy or dash-okta-auth

    Shared app instance, session callbacks

    Scales with Gunicorn/k8s; Enterprise adds features

    High (React/HTML under hood)

    Great for custom dashboards; enterprise features gated behind Dash Enterprise

    Streamlit (Python)

    Easy iframe embedding (?embed=true)

    No native; SSO via proxy (e.g. Okta/NGINX/Auth0)

    Each user has session state

    Lightweight, moderate scaling

    Low–Medium (simple layouts, limited theming)

    Very fast to build but less enterprise-grade

    Panel (Python)

    Embed via server_document() script

    Yes (built-in OAuth2, supports Okta)

    Multiple processes (num_procs), sessions per user

    Production-ready (Bokeh server backend)

    High (templates, layouts, multiple viz libs)

    Flexible; strongest self-hosted Okta story besides JupyterHub

    Bokeh Server (Python)

    Embed via script (server_document)

    No native; must wrap with proxy/auth layer

    Each browser session = new Bokeh doc

    Stable, can scale with multiple servers

    High (low-level HTML/JS control)

    More boilerplate; great for custom interactive plots

    Voilà (Python)

    Embed via iframe

    No native; usually via JupyterHub/OAuthenticator

    Per-user notebook execution

    Simple to run; scales under Hub

    Medium (widgets, custom nbconvert templates)

    Easiest if dashboards already exist as notebooks

    R Shiny (R)

    Easy iframe embedding or Shiny Server Pro integration

    Shiny Server Pro supports SAML/OIDC (Okta included)

    Each user session runs separately

    Mature; scales with Shiny Server Pro or Kubernetes

    Medium–High (UI via R + HTML templates)

    Best for R-centric teams; Pro adds enterprise auth and scaling

    Superset (Python/SQL)

    Embeddable dashboards via iframes/JWT auth

    Yes (OAuth2/OIDC including Okta)

    Multi-user by design (role-based access)

    Production-grade BI platform

    Medium (prebuilt BI UI, not fully custom)

    Good for SQL-driven analytics dashboards, less flexible than custom code

    Metabase (Java)

    Embeddable dashboards (signed JWT for secure embed)

    Yes (SAML/OIDC incl. Okta)

    Multi-user, role-based access

    Enterprise-ready

    Low–Medium (limited customization)

    Great for quick self-serve BI; more “BI tool” than app framework

    Grafana (Go)

    Embeddable panels/iframes, signed URLs

    Yes (SAML/OIDC incl. Okta)

    Multi-user with fine-grained roles

     

     

     

    Flask

    Easy to embed

    Yes (Authlib, Flask-Login, Flask-OIDC, reverse proxy with Okta/OIDC)

    Session-based users; scales with standard WSGI deployment

    Highly production-ready (Gunicorn/uWSGI, Docker, Kubernetes)

    Very High (full control over HTML, CSS, JavaScript, React, Vue, etc.)

    Best when building a bespoke analytics application that integrates with Drupal rather than relying on a dashboard framework. More development effort, but offers maximum flexibility and straightforward integration with existing Drupal authentication and services.