When the Dask Distributed project wanted to develop a diagnostic interface to allow users to visualize the progress of their distributed computations and identify performance bottlenecks, they chose to use Bokeh because of its ability to define dashboard elements and their streaming update logic in Python.

This talk will highlight the best practices of developing Bokeh dashboard applications, which arose from writing the Dask Distributed UI, while creating a new dashboard application to identify faces using OpenCV. This includes how to: Use the bokeh.models API to organize your dashboard code into discrete elements and update logic Connect to data sources like files, databases, or message queues Use layouts including templates for custom layouts and responsive behavior Add pleasing and consistent styling using themes * Deploy your standalone Bokeh dashboard or embed it within larger applications