I am in the middle of developing an internal web application for my company. The idea is you can log into the app, fetch some data from our databases via an API and display it as a dataframe inside the app. From there, you can select certain columns and plot them against each other in different ways or download the dataframe as a csv.

My original approach was to have a second API running on the server which does all the heavy lifting via pandas. You would first query the database API, receive the results as JSON, then forward that to the analysis API and receive the results of the analysis as JSON and display it in the front end. We have a lot of proprietary analysis code written in python which we obviously want to leverage so rewriting everything in JS is not really an option. My boss however wants to see if we can run the analysis code directly in the browser via something like pyscript.

It seems feasible but just feels…wrong? My boss is concerned that my architecture could lock us into something that isn’t very scalable. He feels we should leverage the processing power of the client machine as much as possible.

I don’t know who is right. What would you do?