PyScript — Going Truly Serverless

I was very excited when Anaconda announced PyScript and finally got to spend a few hours tinkering. The result is a simple webpage using NMF from Scikit-Learn to perform topic modeling! That’s correct Numpy and Scikit-Learn running in a browser and finding topics from sentences entered into a textarea.

Try NMF | View Source Code

Potential to go truly serverless

So obviously serverless still involves a server somewhere in the cloud but I see PyScript as a way to go truly serverless. Functionality leveraging great Python libraries such as Scikit-Learn, Pandas, Numpy and Matplotlib that I would usually implement within an AWS Lambda, Container or EC2 can now be included in regular HTML (with an additional initial load time). I could for example build a plugin for a no/low code app builder (e.g., a visualisation recommendation system) that runs completely client side.

Caveats

PyScript is still in alpha and it was not all smooth sailing. Here are some issues I encountered:

• There is not a lot of documentation. I mainly looked at the examples to figure out how to trigger Python from HTML and write back to HTML.
• The default stylesheet overwrites default HTML setting eg the H1 tag padding is changed.
• When I placed the textarea within a form tag, each time the button was clicked all libraries were reloaded. I had to remove the form tag.
• Outputting HTML using .innerHtml() caused an error.

More experiments going forward

I’ll be publishing more of my PyScript experiments.