Sitemap & RSS Feed Tags

Doing web without JavaScript, HTML nor CSS for data science

I’ve recently played with a tool named Streamlit. I wanted to share this experience with you.

If you want to build a simple web page to expose your models using only Python, this library is for you.

Streamlit was launched the first of october in 2019. You can get more information in the launch blog post.

Here is the traditional hello world:

import streamlit as st

st.write("Hello world!")

With that, you get this:

Hello world with Streamlit

To dive deeper in this tool, I recommend the official website.

My goal with this article is telling you the benefits and disadvantages I found after using it.

Benefits:

  • Useful when you want to build a simple page quickly.
  • Interesting when you want to expose a simple prediction.
  • With Streamlit, there’s no need to know HTML, CSS, etc. You only have to know Python.
  • Many graphs to represent data.
  • A large and active community.

This tool is built for data scientists and data science. It is helpful for simple pages, demo pages and simple predictions.

Disadvantages:

  • As always, these kinds of tools are not very flexible. This is not the purpose though.
  • For instance, it’s difficult to get different forms in one page. You can find your way with checkboxes instead of submit button. But it’s a bit weird.
  • It’s also difficult to get different pages in one application. The option st.cache can save your life. But it’s not enough to get a complete application.

Conclusion

Streamlit was designed for data applications. It’s not a perfect tool but a very promising one in my opinion.