Code Clinic | Hosting your ChatGPT Clone on Streamlit Cloud (and local)

When you want to show your agent to the world but your front-end dev skills suck

Intro

You have developed this really amazing app that you want to show the world.

Now what? Many people, myself included, have terrible front-end skills. What to do in this situation?

Since I didn’t know either, I asked Reddits Langchain community to give their advice. The most recommended solution was streamlit ahead of gradio, huggingface spaces, and chainlit. So I figured, streamlit is the way to go.

Streamlit

As I learned, Streamlit Cloud is a managed platform for deploying and managing streamlit apps. This makes it easy to deploy my apps online and is based on an open-source framework. Streamlit Cloud is a commercial product there is a free tier — that’s the one I used for this.

The downside is that it’s built on top of Google Cloud Platform. While that might give some benefit in terms of scalability and reliability (to a certain extent), one can be assured that customer service for the cloud components doesn’t exist and you can’t be certain if important components are disabled on a whim.

Ergo, Streamlit Cloud seemed to be easy enough to do the job.

How does Streamlit work?

For me, the main feature is that Streamlit converts a Python script into a web page. The biggest gotcha for me was that Streamlit reruns the Python script that defines the app every time the user (or agent) interacts with a widget.

But I am getting ahead of myself again.

If you plan to host your app on Streamlit as well, I would recommend first trying it out in a local implementation. Debugging online is really really cumbersome.

Therefore you need to install Streamlit locally. I assume that you have knowledge of Python and are not on Windows.

Let’s start building