Publish a web application for viewing data created with Streamlit on heroku

Overview

-Last time, you can now create web apps using Streamlit ――Because it's a big deal, let's publish it --So what you made → Chocolate Ball Viewer

Premise

--Create an account on heroku (free OK) --I have a github account

App creation

Create an app by referring to here. This article does not touch on Streamlit notation.

$ streamlit run [python-file]

Now, make sure that the app is created as expected.

Deploy to heroku

This is the main part of this article. The outline of the procedure is as follows.

1. Prepare the files needed for heroku deployment
2. Create an app on heroku
3. Deploy
4. Let's check the automatic deployment

Preparation of necessary files

You need the following two files to publish your python app on heroku.

• Procfile --Write an execution command
• requirements.txt --Write down the necessary libraries, etc.

This is enough, but let's create a script to create a streamlit configuration file. It specifies the port number and so on. (It is OK to specify it as a startup option in Procfile)

• setup.sh

Someone has put these together and created a template for publishing the Streamlit app on heroku. https://github.com/patryk-oleniuk/streamlit-heroku-template

This time I made it like this. I almost copied it from streamlit-heroku-template.

Procfile

web: sh setup.sh && streamlit run src/choco_view.py

requirements.txt (It is safer to fix the version)

pylint
pandas
scipy
matplotlib
seaborn
streamlit

setup.sh

mkdir -p ~/.streamlit/
echo "[server]
headless = true
port = $PORT
enableCORS = false
" > ~/.streamlit/config.toml

Put these together and push them to github. Private repositories are also OK.

Create an app on heroku

You can also create heroku apps with commands using the CLI (the heroku tutorial is the CLI), but if you want to link with the github repository, it's easy from the web management screen. This time I will create it from the Web.

Set the service name from "Create New App" on the upper right and click "Create app".

only this

Deploy

When you create, you will be able to select the deployment method. This time, select the "GitHub" tab and specify the repository to be linked.

If it can be linked, it will transition to the deploy screen.

Here, if you press "Enable Automatic Deploys" to enable automatic deployment, any updates will be automatically deployed to the specified branch of the linked GitHub repository.

When you're ready, press "Deploy Branch" from the Manual Deploy menu to start deploying.

Confirm automatic deployment

Let's update the data and push it to github.

If you look at the management screen of heroku, you can see that it is deployed (fun).

did it

I am measuring chocolate ball data as a hobby, so I released an app that visualizes the measurement data (see Chocolate Ball Statistics for details). https://chocolate-view.herokuapp.com/

Flavor can be selected

Simulation of the number of purchases required to hit the silver angel

end

reference

• Show your ML Project to the Internet in Minutes
• streamlit-heroku-template -I made a data viewer for chocolate ball measurement data