pplonski86

Web App is using two open-source packages that I've created:

- MLJAR AutoML - Python package for AutoML on tabular data https://github.com/mljar/mljar-supervised

- Mercury - framework for converting Jupyter Notebooks into Web App https://github.com/mljar/mercury

You can run Web App locally. What is more, you can adjust notebook's code for your needs. For example, you can set different validation strategies or evalutaion metrics or longer training times. The notebooks in the repo are good starting point for you to develop more advanced apps.

Example scenario: you are medicine doctor and would like to research a disease. You have csv data. You train ml pipeline to check which features have predictive power.

We have many researchers that are using our open source python package and to make it even easier to use, we created web app with gui that you can run locally.

It is possible. Please take a look at Jupyter Real Time Collaboration https://jupyterlab.readthedocs.io/en/stable/user/rtc.html

Thank you! I've fixed deployment box.

You can publish multiple notebooks on 1 web service. There is no limit. We provide default view where all notebooks are displayed as cards with clickable title. Please check example in the documentation https://runmercury.com/docs/multiple-notebooks/.

Mercury was created for Jupyter Notebook. You don't need to rewrite your notebook to Python script to use web framework (Stremlit).

Mercury is for quick demos. It is not a full web framework. The simplicity of the framework is very important for us.

We hope to extend widgets offering in the near future.

You are right. We are working on Mercury Cloud service where you can just upload notebook to deploy it. We have free and paid plans. But you can also self-host Mercury with docker-compose.

That was my first website in Next.js framework ... sorry! I've used nextra.site for creating docs.

Main differences to Voila:

- Mercury has no callbacks - we automatically re-execute cells below updated widget,

- Mercury has no layout widgets, all input widgets are always in the left sidebar and output content appear in the same order as in notebook.

Thanks to above decisions you don't need to rewrite your notebook to fit UI framework paradigms.

We are working on open-source web framework Mercury that converts Python notebooks to Web Apps.

It is very opinionated:

- it has no callbacks - we automatically re-execute cells below updated widget

- it has no layout widgets, all input widgets are always in the left sidebar

Thanks to above decisions you don't need to change notebook's code to have web app and fit to the framework.

The simplicity of the framework is very important to us. We also care about deployment simplicity. That's why we created a shared hosting service called Mercury Cloud. You can deploy notebook by uploading a file.

The GitHub repository https://github.com/mljar/mercury

Documentation https://RunMercury.com/docs/

Mercury Cloud https://cloud.runmercury.com

What is advantage of Mercury over Streamlit? You don't need to rewrite your code into Python script. You can serve notebook with its charts, dataframes and Markdown. If you want you can show/hide code. You can share multiple notebooks (websites). There is authentication and PDF export.

What is advantage over Viola? Mercury offers cloud service for deployment. You just upload notebook to make a website. That's it.

What is advantage over Gradio? Gradio is a framework for show casing ML models. You provide input for ML model with widgets and compute the prediction (output). Mercury can be used for show casing ML models, data science reports, dashboards and even scientific presentations (with slides recomputing after widget update)

Im working on Mercury framework to solve this problem. Notebooks can be static or dynamic. You can hide code in notebook and show only outputs. You can easily add authentication.

The framework repository at GitHu https://github.com/mljar/mercury

Im working also on cloud offering so you will upload notebooks and set domain. It is available in alpha at https://cloud.runmercury.com

Clickable links:

- Our website: https://RunMercury.com

- Our GitHub: https://github.com/mljar/mercury

We just added authentication to Mercury. You can share Jupyter Notebooks behind login. You manage users with Admin Panel.

Congrats on launch! Do you have examples of monorepos, for example React with TypeScript and Django with Python?

Mercury will be open source. I plan to create a cloud service for easy deployment. There will be free and paid tier. Mercury is AGPLv3. You can buy commercial friendly license as well.

We released Mercury V2 this week. The V1 was using YAML to generate widgets. The V2 is using Python to add widgets in the notebook. Widgets can be used during development and in web app.

There is simple cell re-execution after widget update.

Mercury GitHub repository https://github.com/mljar/mercury

I have desk with treadmill. It is not attached. So I can use desk with normal chair as well. I'd recommend it.

I'm working on an open-source framework for converting Jupyter Notebook to interactive web applications. I was using free Heroku dynos for demo app deployments. I have 19 demo applications running (example https://automated-pdf-reports.herokuapp.com/)

All apps use 1000 free dyno hours in ~25 days, so at the end of the month, Heroku switched them off.

I didn't find any free alternative for my users and me. I'm thinking about building my own service where users can upload notebooks and publish them as web applications.

Great story. Thank you for sharing. Did you have programming experience before starting a startup?

I'm working on an automatic designer. You just focus on the content and have instant preview how your website will look like in several different templates/designs.

There is Genetic Algorithm that can mutate styles of the current templates (that you liked) and generate endless designs till you find the one you like the most.

There is a web version https://jspaint.app

What backends do you support? Do I need Redis or it can work with PostgreSQL? Cant find this info in readme.

The course is unfinished, I'm working on it.

Definitely, it is not low-effort. I'm writing code for SaaS app and describing it in the tutorial. It is quite a lot of work.

The text on the website might be better, I hope I will improve this one.

There are many people looking for advanced tutorials, not simple todo apps. This course is advanced. It shows how you can build a full SaaS app from scratch. Additionally, there are instructions how to deploy/update app in the cloud.

The website looks great! Congrats!

What is your goal with the project? What tech stack have you used?

I'm working on tutorial on how to build a SaaS application from scratch with React (TypeScript) and Django (Python). During the course, you will learn how to build server uptime monitor.

It is different course, because you build real, working product. The service that you will build is available online at https://monitor-uptime.com Most of the React+Django tutorials create simple todos apps. In this tutorial, you build real product and deploy it to the cloud.

The course is in text-form (no videos). You need to know some basics of React and Django. I will not teach basics of TypeScript and Python. I will show you how to join both technologies and quickly develop a full application. The final code in deployed with docker-compose on AWS EC2 machine. I'm using niginx, PostgreSQL and Let's Encrypt in the docker-compose.

The course is not fully ready yet! That's why I offer 50% discount for early supporters.

If your data doesn't require preprocessing ... Very often you need to prepare data to use xgboost/lightgbm. You need to fill missing values, convert categoricals/dates/text to numeric, do feature engineering.

What is more, MLJAR AutoML is checking much simpler algorithms for you, like Dummy Models (average response or majority vote), linear models, simple decision trees - because very often you don't need Machine Learning. The xgboost/lightgbm cant do this :)

It is hard to find a full tutorial on how to build a full SaaS product. There are many small tutorials on how to use React or Django. There are very few examples how to combine both of them (simple todos apps).

I'm working on tutorial how to use React and Django to build SaaS for server uptime monitoring. The tutorial is starting from scratch and shows all steps needed to build a product and deploy to the cloud. The SaaS is running at https://monitor-uptime.com

The frontend is created with TypeScript in React. The backend is done with Python in Django (Django Rest Framework). The service is deployed on AWS with docker-compose, nginx and Let's encrypt.

The tutorial is not fully ready. Looking for early users interested in the content. I'm offering 50% discount for early supporters.

Low quality Data Science code is not a fault of Jupyter.

The Jupyter allow you to load big chunk of data or some large model only once, and then use it for experiments in other cells. It is hard to replace this feature with plain `*.py` file. For me, this is the killer feature.

Do you know other B-corps in data science space?