Skip to content

Wizard

Warning

Wizard is a living project, and is constantly being improved and new features are being added. Consequently, this documentation might be slightly out of sync.

The Wizard is OWID's ETL admin tool. It is an interactive streamlit-based web app that provides a user-friendly interface to manage our ETL catalog.

It was initially developed to ease the creating of ETL steps by means of templating, but it has evolved to more than that and now provides a wide range of functionalities in the ETL workflow.

Run it locally with the following command:

etlwiz

and then visit localhost:8053.

Run it from admin and staging servers

  • Wizard is available from any server that bakes OWID's site (e.g. staging servers), including live admin page. Just look for "Wizard" in the navigation menu.
  • The production version runs at etl.owid.io/wizard (needs Tailscale).
  • Note that some of the functionalities might not be enabled in a remote setting. For instance, creating steps is currently only available when running locally.

The different pages in Wizard

Wizard is structured into different sections, each of them grouping different pages (or apps) depending on what they do.

In the following sections we try to give a brief overview of each of the sections and the pages they contain.

Get latest details from the app

The best way to know what Wizard can do is to run it and check the available options. The app is constantly being updated and new features are being added.

Wizard
Wizard as of 5th June 2024.

Create ETL steps

This section is dedicated to the creation of new ETL steps, including Snapshot, Meadow, Garden and Grapher steps. Additionally, Fast-Track steps can also be created using the Wizard.

Using Express

Express mode is helpful when working with very canonical datasets. It is a one-click solution to create a meadow, garden and grapher step in one go.

In each step creation, a form is presented to the user so that they can fill in the necessary metadata fields. Based on the input, new files (e.g. python scripts, metadata YAML files, etc.) are created and modifications to existing ones (e.g. the DAG) are done.

After submitting each of the forms, a short guideline is shown so that the user knows what they need to do next.

ETL-steps
The various pages you have to create ETL steps.

Expert

GPT-based assistant to help resolve doubts. Doubts can include anything ETL-related (metadata structure, environment setup, etc.). This documentation is fed to the Expert, so it should be able to answer most of the questions concerning this documentation.

Additionally, Expert can also help out create Datasette queries!

Expert
Asking the Expert to generate a Datasette query to get the charts with most views.

Data tools

Pages to help us improve our charts (e.g. keeping them up to date). The current pages are:

  • Indicator Upgrader: Upgrade old indicators with their corresponding new versions to keep the charts up to date. You will need to (mostly) manually map "old indicators" to "new indicators". Then, the tool will update all affected charts with the new indicators. These modified charts can be reviewed with Chart diff.
  • Chart diff: Shows all charts in your environment (e.g. staging server) that have been modified compared to the production. This is useful to review the changes before they are pushed to the production.
  • Harmonizer: Harmonize the entity names of a table.

Learn more about updating charts section

Monitoring

  • Dashboard: Monitor all our datasets and update them quickly!
  • Dataset Explorer: A tool to explore the datasets in the ETL catalog. You can check the step dependancies and its metadata. If it is a Garden step, you can also perform some actions with it.

Research

  • Insighter: Generate insights from a chart using LLMs.

Misc

  • News: Brief summary of the latest activity in the ETL repository. This is only available in production.
  • owidle: Daily challenge where you have to guess the country based on the data provided.

Metadata

  • Meta Upgrader: Upgrade v1 metadata YAML files to v2. This tool uses chatGPT to suggest the new YAML structure.
  • Meta Playground: A playground to test the metadata of a step. It is useful to check if the metadata is valid and to see how it will look like in a data page of an indicator.

Adding new functionalities to Wizard

The code for the Wizard lives in apps/wizard. It is a streamlit app, so you can also run it with streamlit run apps/wizard/app.py.

Adding a new page

We are trying to keep Wizard as modular as possible, so that it is easy to add new pages to it.

We encourage everyone to experiment with tools from which the team can benefit. Make sure to discuss your ideas with the rest of the team, so that you can make a good use of your time.

To add a new page, follow these steps:

  1. Create a new python script, and place it under apps/wizard/app_pages. This script should be a streamlit python script, and should contain the code to render the page.
  2. Add an entry in the config file apps/wizard/config/config.yml describing your new page. You should first decide in which section you should add your page to or create a new one. You will find more details on how to add your page in the config file itself.