<!here> Today is a good day! 🚀 We're celebrating many things, including @Yolan Honoré-Rougé finally joining Kedro's Technical Steering Committee (TSC). He has made significant contributions to Kedro via

kedro-mlflow

,

kedro-pandera

, and he is thoughtful about progressing Kedro's roadmap and development. You'll experience him as a helpful community member in channels like #C03RKPCLYGY. You can read more about our TSC. We're also celebrating the open sourcing of a new project that already integrates with Kedro. This is a QuantumBlack project that we've been excited about for a while, and it is called Vizro - I swear we did not name it. 🤡 Vizro is a toolkit for creating modular data visualisation applications, and it is built on Plotly and Dash. It integrates with Kedro and allows you to trigger pipeline re-runs from the dashboard views. 🚀 You can: • 📊 See a live demo • ⭐ Visit and star the GitHub repo (if you want) • 📚 Check out the documentation • 🦠 Read a Medium blog post about Vizro

A more native integration with vizro is also coming.

Just saw @Juan Luis’s post on LinkedIn! Congrats y'all it looks very cool! Related to the integration, @Nok Lam Chan would I be correct in guessing that some of the standalone datacatalog functionality might be available here in the future? 😄

What integration do you want to see? A pure DataCatalog powering a vizro dashboard? I suspect this works already.

In the past I've explored using the datacatalog component as the dataloader for streamlit visualizations. It was relatively straightforward to implement a basic functionality which for example allowed to visualize different datasets by selecting them from a dropdown of the datasets configured in the catalog

Drop your comment here! We will be discussing this tomorrow, actually a streamlit integration is also mentioned but vizro will be the first.

yeah I wouldn't really expect anything beyond that! I did briefly look into your last point relating to using kedro pipelines as the source of transformation steps for the actual running dashboard but 1. it was too complex, 2. it introduced too much overhead for streamlit's constant recalculation of stuff and 3. it simply did not add that much functionality beyond just using the datacatalog and doing the final filtering on the streamlit side.

Yeah - you wouldn’t want to do heavy lifting there. I can think of use case that may requires that, i.e. fetching some data from API connecting to a specific filter. That said DataCatalog will still be the first integration, then we can see what else are needed :)

After reading the article, two questions pop up: 1. how is this different from streamlit? 2. Do you see a future where kedro-viz and vizro are merged? I.e. where kedro-viz is an out-of-the-box interactive dashboard for kedro project metadata

Do you see a future where kedro-viz and vizro are merged? I.e. where kedro-viz is an out-of-the-box interactive dashboard for kedro project metadata

An integration make sense, and we are thinking about a kedro-viz plugin where vizro will be supported. They don’t need to be merged because Kedro is just one of the thing that Vizro support, it’s not coupled tightly. A Kedro project is mainly DataCatalog + ConfigLoader + Runner + Pipeline&Nodes + Hooks. Vizro only need the Data Catalog (or maybe ConfigLoader too)

@Matthias Roels I forwarded your questions to the Vizro team 🙂 hope they can reply soon (either here directly or I'll forward their thoughts as well)

@Nok Lam Chan and what about pipeline viz? I.e visualising the DAG?

Very cool! Curious to know more about how this tool fits into the open source ecosystem. For example,

streamlit

looks like a direct analogue. In which cases would

vizro

be a more suitable option? Asking to potentially try it out on the next project.

hi @Iñigo Hidalgo and @Yury Fedotov, thank you for the questions - in that respect Vizro provides a configuration driven assembly system which implements built-in visual styling and an opinionated approach in terms of both application architecture and UX, in order to enable standardisation in code and user journeys for multi-screen applications and complex interactions between components This approach of using a configuration layer to enforce standardisation in the way that components and code is assembled utilises a grammar which is largely tech agnostic, and while it is currently optimised to leverage Plotly and Dash, it could theoretically utilise components from other packages such as StreamLit in future

more “Vizro vs streamlit” on our socials, I forwarded your response @Joseph Perkins https://data-folks.masto.host/@benlindsay/111140404441511586