As I experimented with AI, I realized how incredible it was at building custom data visualizations. The problem was that the AI outputs couldn't really live anywhere without having to manage all of the infrastructure yourself.
The generated code is only one piece of delivering an interactive visualization. You still need somewhere for the data to live, a way to keep the data fresh, hosting, access control, and version control for the code.
I started building Rhubarb as the platform that I wanted to use to deliver analytical products. Having been forced into using legacy point and click data visualization tools for years, I wanted to build the platform that would let me build the kinds of visualizations I wanted to build.
Over the last year the models improved so dramatically, that between that and my development background, Rhubarb turned into a real project.
My intent was to build an environment around AI’s ability to write code for visualizations, so that you could get from data and a concept to a published visualization all in one workflow.
The Rhubarb Assistant writes the code for your data transformations and data visualizations and lets you inspect the code to verify and validate what it produces. The environment pulls all of that together so that at the end of each request, there is a live version of the change rather than just a block of generated code.
For me, this has made it dramatically easier to build and publish interesting analytical visualizations while maintaining control over the analytical pipeline.
This has also produced some results I wasn’t really planning to build. Having been so impressed by it's ability to build any type of custom visualization I'd asked for so far, I asked for something a little absurd to test the boundaries. I asked the assistant to make something inspired by Monet’s water lilies, then I asked it to add frogs and to make them jump when you startled them.
In a couple of prompts I had created the silly interactive frog visualization that's in the demo, entirely from my phone. It is only a couple of prompts away from being a playable game, maybe you can get it there.
The public demo is deliberately limited. You can use the Rhubarb Assistant to modify live visualizations, inspect what it produces, and try building something from scratch. The full thing handles much more of the life cycle from connecting to data, managing and transforming it, to publishing your results.
Give it a try, I would like to hear what you think.
Try it: https://rhubarbviz.com/try?launch=1
View examples: https://rhubarbviz.com/demo-visualization-gallery