Thank you for this pull request and all the other fixes you have sent. I'd love to learn more about what you are doing and what's missing from Draco 2.

You are welcome; I am glad that my contributions to the repo were accepted.

I am actively working on a project I am excited about, visrecly, a user-task-based visualization recommendation tool that also happens to be the topic for my Bachelor's Thesis at Uni Wien. It is heavily inspired by your projects and papers, so thank you very much for your academic contributions!

The main reason I decided to depend on Draco in my project instead of Draco 2 is that - as outlined in cmudig/draco2#341 - the latter does not have a default spec-renderer at hand just yet. On the other hand, with Draco, I just need to plug the generated vega-lite specs into react-vega, and the recommendations will appear automagically on the frontend (:

Another compelling (but not decisive) argument for Draco was that its API is written in TypeScript, making it an excellent fit for my web-based stack. Also, as clingo-wasm solves my ASP problems in the browser, the need for a Python-based rec server got eliminated; hence I will be able to explore offline PWA options as future work.

Long story short, a default renderer and a web API for Draco 2 would be very useful imho, due to the inherent nature of visualizations: most users want to play around with the results immediately, right in their browser.