Looks really interesting. It reminds me a lot of modern arco approaches (kerchunk). There is quite a bit of interest in tests and studies of federated query over file serializations (vs managed database solutions) (e.g. parquet, zarr) with contextual headers storing mapping files R2RML/RML, kerchunk). I could see a lot of performance potential here vs say virtualized queries via athena over json-ld files, while still maintaining benefits of file-store data vs managed service. Thanks for the share.
