Why sidecar metadata files? In general I've tried to minimize the number of files on disk since that makes copying slow and error-prone. (A future version of Timelinize will likely support storing ALL the data in a DB to make for faster, easier copying.) We'd still need a DB for the index anyway, which essentially becomes a copy of the metadata.
Don’t know if you have seen the work DuckDB is doing on ducklake. Maybe there is an overlap in vision for versioning data across multiple data sources - and similar to SQLite it’s not proprietary and easily drilled down on. I’m sorry, don’t have technical knowledge :/
