@josh_wills - Have you given any thought to creating a tap for DuckDB? I'm using DuckDB in my workflow because it provides me efficient temp storage (4GB of .csv files in 600KB) and loading the .csvs is much faster than pandas` read_csv(). Since Metabase is not yet ready to consume the DuckDB database directly, I'm pushing my data into Postgres for reporting. I'm working around not having the DuckDB tap by just pushing the data to Postgres with pandas right now, and it's working fine, but I'm curious if the tap is something you're already considering.

Isn't something I have on my roadmap as of yet, but I'm happy to take a look at it in the near future; I'm also happy to look at just adding a DuckDB connector for Metabase as I did for Evidence: https://github.com/evidence-dev/evidence/pull/435

tangentially related - have you considered parquet? Obviously this can be a bit annoying since 1 table = 1 file, but it is another option.

yeah I checked-- parquet doesn't have out-of-the-box support with metabase

Apologies - my question wasn't as clear as it should have been, or you jumped past my immediate problem to the final solution. I'm interested in a DuckDB tap/loader for Meltano so it can move data from DuckDB into Postgres. I view this as a potential interim solution, but on if the Metabase driver can eventually meet my needs, some of which I described in this post: https://affordabledata.substack.com/p/postgresql-to-duckdb I wouldn't ask if I thought I would be the only one using the tap, I also view it as something that would benefit the community as DuckDB adoption increases, but I'm still relatively new to this.

Yes, this is why I asked about parquet format. I don't expect much development on duckdb as a tap until the file format matures, which might be quite some time.