have you looked litestream backups for streaming replication? you can point the replica to a filesystem instead of s3, at the latest version gives point-in-time recovery, so you don't need to store N while copies

will you VACUUM at the end to reclaim disk space?

it would lock any writes and take forever ... but at least the DB will stop growing regardless

delete this

oooooh... taking this to the Discord now!

Id imagine any dialect, so DuckDB-style SQL would work fine. But the API response format, params like _shape, etc would be the same as Datasette

And other Datasette-isms, like `.csv`/``.json` to return results in different formats, the facetting API, `/$DB/$TABLE` urls, etc

In the past, Simon has talked about making Datasette a "specification" of HTTP APIs. Meaning one can make an implementation of Datasette that uses the same HTTP API, but is backed by DuckDB/Postgres etc

all to say: a duckdb-backed Datasette compliant tool would be very cool to see!

I think if DuckDB was around when Datasette was first created, it very likely would have!

Tho I'd say the "new" Datasette features we're working on, mostly writes + interactive UIs, it wouldnt make much sense on DuckDB

(and we're doing a terrible job at promoting the new features btw)

a semantic search over embedded images in political emails sounds VERY interesting...

apparently 191! github.com/paolobozzola...

ALSO: I made a Google Group for libfec. If you have questions, bug reports, feature requests, or just want to talk about federal campaign finance, check it out!

New features and releases for the libfec CLI will be announced here

Mailing lists are SO back

groups.google.com/g/libfec