This isn't currently possible. You can override the kv-metadata written by the spatial extension by passing a struct to the KV_METADATA parameter, but if you give that struct a geo key its going to overwrite the geoparquet kv-metadata already written. E.g.

COPY (
    SELECT
        42 AS number,
        true AS is_even
) TO 'kv_metadata.parquet' (
    FORMAT PARQUET,
    KV_METADATA {
        number: 'Answer to life, universe, and everything',
        is_even: 'not ''odd''' -- single quotes in values must be escaped
    }
);

Adding support for specifying CRS/bbox covering columns is planned but most likely won't make it in before next release.

But could I just read the full geo key and then modify the whole thing, and write it all back in?

Like get the full 'geo' key's json, then do json manipulation in python to insert crs and bbox covering info, then write that full json back to the geo key?

Maybe? But I don't think you can update parquet kv-metadata in-place after the file has been written right?

Ah, probably not. But my current approach with pyarrow rewrites the file anyways, so maybe I can read it in again with duckdb, add the metadata and write it out. Or is there a way to get the geo key that the geoparquet writer would write out?