Many records as JSON Lines: one compact canonical record per line, nothing else
on it. ka export --format jsonl used to join the pretty-printed on-disk form,
~80 lines per record with { alone on the first, which jq -s tolerates and
every line-oriented reader (pandas lines=True, DuckDB, split -l) does not.
Keys stay sorted, so a line is the stored record minus its whitespace.
Many records as JSON Lines: one compact canonical record per line, nothing else on it.
ka export --format jsonlused to join the pretty-printed on-disk form, ~80 lines per record with{alone on the first, whichjq -stolerates and every line-oriented reader (pandaslines=True, DuckDB,split -l) does not. Keys stay sorted, so a line is the stored record minus its whitespace.