The output gets automatically compressed (when the format supports that), based on the extension that you use for the output.

The following uses Gzip to compress the CSV file generated from the Alpaca JSON dataset:

llm-convert \
  from-alpaca \
    --input ./alpaca_data_cleaned.json \
  to-csv-pr \
    --output alpaca_data_cleaned.csv.gz

The input gets automatically decompressed based on its extension, provided the format supports that.