You can simply install scrapy-contrib-bigexporter to directly save to parquet from Scrapy: https://github.com/ZuInnoTe/scrapy-contrib-bigexporters

It supports parquet, orc and avro. It has a couple of configuration options that allow you to be flexible (e.g. automatic inference of schema, compression etc.).

See here also an example on how to use parquet: https://github.com/ZuInnoTe/scrapy-contrib-bigexporters/tree/main/examples/quotes_parquet