I think you misunderstand. The non-text files remain in S3, my service just parses them for text, then stores it in a Lucene index to become searchable.