An order event often needs some preparation before other applications can use it. A reporting pipeline may standardize a currency code, select orders in a supported currency, extract individual order lines, and send expensive orders to a separate analysis topic.
Each of these decisions can use the current record without consulting earlier orders. Kafka Streams provides stateless transformations for this kind of work.
This chapter explains the main operations, how they affect record keys and output counts, and how to combine them into a clear pipeline. We'll also look at the limits of stateless processing, especially when records are invalid, the application repeats processing, or a callback calls an external system.