Thanks for shared your experiences. Nowadays, I interested in Kafka Streams API and our road map is similar. I want to aggregate some real-time data and want to push aggregation results to the socket. So, I decided to use Kafka Streams API even wrote a POC. That's solved the problem and we will take it to the production environment. But, as far as I can see, your choice was KSQL. Is there a specific reason for your choice of KSQL?