Streaming is a way of delivering and processing data continuously as it is produced instead of collecting it and transferring it all at once. In data engineering it refers to an architecture in which events such as orders, payments, and sensor readings are delivered to and processed by many systems the moment they occur, in contrast to batch processing, which gathers data and handles it at fixed intervals. Apache Kafka and Apache Flink are the best-known open-source technologies for it.
In AI the word is used in two contexts. One is data streaming that feeds models and agents with current data in real time; the other is response streaming, in which a language model sends tokens as they are generated instead of waiting for the full response. In everyday use it also means playing video or music as it is received without downloading it first.