AI Dictionary › Fondamenti AI
Streaming is the processing or transmission of data as a continuous flow, item by item, as it arrives, rather than in accumulated blocks. It applies to two related worlds: content streaming (video and audio playing while downloading) and stream processing (real-time analysis of event flows). In both cases the principle is the same: act immediately, without waiting for everything to be available.
Streaming is the processing or transmission of data as a continuous flow, item by item, as it arrives, rather than in accumulated blocks. It applies to two related worlds: content streaming (video and audio playing while downloading) and stream processing (real-time analysis of event flows). In both cases the principle is the same: act immediately, without waiting for everything to be available.
In stream processing, data flows as events through platforms like Apache Kafka or Flink: each event (a click, a transaction, a sensor reading) is processed within milliseconds, aggregated over sliding time windows and forwarded downstream. In content streaming, the file is split into segments the client plays while downloading the next ones, adapting quality to available bandwidth.
In AI, streaming is an everyday experience: LLM responses arrive token by token, which is what makes chats feel responsive even when full generation takes seconds. On the data side, real-time flows feed recommendation systems, fraud detection and agents reacting to events. Beyond AI, streaming has redefined music, TV and finance.
The English word stream, a current or brook, comes from Old English and Germanic roots. In software, the data-flow metaphor is as old as Unix and its byte streams; mass multimedia streaming took off in the 1990s with RealAudio (1995), while large-scale stream processing established itself in the 2010s with Kafka, born at LinkedIn and open-sourced in 2011.
Lo streaming è l'elaborazione o la trasmissione dei dati in flusso continuo, elemento per elemento, man mano che arrivano, invece che in blocchi accumulati. Si applica a due mondi affini: lo streaming di contenuti (video, audio riprodotti mentre si scaricano) e lo stream processing (analisi di flussi di eventi in tempo reale). In entrambi i casi il principio è lo stesso: agire subito, senza aspettare che tutto sia disponibile.
Nello stream processing, i dati fluiscono come eventi attraverso piattaforme come Apache Kafka o Flink: ogni evento (un click, una transazione, una lettura di sensore) viene elaborato in millisecondi, aggregato su finestre temporali mobili e inoltrato a valle. Nello streaming di contenuti, il file viene suddiviso in segmenti che il client riproduce mentre scarica i successivi, adattando la qualità alla banda disponibile.
Nell'AI lo streaming è esperienza quotidiana: le risposte degli LLM arrivano token per token, ed è questo a rendere le chat percepite come reattive anche quando la generazione completa richiede secondi. Lato dati, i flussi in tempo reale alimentano sistemi di raccomandazione, rilevamento frodi e agenti che reagiscono a eventi. Fuori dall'AI, lo streaming ha ridefinito musica, TV e finanza.
La parola inglese stream, corrente o ruscello, viene dall'antico inglese e da radici germaniche. Nel software la metafora del flusso di dati è antica quanto Unix e i suoi stream di byte; lo streaming multimediale di massa decollò negli anni '90 con RealAudio (1995), mentre lo stream processing su larga scala si affermò negli anni 2010 con Kafka, nato in LinkedIn e reso open source nel 2011.
From our network
AGORÀ Intelligence — Enterprise AI Governance Platform
Govern AI at scale: policies, adoption and measurable results on your data. Built for boards and C-suite.
Visit agora-intelligence.com →From the Agora Intelligence blog
📱 Download the Android app (beta) iOS coming soon
Say what you mean. Get what you need.
Grace Certified — the AI coach that trains and certifies your prompt engineering — by Agora Intelligence.