Episode
An episode is reinforcement learning's unit of experience: one complete run through an environment from initial state to termination, containing the sequence of states, actions, and rewards. A recorded browser session is an episode with the web as the environment, which makes the term the bridge between session data and RL tooling.
How it works
RL infrastructure operates on episodes: replay buffers store them, algorithms sample them, evaluation aggregates over them. Mapping sessions onto the abstraction is mostly direct, with states as interface snapshots, actions as interaction events, and termination as session end or task completion; the reward signal is the piece sessions do not natively carry and must be annotated (task completed, conversion reached, error avoided), usually from the same custom events analytics already emits. The mapping matters because it lets the two worlds share tooling: an archive of rrweb recordings is, to an RL pipeline, a pile of offline episodes, and offline RL methods train policies from exactly such piles without touching production.
Where you encounter it
RL papers and frameworks, offline RL dataset formats, and pipeline design where session archives meet training systems.