glrs
    Preparing search index...

    turns

    a turn is one exchange: your message, the model's work, its answer. a step is one model call inside it.

    queue key delivery setting
    follow-up enter its own turn once the current one finishes followUpMode
    steering alt+enter joins the running turn at its next step steeringMode

    one-at-a-time (the default) delivers the oldest waiting message. all delivers everything waiting, joined by a blank line. with nothing running, alt+enter is just a turn.

    a provider charges less for a prefix it has seen before, so the prefix is kept stable: the system prompt is byte-identical every turn, and steering is appended rather than inserted.

    provider how
    openai, google caches a prefix without being asked
    anthropic needs a breakpoint written into the messages
    amazon bedrock needs a cachePoint

    the breakpoint goes on the second-to-last message, the newest point still present next turn. it advances each turn, which extends the cached prefix rather than replacing it.

    why it is shaped that way: a turn.

    see also: resume and fork, a turn