ReadonlycodecsThe snapshot form of the composition's caches: how the scheduler copies live serial state it publishes.
request's state: its resumed generation, else the longest reusable prefix
of its prompt in namespace (its adapters, prepared media, method), else
fresh state. Ownership passes to the caller.
Admit request beside the requests already holding state: true admits it
(it holds its projected KV bytes until release), false means it waits.
Throws for a request that can never be admitted: one past the caches' own
position limit, or one over the KV budget alone.
An execution boundary passed: the cache may give optional state back to the allocator. Never changes a running request.
Save a resumable point of request's generation (owned caches; the cache
disposes them whatever happens).
Whether admit would admit request now; holds nothing.
Whether state request publishes may be kept: the scheduler skips
copying state the cache would drop.
Offer state covering tokens for reuse in namespace; the cache keeps or
drops it by its own policy. Ownership transfers on return; on a throw it
stays with the caller.
request is done: its KV bytes return to the budget; a completed
generation's checkpoints are removed, an interrupted one keeps them.
The scheduler's port to the composition's cache service.
acquireandreleasebracket a request; between them the scheduler offers state it produced (publishat prefix boundaries and when a row finishes,checkpointfor a replayable generation) and marks execution boundaries. Admission under the KV budget runs beforeacquire. Nothing here says what is resident or where.