Warm up a Co-STORM session asynchronously — tempest_session_warmup

Usage

tempest_session_warmup_async(
  session,
  timeout_s = getOption("tempest.costorm.warmup_timeout_s", 120),
  max_parallel_experts = getOption("tempest.costorm.warmup_max_parallel_experts", 3L),
  is_current = function() TRUE
)

Arguments

session: A TempestSession created by tempest_session().
timeout_s: Maximum seconds allowed for each expert orientation. Use NULL, a non-positive value, or a non-finite value to disable timeouts.
max_parallel_experts: Maximum number of expert requests started at the same time.
is_current: A zero-argument function that returns TRUE while the originating host session is current. Stale operations resolve without committing late results.

Value

A promise that resolves to an internal tempest_warmup_result S7 object with aggregate counts and per-expert audit records.

Details

Runs one bounded, evidence-seeking orientation for each active expert without blocking the caller. Expert requests run in bounded parallel batches, while transcript and evidence commits occur deterministically in expert order. A single mind-map update follows the completed evidence commits.