Pluggable "worth saying" scorer for Active mode.
Returns a score in [0, 1]; the policy speaks when the score meets the configured threshold AND the silence/throttle gates pass. The scorer itself is injected (it may be a heuristic or an LLM call) so the policy stays pure and deterministic in tests.
[0, 1]
The most recent transcript segment.
A worth-saying score in the range [0, 1].
Pluggable "worth saying" scorer for Active mode.
Returns a score in
[0, 1]; the policy speaks when the score meets the configured threshold AND the silence/throttle gates pass. The scorer itself is injected (it may be a heuristic or an LLM call) so the policy stays pure and deterministic in tests.