DiscussionState
autogen.beta.network.adapters.discussion.DiscussionState dataclass #
DiscussionState(participant_order=list(), expected_next_speaker=None, last_speaker_id=None, last_envelope_id=None, turn_count=0)
Folded state for a discussion session.
participant_order is snapshotted at initial_state so fold can compute the next speaker without access to SessionMetadata (the adapter Protocol passes only state + envelope into fold). The hub's hydrate() re-folds from the WAL, which is deterministic because initial_state reads the same metadata.participants snapshot every time.