+ jHjVNRt^RIt^RIHtHtHt^RIHtRtRt Rt Rt Rt R#) aACP v0.9 has no per-session destroy: the last viewer leaving an idle ACP session ends it (#118216). The in-process adapter stamped ``ended_at`` at its stdio shutdown; under the gateway the ACP process is only a viewer, so the owner stamps it when the last subscription goes or the gateway stops. Without a writer, source='acp' rows stay open forever and the ended-session guard keeps prune/archive away. A later ``session/load`` + prompt reopens the row (``session_local_recovery.reopen_local_session``). N) POLICY_PREFIXend_idle_local_session local_receipt)RuntimeStoreErroracp_disconnectc VPPV4pVeVP'dR#\VPV4pTPR/4PR4R8wdR#\ \TPTPTTR,R,\R74# \ dR#i;i \ \\3d1\P!\4PRTR R 7R#i;i) zXEnd *session_id* iff it is an ACP-created local session with no viewer and an idle FIFO.Fpolicysourceacpentry session_idepochr target_idreasonzACP session %s not endedTexc_info)sessionsget subscribersrdbrboolrrACP_END_REASONKeyError TypeErrorlogging getLogger__name__warning) authorityr livereceipts&& B/home/ubuntu/.hermes/hermes-agent/gateway/session_acp_lifecycle.pyend_idle_acp_sessionr#s    ! !* -D D,,, j9{{8R $$X.%7*9<C3 CCAD  D c &VPPV4pVeVP'dR#\VPV4pTPR/4PR4R8Xd \ Y4#\\TPTPTTR,R,TR74# \ dR#i;i \ \\3d1\P!\4PRTR R 7R#i;i) a``session.detach {end_reason}`` from the last viewer: end *session_id*'s idle local row with the viewer's own reason (main's classic CLI ``end_session(sid, 'cli_close')`` and its switch reasons). ACP rows keep their own reason; a busy FIFO keeps the row open, as for ACP.Frr r r r r zlocal session %s not endedTr)rrrrrrr#rrrrrrrrr)rr rr r!s&&& r"end_idle_viewer_sessionr%#s    ! !* -D D,,, j9{{8R $$X.%7#I::*9<(end_idle_acp_sessions..HsCs#Is33ss)r _read_ctxexecuterlensum)rconnrowidssf r"end_idle_acp_sessionsr;Cs    ! !T6:ll 9MCrAs2SS2!&*Dr@