Se creo el modulo runtime/control/ que agrega estado operacional desde los modulos existentes (routing_history, cluster_state, mode_manager, health_score, Qdrant) sin duplicar logica. Se expone via 6 nuevos endpoints en la Live API (:8084).

EndpointDescripcion
GET /api/control/runtimeStatus ultra-compacto (mode, health, nodes, latency, governance)
GET /api/control/statusEstado completo (runtime + uptime + Qdrant + nodos offline)
GET /api/control/nodesEstado por nodo (online, modelos, latencia, tool_use)
GET /api/control/routesUltimas decisiones del router (modelo, nodo, task, latency, failover)
GET /api/control/policiesPoliticas activas (execute_policy, observe_policy, tool_fastpath)
GET /api/control/explain/last-routeExplica la ultima decision de ruta (reason_codes, fallback)

El endpoint /api/control/runtime incluye el campo governance con valores:

EstadoCondicion
NORMALSin bloqueos, Qdrant healthy, fallback_rate bajo
ELEVATED>0 bloqueos en 1h o fallback_rate > 10%
DEGRADEDQdrant degraded o tool parser failures > 5
LOCKDOWN>5 bloqueos en 1h + fallback_rate alto + Qdrant caido
ArchivoAccion
runtime/control/__init__.pyNuevo
runtime/control/control_plane.pyNuevo (~180 lineas)
runtime/state/live_api.pyModificado (6 endpoints + handlers)
Terminal window
# Runtime compacto
curl -s http://192.168.1.30:8084/api/control/runtime
# Estado completo
curl -s http://192.168.1.30:8084/api/control/status
# Nodos
curl -s http://192.168.1.30:8084/api/control/nodes
# Ultimas rutas
curl -s http://192.168.1.30:8084/api/control/routes
# Politicas
curl -s http://192.168.1.30:8084/api/control/policies
# Explicacion de ultima ruta
curl -s http://192.168.1.30:8084/api/control/explain/last-route
#SubfaseEstado
18.2Model & Node Health ScoringPendiente
18.3Tool Reliability MetricsPendiente
18.4Memory Usefulness EnginePendiente
18.5Governance Audit v2Pendiente
18.7Recovery ManagerPendiente
18.8Professional Ops Dashboard /ops/controlPendiente
18.9Stability Validation CriteriaPendiente