Struct RecoveryCoordinator 

pub struct RecoveryCoordinator { /* private fields */ }

The recovery coordinator. Holds the shared routing stats (for FallbackEvent recording) and the live fallback-model list. The lifecycle is passed per-call (the orchestrator owns it).

Implementations

impl RecoveryCoordinator

pub fn new(routing_stats: Arc<RoutingStats>, config: ResilienceConfig) -> Self

Create a coordinator. routing_stats is shared with EngineApi / AgentRuntime so fallback events surface in the Web UI.

pub fn set_health(&self, health: Arc<ProviderHealthRegistry>)

Wire the per-provider circuit breaker (RFC-029 P3). When set, L2 skips every model whose provider’s breaker is open.

pub fn set_a2a_breaker(&self, breaker: Arc<A2ACircuitBreaker>)

Wire the A2A circuit breaker for L4 delegation gating (RFC-029 P4). When set, L4 checks is_allowed before attempting delegation.

pub fn set_fallback_models(&self, models: Vec<String>)

Update the fallback-model list (called when engine routing config changes, e.g. set_routing / hot-reload).

pub fn set_config(&self, config: ResilienceConfig)

Update the resilience config (hot-reload).

pub async fn execute( &self, lifecycle: &AgentLifecycleManager, directive: &Directive, env: &ExecEnv, ) -> Result<ExecutionResult>

Execute a directive with the recovery ladder.

On success or non-provider failure (cancellation/abort), returns the result as-is. On a provider failure (failure_class: Some), escalates through L1 (same-model backoff) → L2 (model swap) → terminal.