Struct ReliableQueue 

pub struct ReliableQueue { /* private fields */ }

Reliable message delivery with acknowledgment tracking.

Tracks pending deliveries by correlation ID. On ack timeout, retries with exponential backoff. After max_retries, routes to the dead letter queue.

Implementations

impl ReliableQueue

pub fn new(config: ReliableConfig, dead_letter: Arc<DeadLetterQueue>) -> Self

Create a new reliable queue with the given config and dead letter queue.

pub fn with_defaults(dead_letter: Arc<DeadLetterQueue>) -> Self

Create with default configuration.

pub fn with_model(self, model: Arc<Mutex<DeadLetterModel>>) -> Self

Install a learned DeadLetterModel on this queue. With a model installed, Self::check_timeouts and Self::expire_max_retries route their backoff / discard decisions through the model. Untrained models fall back to the hardcoded formulas, so with_model is drop-in safe.

NOTE(eml-swap): wired — Finding #5 (DeadLetterModel).

pub fn dead_letter_model(&self) -> Option<&Arc<Mutex<DeadLetterModel>>>

Returns a handle to the optional dead-letter model.

pub fn config(&self) -> &ReliableConfig

Get the configuration.

pub fn register(&self, msg: KernelMessage) -> (String, KernelMessage)

Register a message for reliable delivery tracking.

The message must have a correlation_id set. If not, one is generated. Returns the correlation ID used for tracking.

pub fn acknowledge(&self, correlation_id: &str) -> Option<DeliveryResult>

Acknowledge receipt of a message by correlation ID.

Returns Some(DeliveryResult::Acknowledged) if the message was pending, or None if not found (already acked or expired).

pub fn check_timeouts(&self) -> Vec<(String, KernelMessage)>

Check for timed-out deliveries and return messages needing retry.

Returns a list of (correlation_id, message) pairs that have exceeded their ack deadline and should be retried.

NOTE(eml-swap): wired — Finding #5 (DeadLetterModel). When a learned model is installed and trained, the next-attempt ack deadline comes from the model; untrained models fall back to the existing initial_timeout * backoff_multiplier^attempt formula bit-for-bit.

pub fn expire_max_retries(&self) -> Vec<DeliveryResult>

Move deliveries that have exceeded max retries to the dead letter queue.

Returns the list of message IDs that were dead-lettered.

NOTE(eml-swap): wired — Finding #5 (DeadLetterModel). When a trained DeadLetterModel is installed, the discard decision consults the model’s should_discard head in addition to the config’s max_retries ceiling. The untrained fallback matches today’s “attempts > max_retries” rule exactly.

pub fn pending_count(&self) -> usize

Number of currently pending deliveries.

pub fn is_pending(&self, correlation_id: &str) -> bool

Check if a specific correlation ID is pending.

pub fn pending_snapshot(&self) -> Vec<(String, PendingDelivery)>

Get a snapshot of all pending deliveries.

pub fn cancel(&self, correlation_id: &str) -> Option<KernelMessage>

Cancel tracking for a specific message.

pub fn timeout_for_attempt(&self, attempt: u32) -> Duration

Calculate the timeout for a given attempt number.

Uses the same EML-aware dispatch as Self::check_timeouts so external callers observing the policy see the same delays the queue would actually wait.