flusso-queue-core

A generic work-queue abstraction for the sync pipeline — domain-agnostic, so the backend swaps without touching pipeline code.

Quick reference

Generic over the payload T, so it depends on neither the source, the sink, nor the engine. Both backends drive the same engine loop: recv → process → ack.

The point of the abstraction is a swappable backend:

• In-process tokio channels (see queue-channel) for single-node runs. Acking is a no-op there — durability comes from the source (the replication slot), and the bounded channel gives backpressure.
• A durable broker (e.g. NATS JetStream) later, where [AckHandle::ack] becomes a real server acknowledgement and the queue itself is the durability boundary for the work pipeline.

ℹ️ Info — because [AckHandle] models the ack identically, the engine loop is unchanged whichever backend backs it; only the durability guarantee shifts.