Superpowers for Rust message consumers

Production-ready retries, DLQs, ordering, audit logs, and autoscaling for your existing messaging stack.

Get startedGitHub

cargo add shove --features inmemory

What you get

A small API on top of the brokers you already run. Four pillars:

• Strongly typed topics — bind a message type to a queue topology once; the compiler keeps publishers and consumers in sync.
• Managed consumer groups with autoscaling — min/max worker bounds driven by queue depth or consumer lag.
• Delivery ordering — strict per-key ordering with a single .sequenced() switch.
• Per-instance broadcast — every replica hears every message, with a single .broadcast() switch and nothing left behind on the broker.

Plugs into RabbitMQ, AWS SNS+SQS, NATS JetStream, Apache Kafka, and Redis/Valkey Streams.

What's next

• Get started in 60 seconds — runs in-process, no Docker required.
• Core concepts — topics, outcomes, handlers, broadcast, the Broker<B> pattern.
• Per-broker setup — production patterns for RabbitMQ, SNS+SQS, NATS, Kafka, and Redis/Valkey.
• Full API reference — auto-generated rustdoc on docs.rs.

Who builds this

shove is written and maintained by Zannis Kalampoukis, a Rust and distributed-systems engineer. It grew out of production messaging work at Synchronicity Labs, a software engineering and consulting studio that builds and rescues backends, data platforms, and onchain systems.