Skip to content
Cloacina

What is Cloacina?

Cloacina is a workflow orchestration engine for Rust and Python. It runs durable, database-backed task pipelines — with retries, recovery, and at-least-once execution — and in-process, event-driven computation graphs. Its only hard dependency is a database (PostgreSQL or SQLite): no broker, no queue, no coordinator.

One engine, two ways to run it

Section titled “One engine, two ways to run it”

The same engine runs two co-equal ways. Neither is the “real” way; you pick by how you want it to live in your system:

  • Embed the library — add cloacina (Rust) or cloaca (Python) as a dependency and run orchestration inside your application, against a database you already operate.
  • Run the service — operate cloacina-server as a multi-tenant control plane with an HTTP/WebSocket API, a web UI, and horizontal scale; ship workflows to it as .cloacina packages.

You don’t graduate from one to the other. Embedding is a permanent, production-legitimate way to run Cloacina — chosen because you want orchestration as a component of your own system, not because you haven’t “grown” into the server yet. Equally, plenty of teams want the service from day one.

The embedded-first principle

Section titled “The embedded-first principle”

Cloacina is embedded-first by design: the engine is a genuine standalone library, and the server is built on top of it — not the other way around. That ordering is a deliberate commitment. It’s why embedding is truly first-class rather than a stripped-down mode: the library is the product, and the service is one way to deploy it.

Two execution primitives

Section titled “Two execution primitives”

Whichever way you run it, Cloacina exposes two primitives (described once, for both languages, in Engine & Primitives):

  • Workflows — durable, database-backed DAGs where the task is the unit of scheduling. For work that must survive restarts and recover from failure.
  • Computation Graphs — in-process, event-driven DAGs where the whole traversal is the unit. For low-latency processing that reacts to a stream.

Next

Section titled “Next”