Engine & Primitives

The concepts here are the heart of Cloacina, independent of how you run it. They apply whether you embed the library or run the service — each object is described once, with both its Rust and Python interfaces.

Who this is for: anyone learning the engine primitives — workflows, computation graphs, scheduling, and packaging — regardless of language. This material is language-agnostic.

Prerequisites: basic familiarity with the concepts from Start Here.

The core primitives:

Workflow — a durable, database-backed DAG; the task is the unit of scheduling.
Task — a unit of work in a workflow.
Context — typed data passed between tasks, persisted and recovered.
Computation Graph — an in-process, event-driven dataflow; the whole traversal is the unit.
Node — a vertex in a computation graph.
Reactor — binds accumulators to a graph and fires it when criteria are met.
Accumulator — turns a source or stream into boundary events (passthrough, stream, polling, batch, state).
Boundary event — the typed event an accumulator emits and a reactor reacts to.
Trigger — fires a workflow (poll or cron).
Cron schedule — time-based workflow scheduling.
Package (.cloacina) — the distributable unit.
Runner — the host that executes workflows against a database.