Graph Scheduler

Graph Scheduler answers one question about a dependency graph: what is runnable now? Given the nodes, their dependency edges, and each node’s status, it computes the readiness frontier — the nodes whose prerequisites are satisfied — and pulls the next runnable work, with rules, ordering, and goal-direction.

It is a small, layered, zero-dependency library. The engine reasons over the graph and status alone; it never runs your work or touches the outside world.

Early days

The package is a scaffold — the layers land task by task. The API Reference in the sidebar is generated from the shipped source, so it grows as the engine does. The full intended API (getReady, trigger rules, filters, sorters, goal-direction) is specified in the design decision D-BPD8-graph-scheduler-api in the repository.

Entry points

Layered so dependencies point down. The whole library is zero-dependency.

Import	Subpath	What it is
graph-scheduler	.	The package root — re-exports the layers and the root getReady engine.
graph-scheduler/graph	./graph	The immutable DAG core: nodes, edges, dependency structure.
graph-scheduler/tracker	./tracker	Readiness tracking: which nodes are runnable given completed dependencies.
graph-scheduler/pull	./pull	The pull surface: draw the next runnable node from the tracker.

API reference

The API Reference section in the sidebar is generated directly from the TypeScript source, so it always matches the shipped surface. It fills in as each layer lands.