DAG Diagrams

Every Ripperoni run is an authored dagonizer DAG. These diagrams are generated directly from the committed *.dag.jsonld documents and plugin DAGs by docs/.vitepress/scripts/render-dags.mjs (via the framework's MermaidRenderer), so they always reflect exactly what ripperoni run dispatches.

These node shapes are the same across every diagram (the renderer assigns one shape per placement type):

Shape	Placement	Meaning
[ name ] rectangle	SingleNode	one node runs once
[/ name /] trapezoid	ScatterNode	fan-out — one child run per source item
[[ name ]] subroutine	EmbeddedDAGNode	invokes a nested DAG
(( name )) double circle	Terminal (completed)	successful end state
> name ] flag	Terminal (failed)	failure end state
( name ) stadium	PhaseNode	a pre/post lifecycle hook

Placements bound to a worker container are tinted blue.

Hover a diagram for its D-pad (zoom · pan · centre · fit), scroll to zoom, drag to pan, or hit ⛶ expand for a fullscreen explore view (esc to close).

Orchestration — aonprd:crawl

The top-level run: embed crawl:discover to populate the URL list, then scatter the per-page DAG aonprd:page over it on the worker container.

Link discovery — crawl:discover

The native cyclic crawl: init-frontier → fetch-and-extract → dedupe-and-enqueue with a back-edge that loops until the frontier is exhausted or the budget is hit.

Per-page chain — aonprd:page

The unit of work for one detail page: fetch HTML, parse via the embedded aonprd:parse sub-DAG, write JSON.

Example: docs scraper — docs:parse

Example: wiki docs — wiki-docs:parse

Full parse taxonomy — aonprd:parse

The complete AONPRD parse DAG — every concept's capability chain compiled from the taxonomy (191 placements). Large by design; zoom or pan to explore.

D&D 5e parse taxonomy — dnd5e:parse

The D&D 5e parse DAG compiled from the shared taxonomy compiler (src/taxonomy/). Content-classified: dnd5e:taxonomy-route inspects the page body to dispatch to the matching concept chain; unrecognised pages route to dnd5e:unknown-end.