Getting Started

Squashage takes a pile of JSON records and squashes them into a deterministic RDF graph. Every build is a directed acyclic graph (DAG) executed by @studnicky/dagonizer. Two things drive a run: a config file and a plugin.

• Config squashage.config.json — where the JSON lives, where the graph goes, how many records run at once.
• Plugin plugins/<namespace>/ — the classifiers, the squash node, the per-record DAG. Bring your own; or run bare for a generic rdf:type-only projection.

Install

git clone https://github.com/Studnicky/Squashage.git
cd Squashage
npm install
npm run build

Season the config

A config file is one run. The root holds input, output, and the run knobs directly — nothing else:

{
  "input":  { "basePath": "./input", "format": "json" },
  "output": { "path": "./graphs/out.nq", "format": "nquads" },
  "concurrency": 4,
  "graphs":    { "default": "https://example.org/graph/default" },
  "subjectIri": { "from": "/url", "sanitize": "url-tail" }
}

Drop JSON records into ./input/, point output.path at where you want the graph, and you're done seasoning. Full field reference: Configuration.

Load the plugin

A plugin lives in plugins/<namespace>/ and exports register(dispatcher). It registers the classifiers and squash node for your domain, and ships its own per-record DAG as a *.dag.jsonld file. Full authoring guide: Plugins.

Pass --plugin <namespace> to load it:

npx squashage-dag build \
  --config squashage.config.json \
  --plugin myplugin

No plugin? Squashage runs bare: every record lands in the graph as rdf:type <vocab>Generic. Good for exploring data before you write classifiers.

Run it

npx squashage-dag build \
  --config squashage.config.json \
  --plugin myplugin

Three files land at output.path:

File	Contents
<output.path>	The success graph — every record that made it through.
<output.path-stem>.prov.<ext>	PROV-O activity quads — one prov:Activity per node execution.
quarantine/<bucket>/<id>.json	One file per failed record, bucketed by failure mode.

13,653 Pathfinder records in, 1.74M quads out, 0 failures. Live demo →

Plate it

Turn any .nq graph into a self-contained, offline, cosmos.gl WebGL browser:

npx squashage-dag viz \
  --in ./graphs/out.nq \
  --out my-graph

Open my-graph.html in any browser. No network, no node_modules. Ships with a node inspector, physics panel, pause/resume, and per-concept rulebook properties loaded on demand.

Where to look next

• Walk-through — one record's full journey from JSON to quad.
• Classifier cascade — ten classifiers, one winner per record.
• DAG — the run-scope + per-record DAGs in full.
• Plugins — register(dispatcher) contract, per-record DAG, classifier building blocks.
• Configuration — every config slot.
• Architecture — module map + class lineage.