holonic — Graph-Native Holonic RDF Systems

Contents

• API Reference
  • HolonicDataset
  • Structural Lifecycle (0.4.2)
  • Ergonomics (0.5.0)
  • Governance & Safety (0.6.0)
  • Consumer Integration (0.7.1)
  • Store Protocol (0.4.0)
  • Model Types
  • Console Model (0.3.1+)
  • Graph-Level Metadata (0.3.3)
  • Scope Resolution (0.3.4)
  • Projection Pipelines (0.3.5)
  • Plugin System (0.3.5)
  • Projection Types
  • Backends
• Projections
  • Two Modes
  • What Gets Collapsed
  • Built-in CONSTRUCT Templates
  • Projection Pipelines
  • Client Integration
  • Utility Functions
  • API Reference
• Backends
  • HolonicStore Protocol
  • AbstractHolonicStore
  • RdflibBackend
  • FusekiBackend
  • Implementing a Custom Backend
  • Optional Native Methods
  • Migrating from GraphBackend (0.3.x)
• CGA Ontology
  • Files
  • Organization (9 sections)
  • Key Classes
  • Key Properties
  • Namespace
  • Other namespaces used
• Holonic and the DOM: A Mental Model
  • Why the DOM is a useful mental model
  • Concept mapping
  • What the DOM model suggests about how holons should eventually be dispatched
  • Where the mapping diverges
  • What the library provides today
  • When to think in DOM terms and when not to
  • Related material

Overview

holonic is a lightweight Python client for building holonic knowledge graphs backed by rdflib, Apache Jena Fuseki, or any SPARQL-compliant quad store.

A holon is an IRI whose associated named graphs exist in an RDF dataset. The dataset IS the holarchy — no separate registry object. Portals are RDF triples discoverable via SPARQL; membranes are SHACL shapes; all state lives in the graph.

✨ Try it in your browser ✨

No installation required. The example notebooks run entirely in your browser via JupyterLite and Pyodide.

13 notebooks covering holons, portals, projections, governance, scope resolution, and dispatch patterns. Notebooks 01-12 run in the browser; notebook 13 (yFiles visualization) requires a local Jupyter install. See 00_start_here.ipynb inside the lab for the full guide.

Quick Start

from holonic import HolonicDataset

ds = HolonicDataset()  # rdflib backend, in-memory

ds.add_holon("urn:holon:source", "Source Data")
ds.add_interior("urn:holon:source", '''
    <urn:item:1> a <urn:type:Widget> ;
        <urn:prop:name> "Alpha" .
''')

print(ds.summary())

Feature Tour

Feature	Introduced	Entry point
Four-graph holon model	0.1.x	add_holon, add_interior, add_boundary, add_projection, add_context
Portal traversal + provenance	0.2.x	add_portal, traverse_portal, traverse, find_path
Membrane validation (SHACL)	0.2.x	validate_membrane
Projection utilities	0.3.0	project_holon, apply_pipeline, project_holarchy
Console-friendly dataclasses	0.3.1	list_holons_summary, get_holon_detail, holon_neighborhood
Graph-level metadata	0.3.3	refresh_metadata, get_graph_metadata
Ontological graph types	0.3.4	cga:HolonicGraph + cga:graphRole; holonic-migrate-registry CLI
Scope resolution	0.3.4	resolve(predicate, from_holon, ...)
Projection plugin system	0.3.5	register_pipeline, run_projection, @projection_transform
Protocol rename + ABC	0.4.0	HolonicStore, AbstractHolonicStore

Backends

• RdflibBackend — Default. Zero infrastructure, pure Python.

• FusekiBackend — Apache Jena Fuseki via SPARQL over HTTP.

• Custom — Implement the HolonicStore protocol, or inherit AbstractHolonicStore for optional-method defaults.

See backends for the full protocol surface.

Migrating from 0.3.x

GraphBackend was renamed to HolonicStore in 0.4.0; the old name remains as a deprecated alias through all of 0.4.x. Migration guide: docs/MIGRATION.md.

Generating Documentation

pip install holonic[docs]
cd docs
sphinx-build -b html . _build/html

Running the Example Notebooks

pixi run serve

This launches JupyterLab with the thirteen example notebooks in notebooks/ available. Notebooks are committed with outputs stripped; executing them locally populates output cells without affecting the committed state. Run pixi run check-notebooks before committing to confirm no outputs leaked in.

Roadmap

See the project README for the headline roadmap. Full requirements tracked in SPEC under R9.11–R9.22.