Skip to content

COMMERCIAL.md

Latest commit

 

History

History
119 lines (93 loc) · 6.39 KB

COMMERCIAL.md

File metadata and controls

119 lines (93 loc) · 6.39 KB
Raw

Commercial and alternative licensing

The public grants in LICENSING.md are the default: AGPL-3.0-or-later for the pipeline, ODbL-1.0 for the corpus, MIT for the explorer. They are not the only terms available.

DreamLab AI Consulting Ltd holds the rights in pipeline/, static/, docs/, examples/, ontology/ and dist/, and can therefore grant those parts on other terms. This is the standard dual-licensing position: copyleft for everyone, negotiated terms for those who need something else.

Contact

TO BE CONFIRMED. No commercial contact address is published for this repository yet. Until one is, open an issue on DreamLab-AI/knowledgeGraph titled licensing enquiry and state which of the three tracks below applies. Do not put confidential terms in a public issue. Ask for a private channel first.

This section will be replaced with a direct address. Treat the issue tracker as an interim route, not as the intended long-term one.

What can be negotiated

1. A proprietary licence for the pipeline

pipeline/ is AGPL-3.0-or-later. §13 of that licence means that if you modify it and let users interact with it over a network, you must offer those users the corresponding source of your modified version. For a hosted product built on it, that is usually the sticking point.

A proprietary licence can remove it. Typically negotiated:

  • Use of pipeline/ in a closed-source product or internal hosted service, with no §13 or §5 source-disclosure obligation.
  • The right to sublicense the pipeline as an embedded component of your own product.
  • A defined support or update window against a named version.
  • Warranty and indemnity terms: the AGPL grants none, and a commercial licence is where those get negotiated if you need them.

The licence covers 13 files of Python built on rdflib: a JSON-LD-in-Markdown parser, a validator, and six emitters (Turtle, WebVOWL JSON, page API, search index, NGG1 binary graph tiers, backlinks). Read the source before enquiring: it is all in pipeline/, and for many uses the AGPL is simply not a problem and you do not need to pay for anything.

2. Corpus use outside ODbL share-alike

ODbL-1.0 requires that if you publicly use a derived database, you offer that derived database under ODbL as well (§4.4), and that you keep the attribution notice (§4.3). If you produce a derivative and distribute it in a form that would be a Produced Work under §4.6, the share-alike bites on the database behind it.

That is the obligation people most often want lifted. Negotiable:

  • Use of ontology/ or dist/data/* in a proprietary product without the obligation to publish the derived database.
  • Removal or relaxation of the attribution requirement where a notice is impractical.
  • Use for training, fine-tuning or evaluating machine-learning models. The rights of text and data mining are expressly reserved: see the reservation clause in LICENSING.md. A licence for that use is available and is the intended route.
  • Delivery in a form other than the published artefacts: a filtered subset, a different serialisation, or the private superset. The public corpus is the vc:public true subset: 7457 of 7471 pages parsed at the last build, 14 withheld.

Read the corpus honesty note in LICENSING.md before licensing it. It is mostly AI-generated synthetic content produced under human direction, by design, and it exists as an ontology testbed. If you need an authoritative human-authored knowledge base, this is the wrong dataset and no licence will change that.

3. Consulting on similar builds

The method here is transferable: Logseq Markdown with embedded JSON-LD as the authoring surface, rdflib as the build layer, OWL and a binary graph format as the outputs, and a static site plus a WASM explorer as the delivery. The last full build produced 252,974 triples across 7457 classes in 18.0 seconds.

Engagements that have been scoped around this shape:

  • Standing up an equivalent pipeline against an existing Markdown or wiki corpus.
  • Schema and JSON-LD context design, and the validation gates that keep a corpus of this size consistent: the last build reported 0 errors and 961 warnings.
  • Migration of an existing ontology into the JSON-LD-in-Markdown authoring model.
  • Integration work against downstream consumers, including RDF stores and reasoners. VisionClaw is the reference second consumer of this corpus; its GitHubSyncService ingests the same pages into an embedded Oxigraph quad store behind SHACL gating.

Consulting is separate from the licences. It does not by itself grant rights in this repository, and the licences do not by themselves buy any support.

What is not negotiable

  • explorer/ is MIT and stays MIT. There is nothing to sell: the MIT grant already permits commercial and proprietary use, sublicensing and closed distribution. The only obligation is preserving the copyright notice of the WebVOWL authors, and that obligation is theirs to enforce, not DreamLab AI's to waive. See the reasoning in LICENSING.md.
  • Third-party dependencies. The packages declared in explorer/package.json, explorer/modern/package.json and explorer/rust-wasm/Cargo.toml are not DreamLab AI's to relicense. Neither is rdflib.
  • Warranties by default. All three public licences disclaim warranties. A commercial agreement can add them; enquiring does not.

What a licence cannot give you

Read the jurisdiction section of LICENSING.md before paying for corpus rights. The rights basis for the corpus is UK CDPA 1988 s.9(3), which has no US equivalent: Thaler v. Perlmutter, No. 23-5233 (D.C. Cir., 18 March 2025) holds that the US Copyright Act requires human authorship. The UK sui generis database right does not extend to the EEA for a database made by a UK company after 1 January 2021, and has no US analogue at all.

A commercial licence from DreamLab AI is therefore worth most as (a) a contract, which binds regardless of what subsists, (b) certainty about the reserved TDM rights, and (c) access to material and delivery formats that are not published. It is not a guarantee that a third party could not lawfully reach the same public data by another route in another jurisdiction. Anyone selling you the opposite story about a machine-generated corpus is overselling.