WEFT

An open knowledge vault even your AI agent can read.

Documentation and specification: https://dhcraft.org/weft/

WEFT is a lightweight, standards-mapped relation ontology that types the links between atomic notes in a human-curated Markdown knowledge vault. It lets an LLM agent navigate and reason over a typed knowledge graph rather than over flat full text. The ontology reuses established vocabularies and mints a property only where a vault-specific relation provably carries more than association.

The name is a coined word, not an acronym. The weft is the set of cross threads woven through the warp; the typed relations are the weft that turns atomic notes into navigable fabric. Web, text and weave share the Latin root texere, to weave.

Status

Draft, v0.1. The documentation site (docs/) is built and presents the design system and the interactive schema diagram. The ontology serialization (ontology/weft.ttl) and the remaining artifacts are pending the finalized relation layer. The public repository and its Zenodo DOI are not yet created.

What it is, and what it is not

• It is an application profile. It binds to SKOS, Dublin Core Terms, PROV-O, CiTO and schema.org, and adds vault: terms only as rdfs:subPropertyOf a real standard term.
• It is not a triple store and needs no SPARQL. The vocabulary is published as a citable static artifact. The relation layer is consumed by an agent as a typed tool interface, not queried.
• The human-readable source of truth is a Markdown specification kept in the author's Obsidian vault. The Turtle here is the derived machine form.

Structure

docs/        Documentation and specification site (static, self-contained, GitHub Pages)
ontology/    weft.ttl, weft.jsonld           (pending)
shapes/      SHACL shapes for the consistency conditions  (pending)
examples/    worked triples and the navigation-contract legend  (pending)
data/        optional FAIR snapshot of an instance graph  (pending)

The documentation site

docs/index.html is a single, self-contained page. It renders the ontology as a specification and carries an interactive schema diagram (the T-Box) in two views, a class-and-relation map and the provenance model, with click-to-focus on each node to verify its exact mappings and outgoing properties. The layout is curated rather than force-directed, so the schema stays readable. Once the OWL serialization exists, WebVOWL can serve as the automatically generated cross-check.

The design system follows the project idea. A violet and black base alludes to Obsidian as the human Markdown substrate, and a three-part accent (teal, amber, rose) carries the formal RDF triple of subject, predicate and object laid on top.

Instrument

The companion tool vault-graph extracts the links from the vault, types and enriches them with confidence and provenance, and exports the instance graph. WEFT is the vocabulary, vault-graph is the instrument that produces instances of it.

Standards reused

SKOS, Dublin Core Terms, PROV-O, CiTO (SPAR family), schema.org, with vault: subproperties for the relations that carry direction, role or curated membership.

License

The vocabulary and the documentation are released under CC BY 4.0. The Digital Humanities Craft brand assets referenced by the documentation site remain proprietary to Digital Humanities Craft.

Citation

To be assigned on the first release (Zenodo DOI, CFF file).

---

Built by Digital Humanities Craft.