HyLAR

• What it is — HyLAR is a hybrid, location-agnostic, rule-based incremental reasoner for the Web. It combines an rdfstore.js triplestore with an incremental reasoning engine and can run as an npm module, a browserified client, or a server.

• Key concepts —

  • Ontologies are loaded and classified before being queried with SPARQL; loading can replace the knowledge base or retain its existing values.
  • Entailment covers RDFS and a documented subset of OWL 2 RL.
  • Custom business logic is expressed as forward-chaining conjunctive rules: one or more cause triples imply a consequence triple. Rule terms can be variables, URIs, or literals, and predicates can also be comparison operators.
  • The Triple Storage Manager separates direct triplestore operations from inferencing: its query() delegates to rdfstore.js and does not infer at that level.

• How you’d use it — Install hylar through npm, instantiate Hylar, call load(rawOntology, mimeType, keepOldValues), then issue SPARQL with query(). The main API documents Turtle, N3, and JSON-LD input; the storage-manager documentation also describes RDF/XML conversion to Turtle. Add or remove named custom rules with parseAndAddRule() and removeRule(). For deployment, generate a browser bundle with npm run clientize, or install the global server and use its classify, query, and rule endpoints; server options select RDFS versus OWL 2 RL entailment and incremental versus tag-based reasoning, with tag-based reasoning providing proofs.

• LLM angle — none stated

• Pitfalls & lessons — HyLAR supports only a subset of OWL 2 RL and RDFS. OWL 2 RL axiomatic triples are not yet supported, while RDFS axiomatic support excludes axioms related to rdf:Seq and rdf:Bag. Direct queries through the Triple Storage Manager bypass inferencing.

• Verdict — A practical JavaScript reasoner when you need incremental RDFS or partial OWL 2 RL entailment, SPARQL access, and custom forward rules across local, browser, or server deployments.

Sources consulted

• README.md
• github-pages/main.md
• github-pages/stor.md