OntoSmith

OntoSmith starts from the schema, not from whatever a model happens to extract. You design a flat ontology (entity and relation types, definitions, boundary examples), a human approves it, and the graph is built under it on Neo4j, with vector and full-text search in the same store.

Why it matters: it separates proposing from judging. Cheap models propose candidate facts, deterministic gates and the Jev decision model judge each one against its evidence, and only judged facts are written, each with its evidence span and provenance. Every build prices itself before it spends. It is deliberately not an OWL editor: no class hierarchy and no inference.

Python, MIT, about 256 stars at the time of listing.