Ontology reasoning fused with vector search, in one process, one query language, one identity space. The neurosymbolic core that holds the mind together.
Research preview · v0.8.5Candidates are pruned against your EL++ ontology while the HNSW graph is traversed, not in a post-filter afterward. One identity space means the reasoner and the vector index share the same entity IDs: zero-copy joins, no glue layer, no drift between two stores.
MATCH?dIS-ADiseaseNEAR?symptomThis is a real ontology exactly as oxid-db returns it: classes, individuals, and every relation between them. The reasoner has already classified it: hover any neuron and the chips you see are inferred memberships, derived by EL++ subsumption, not stored.
Drag to rotate · hover a neuron to inspect · every lit synapse is a subclass or membership relation
Ontology constraints and vector search live in the same statement: symbolic predicates and semantic nearness, resolved in a single pass.
// retrieve cardiac drugs semantically near a target, constrained by ontology
MATCH(?drug:Drug)-[:TREATS]->(?c:Condition)
WHERE?cIS-ACardiovascularDisease
AND?drugNEARembed("selective beta blocker")
AND?drug.phase>=3
RETURN?drug,similarityASscore
ORDER BYscoreDESCLIMIT20
AS OF'2026-06-01T00:00:00Z'
Transactional guarantees and recovery you can audit: the substrate the mind is written to.
Every change is journaled before it lands.
Replays the log to a consistent state on restart.
Readers never block writers, ever.
Rewind the store to any committed moment.
Read your knowledge as it stood, then.
A single self-contained Rust binary with an embedded knowledge-graph web UI and zero runtime dependencies. Drop it in and it runs: reasoner, vector index, embeddings, and console, all in one process.
One engine for logic and similarity. Start building on the neurosymbolic core.
Research preview · v0.8.5