THE STATUS QUO
Embeddings & similarity search
Entities & explicit relations
Two stores means two truths. Embeddings drift from the graph, identity gets duplicated across systems, the sync glue turns brittle with every schema change, and every query ends in fragile post-filtering: accuracy quietly falling through the cracks.
HYBRID OXQL · ONE PASS
// narrow by logic, then rank by vectors
FIND ?x
WHERE ?xIS-A Beer
NEAR ?x TO"crisp pale lager"
LIMIT 5
→ 5 results · single pass · 12ms
The ontology filters candidates during the search, not after. No noise, no post-filtering, no second system.
Perception, cognition, and expression: three systems that pass meaning around a single loop.
It reads the world so your database can understand it.
The perception layer: turning text, code, PDFs, and the web into structured meaning.
Memory that reasons.
The neurosymbolic core that stores, connects, and reasons over everything it knows.
Where agents meet the mind.
The expression layer where agents query, prompt, and act on the database.
The neurosymbolic features ride on a storage engine that behaves like one: durable, transactional, recoverable.
Write-ahead log, crash recovery, point-in-time recovery.
AS OF snapshot queries over your knowledge's history.
Lock-free reads under concurrent writes.
Every entity is one ID to reasoner and vector index: zero-copy joins.
Insert text, get vectors: local ONNX, no external service.
Train TransE / RotatE structural vectors in-engine.
Embedded graph UI, zero runtime dependencies.
Node.js & Python.
Agents ask in language. Oxid turns the question into an OxQL query, reasons over the graph, and hands back an answer it can defend.
One server, every client. Your agent gets typed tools, not a REST wrapper.
Which suppliers shipped late to our Berlin plant last quarter, and what did they have in common?
FIND?supplier
WHERE?sSHIPPED-TOBerlin_Plant
AND?sDELAYED-BY> 0
NEAR?causeTO"shared delay"
All late shipments trace to one shared upstream: the Hamburg hub, after the March port closure. The link came from the graph, not a keyword match.
Perception, reasoning, and memory in one system.
This is the future of databases.