DuneSpan / Lexora · remembers and reasons
Lexora · remembers and reasons
The industrial agentic layer, built on a semantic backbone
Lexora connects drawings, documents, standards, ERP records, QC reports and past projects into one secure knowledge layer, then puts governed copilots and agents to work across it.
Copilot
Ask the question your best engineer would ask
Answers cite the drawing, page or record they came from. When the evidence is not there, Lexora says so instead of guessing.
Semantic backbone
The semantic layer is the enabler
Agents are only as good as the meaning underneath them. Lexora keeps that meaning in a versioned registry, not inside prompts.
Ontology
A core spine, extended by industry, then by client, then by instances such as projects, panels and work packages.
Knowledge graph
Relationships between orders, drawings, devices, suppliers, assets and certificates.
Semantic registry
Synonyms, rules and jurisdiction overlays with a lifecycle: draft, certified, deprecated.
Federated queries
One question across ERP, documents and shop-floor data, with access rights applied.
How answers are made
Retrieve. Constrain. Verify. Abstain.
Retrieve
Pull evidence from permitted sources only, revision-aware.
Constrain
Apply the ontology, rules and the user's access rights.
Verify
Check the answer against its sources before it is shown.
Abstain
When evidence is missing or conflicting, say so and route to a person.
Model flexibility
Your choice of model, your choice of where it runs
Frontier models via API
Fastest start, through governed API connections.
Open-weight, self-hosted
Selected open-weight models in sovereign cloud or on your site.
Fine-tuned private model
A model tuned on your domain, kept inside your environment.