Skip to content
Glossary

Technical terms, simply explained.

From agent to knowledge graph: the key terms around MAKOR, data and AI. Each term in one sentence, then in a little more detail, with a note on where it appears in MAKOR.

A term is missing? Write to us and we will add it to the glossary.

A

AI ActEU AI Act

European law with rules for AI systems.

Depending on the risk, the regulation sets requirements for AI systems, for example on transparency, logging and human oversight. Provenance for every answer, an audit trail and approval by people support such requirements. Whether and how it applies depends on the deployment.

AI agentAgent

A language model that carries out tasks in several steps.

An agent does not just answer, it acts: it queries data, checks results, prepares reports or requests and gets in touch when something stands out. For this to happen reliably, it needs a clear assignment, limited permissions and a log. In MAKOR, it submits decisions that carry weight to a person for approval.

Audit trailAudit log

A complete log of all processing steps.

An audit trail records which source, which mapping, which rule and which step led to a result. That makes it possible to trace later how an answer came about, for people and AI agents alike.

C

Context

Everything a language model needs to know about the situation.

A language model is only as good as what it learns about your situation. In MAKOR, the context consists of five parts: facts, terms, processes, place and time, and rules.

D

DataCubeData cube

Data organised along several axes.

In Makor Atlas, these are space, time and attribute. Every question is a slice through the cube: an area, a period, a measure. A query only reads the part it needs.

Digital sovereignty

Deciding for yourself where data is stored and who accesses it.

For MAKOR, this means: developed, operated and hosted in Germany, no routing through third countries, no externally controlled services in the data path, open standards instead of a silo.

G

Generative interfaceGenerative UI

The interface builds the right view from a question.

Instead of hunting through menus, you ask a question. Makor Workspace chooses the right tool, such as a network, time series or map, builds the view and shows what it is based on. Every generated view remains a full tool.

GISGeographic information system

Software for data with a spatial reference.

A GIS processes maps, areas, sites and movements. Makor Atlas combines GIS with a DataCube, so that even very large datasets can be analysed in real time.

Graph database

Stores data as nodes and relationships.

Instead of in tables, data sits as a network: customer, has order, concerns site. That makes questions about connections fast, such as who is connected to whom across two steps. Makor Logos comes with its own graph database.

GroundingRAG

A language model gets verified facts before it answers.

With grounding, often via retrieval-augmented generation (RAG), a language model receives the relevant extract from verified sources before every answer. The answer is then grounded, not guessed. In MAKOR, these facts come from the knowledge graph, together with their provenance.

H

Hallucination

A language model makes up something that sounds plausible but is not true.

Language models write what sounds likely. If they lack knowledge, they sometimes fill the gap with a guess. MAKOR counters this: answers rely on your data, and where data is missing, MAKOR says so openly.

Harness

The framework in which a language model or agent works.

Literally the gear that steers power in one direction. In AI development, the harness defines which data a model sees, which tools it may use and which rules apply. Makor Äther is the harness that connects data, processes, meaning and AI.

I

Isolated operationAir gap

Operation without any connection to the internet.

Systems run in a network that is separated from the outside world. This is common in security-critical areas. If you wish, MAKOR runs completely disconnected from the network.

K

Knowledge graphSemantic graph

A network of facts and relationships with a defined meaning.

A knowledge graph connects data from many sources through the terms of the ontology. For people it is a query, for language models and agents it is the memory with verified facts.

L

Language modelLLM, large language model

A program that understands and writes language.

It has learned from a vast number of texts how language works, but not how your organisation works. That is why it needs context: your data, your terms, your rules. In MAKOR it works with this context and names the source.

O

On-premisesYour own hardware

Operation in your own data centre.

The software runs on your hardware, under your control. MAKOR runs on your own hardware, in a private cloud or on a European PaaS.

OWL, RDFS and SHACLOpen standards

Standards for describing terms and validation rules.

RDFS and OWL 2 are used to describe ontologies, SHACL to describe validation rules for data. Because MAKOR uses these W3C standards, your ontology stays portable and your property.

P

People decideHuman in the loop

A person checks and approves before anything that carries weight happens.

The technology recommends, the decision stays with you. In MAKOR, agents submit proposals for approval, and specialists give the data weight and context.

ProvenanceData lineage

Where a piece of information comes from, in which version and from when.

In MAKOR, every piece of information carries its source, version, time and reliability. For every result, Makor Logos supplies the chain: which source, which mapping, which rule.

R

ReasoningInference

Deriving what follows from terms and rules.

If sub-unit E-9 is part of unit V-4 and vehicle F-112 is part of E-9, then F-112 also belongs to V-4. Makor Logos adds such inferences without changing the dataset, and every inference can be traced back to its source.

S

Semantic layer

A layer of meaning over all data sources.

It connects databases, files and streams through shared terms into a queryable graph, without replacing the source systems. In MAKOR, Makor Logos takes on this role.

Semantics

The meaning of data.

A column holds the value 3. Often only a few people know what it means. Semantics makes this meaning explicit, so that people, analyses and AI speak the same language.

SPARQL and GQLQuery languages

Standardised languages for querying graphs.

SPARQL is the W3C standard for RDF graphs, GQL the ISO standard for property graphs (ISO/IEC 39075). Makor Logos supports both, so that your queries are not tied to one product.

Start with a question, not with a platform.

Talk to us about a specific question and the sources you already have today. We will show you how MAKOR makes it answerable.

Contact info@bonn-systems.com