Glossary¶
Terms used across this documentation. Each entry gives the short answer; the links go to the page that explains the idea properly.
Data and its shape¶
RDF — Resource Description Framework, the W3C standard for representing data as a graph of statements rather than as tables. It is a data model, not a file format. See Linked data.
Triple — one statement, made of a subject, a predicate and an object: this thing has this relation to that. The smallest unit of RDF. See Every fact is a triple.
Statement — another word for a triple. Used in TriplyDB where a count is involved ("statement count").
Quad — a triple plus the graph it belongs to, so four terms instead of three.
IRI — Internationalized Resource Identifier, a globally unique name for a thing, shaped like a URL. IRIs are what make it possible to say something about the same thing across datasets. See Three kinds of term.
Literal — a value rather than a thing: a number, a date, a piece of text. Literals can only appear as the object of a triple.
Blank node — a term that names something without giving it an IRI, used for structure that has no identity of its own.
Namespace and prefix — a namespace is the shared beginning of a set of IRIs; a prefix is
a short alias for it, so skos:Concept stands for the full IRI. See
Namespaces and prefixes.
Graph — a named set of statements. Data is split into graphs so parts of it can be loaded, replaced or queried independently. See Graphs group statements. In TriplyDB a graph lives inside a dataset — see TriplyDB concepts.
Named graph — a graph identified by an IRI, so a query can ask about it specifically.
Knowledge graph — a graph that holds both the data and the model describing it, built to be connected to other data rather than to stand alone. See Knowledge graphs.
Linked data — data published as RDF with resolvable IRIs, so following an identifier leads to more information about the thing it names. See Linked data.
Serialisation — a file format for writing RDF down. Turtle, TriG, N-Triples, N-Quads, JSON-LD and RDF/XML all express the same model. See Supported file formats.
Querying¶
SPARQL — the W3C query language for RDF. See SPARQL: the basics.
Triple pattern — a triple with variables in one or more positions. Matching patterns against the data is how SPARQL selects things. See Triple patterns and variables.
SELECT query — returns a table of values, one row per match.
CONSTRUCT query — returns RDF rather than a table, building new triples from what it matched. Used for transforming data as well as reading it.
Endpoint — the address a SPARQL query is sent to.
Federation — running one query across more than one endpoint. See SPARQL: construct and federation.
GraphQL — a query language originally from application development. In TriplyDB the schema is generated from your SHACL shapes. See GraphQL.
Models and validation¶
Ontology — a formal description of the kinds of things in a domain and how they relate. See OWL and ontologies.
Class and instance — a class is a category, an instance is one member of it. See What an ontology is.
Property — the predicate part of a triple, seen as part of a model: the kind of relation that may hold between things.
OWL — the Web Ontology Language, for stating what is true of a domain so that new facts can be inferred. See What OWL adds.
Inference — deriving statements that are not written down, from what is. See Inference or property paths?
SHACL — the Shapes Constraint Language, for stating what data must look like and checking it. Where OWL describes the world, SHACL describes the file. See SHACL and SHACL and OWL.
Shape — one SHACL description: which properties instances of a class must or may have, and with what values.
Shapes graph — the graph holding a dataset's shapes. The Editor generates its forms from it.
Violation — a place where the data does not satisfy a shape, reported by validation. See What a violation looks like.
Validation report — the result of validating, itself RDF, so it can be queried. See The report is data too.
Vocabularies¶
SKOS — Simple Knowledge Organization System, for thesauri, taxonomies and controlled vocabularies. See SKOS.
Concept — one entry in a SKOS vocabulary: a term or category, not a class. See Concepts, not classes.
Concept scheme — a collection of SKOS concepts that belong together. See Schemes hold concepts.
Broader and narrower — the SKOS relations building a hierarchy of concepts. See The direction that trips everyone up.
Deprecated term — a class, property or concept marked as no longer to be used, without being removed. See Retiring a concept.
In TriplyDB¶
The product's own vocabulary, defined at greater length in TriplyDB at a glance.
Instance — one installation of TriplyDB. Not to be confused with an instance of a class.
Account — the owner of everything on an instance: either a user or a group.
Dataset — the unit you manage, publish and export. It holds graphs, and may hold assets.
Asset — a file that is not RDF, stored in a dataset so the graph can refer to it.
Service — a copy of a dataset prepared for a particular query language. SPARQL, text search and GraphQL each need one running. See Manage services.
Access level — Public, Internal or Private: who can find and open something. See Access control.
Saved query — a query with a persistent URL, metadata and version history, which is also a REST API. See Work with saved queries.
Story — queries and their results presented as a readable narrative. See Build a data story.
View — in the Editor, a configured perspective on a dataset: which classes and concept schemes are shown, and how. One dataset can have several. See Edit SKOS data.
Node shape — the SHACL description of one class, which is what the Editor builds a form from. A class without one cannot be edited until a shape exists.
Modification history — the record of who changed an instance and when, kept for every change made through the Editor.
AI assistants¶
AI assistant — a chat application that answers questions and carries out tasks in plain language. Once connected to TriplyDB, it can work with the content your account has access to. See Connect an AI assistant.
MCP — Model Context Protocol, an open standard through which an AI assistant connects to other systems and uses the tools they offer. See MCP server.
FlowsPrivate preview¶
The vocabulary of Flows, TriplyDB's pipelines.
Flow — a pipeline built on a canvas in the console: tasks, the connections between them and its variables. It runs by hand, on a schedule or over HTTP. See Flows.
Blueprint — another word for a flow's definition, used in the API (/flows/blueprints) and
on the Flow blueprints admin page.
Flow builder — the canvas a flow is built on, with the task catalogue beside it. See Build a flow.
Task — one step of a flow, such as an import, a query, a validation or an email. See Flow tasks.
Input and output — the datasets, graphs, queries, stories, groups or services a task reads and writes. Tasks act on these directly; no data passes along a connection. See Inputs and outputs.
Connection — an arrow from one task to another, meaning finish the first before starting the second. Connections decide the order; where the cards sit does not. See Edges set the order.
Decision — a task that sends a run down one branch or another, depending on a condition about variable values. See Branches.
Variable — a named value a flow's tasks use, written ?name, given a value for each run.
See Variables.
Masked value — a variable holding a secret, such as an API key, that tasks can use but only permitted people can see.
System variable — a variable TriplyDB sets on every run: who started it, and an API token for
that person. Its name starts with system_.
Preset — a saved set of variable values to start a run with. See Presets and schedules.
Schedule — a flow's recurring start time, with the variable values it runs with.
Run — one execution of a flow, with its status and a record of what each task did. See Run a flow.
Revision — an earlier saved version of a flow, which can be loaded back from Version history.
Artifact — a file a task produces and keeps with its run, such as a connector script's
script.log.
Connector agent — a program run outside TriplyDB, on a server or a laptop, that lets a flow run scripts on infrastructure TriplyDB cannot reach. See Run custom code in a flow.
Handshake token — the secret an administrator gives a connector agent so it can register with the instance. See Connector agents page.
Explained at length in: Linked data · Knowledge graphs · SPARQL · OWL and ontologies · SHACL · SKOS · TriplyDB at a glance · Flows