Glossary
One line each. Follow the link where you want the full explanation.
The core vocabulary
| Term | In one line |
|---|---|
| Resource | Anything your operation refers to by name: equipment, a site, a procedure, a risk, a mitigation, a KPI. → |
| Time series | One signal, measured repeatedly: a stream of timestamped values. → |
| Event | A record that something happened, at a point in time, to something specific. → |
| Data set | A container grouping related data, and the unit of access control. → |
| Relationship | A named, directional link between two resources, contains, feeds, monitors. → |
| Label | The class a resource belongs to, Pump, Site. Your taxonomy. → |
| Metadata | Free key/value context on a resource, series or event. No schema change needed. → |
| External id | A resource's permanent, unique handle, stored exactly as written. Never changes, never reused. On an event it means something else: a shared, non-unique key for the subject. → |
| Name | The readable label a person scans in a list. Changeable at any time. → |
Model concepts
| Term | In one line |
|---|---|
| Ontology | A written-down, machine-readable agreement about what things are and how they relate. → |
| Taxonomy | A classification of what kinds of things exist, arranged as a tree. → |
| Knowledge graph | The ontology filled with your real assets and their real connections. → |
| Digital twin | A live digital counterpart of the operation: the knowledge graph plus the data feeds that keep it in step with reality. → |
| Contextualization | The activity of linking each record to the real-world thing it describes. → |
| Data governance | The rules and accountabilities for data: what exists, who owns it, who may use it, whether to trust it, how long it lives. → |
| Traversal | Answering a question by following relationships from one thing to the next. → |
| Asset | Layer 1 of the model, a physical or logical thing your operation is made of. → |
| Function | Layer 2, what assets do: processes, activities, computations. As an executable block (cleaning, transformation, windowing) it is on the roadmap. → |
| Business knowledge | Layer 3, KPIs, policies, obligations, commercial context. → |
| Lineage | The planned record of what a derived value was computed from, back to raw data. On the roadmap. → |
| Data quality signal | A planned flag saying whether an input was complete, gap-filled, interpolated, flagged or missing. Arrives with lineage. → |
| RDF | A standard way of describing things as subject–predicate–object statements; DataHub's model aligns with it. → |
rdf:type | The formal name for what DataHub calls a label. → |
rdfs:label | The formal name for what DataHub calls a name. → |
| IRI | A globally unique identifier for a thing; the external id is its local part. → |
Working with data
| Term | In one line |
|---|---|
| Value type | How each datapoint is stored, fixed at series creation, drives precision and cost. → |
| Unit | What a measurement is in, bar, °C, kWh. → |
| Step series | A series that holds its value until the next reading, rather than being interpolated. → |
| Datapoint | One timestamped value in a time series. → |
| Nyquist rate | Sample faster than twice the fastest behaviour you care about, or it disguises itself as slower behaviour. → |
| Insights | The console view for charting, comparing and live-streaming time series. → |
| Live mode | Streaming new datapoints onto a chart as they arrive. → |
| Data subscription | A standing request: new datapoints are pushed to you on arrival, with your position held so a reconnect resumes without loss. → |
| Durable subscription | The kind whose position is held by the platform, so nothing is missed across a disconnect. A live tail, by contrast, just watches. → |
| Raw correlation | How closely two traces move together as they are, easily fooled by shared trends. → |
| Whitened correlation | The same comparison after removing each signal's own trend and rhythm. More honest. → |
| Lead / lag | Whether one signal moves before or after another, and by how much. → |
| Coherence | The timescale at which two signals move together most strongly. → |
| Lasting link | Label meaning two signals stay tied together over the long run rather than drifting apart. → |
| Consistent / inconsistent | Whether a relationship holds across the whole range analysed, or only part of it. → |
AI agents
| Term | In one line |
|---|---|
| AI agent | Software that reasons over your data and acts or recommends, far more accurately when it can use the knowledge graph. → |
| Machine learning | Working the rule out from examples instead of writing it down. → |
| Feature | An input to a model: a number that means something, such as vibration energy or hours since last start. → |
| Label | The recorded answer for a past example, which in this platform is usually an event. → |
| Supervised / unsupervised | Learning from labelled examples, against finding structure with no answers given. → |
| Overfitting | Memorising the examples rather than learning the pattern. Excellent on old data, useless on new. → |
| MCP | Model Context Protocol, the open standard by which an AI assistant discovers and calls a system's tools. → |
| MCP server | The side that publishes the tools. DataHub has two, on the same token and checks as every other call. → |
| Tool | One named capability an agent can call, such as event_filter, with typed arguments and a description written for the model. → |
| Console assistant | The console's own read-only agent, reached from Ask AI, answering from your tenant's data. → |
| Agent loop | Observe, decide the next step, act, check, repeat until the goal is met or a limit stops it. → |
| Human gate | The approval an agent's draft passes through before anything irreversible happens. → |
| Perception agent | Watches continuous signals and emits typed events when something significant happens. → |
| Triage agent | Collapses an alarm flood into the causal chain and one clear action. → |
| Coordinator | Sequences several agents' output so their actions do not collide. → |
| Sourced answer | An agent answer where every claim points at the resource, event or value behind it. → |
| Dynamic incentives | Steering agents by scoring observed outcomes and feeding the score back. → |
Integration
| Term | In one line |
|---|---|
| CDC | Change data capture, tailing a source database directly so its changes flow in. → |
| Publication | The name a source database exposes its change log under, needed to configure CDC. → |
| Flush interval | How often a CDC integration writes captured data down. Shorter is fresher; longer is cheaper. → |
| Retention | How long measurement history is kept. Automatic expiry is on the roadmap, so today nothing ages out and volume is cumulative. → |
| Access token | A short-lived credential carrying your identity and permissions. Treat it like a password. → |
| Service account | A non-human identity used by an integration, with its own narrow permissions. → |
Administration
| Term | In one line |
|---|---|
| Tenant | A separate world with its own databases and file storage, and its own credentials where the deployment configures them. → |
| Organisation claim | The field in a user's token that resolves which tenant a request belongs to. → |
| Organisation (Keycloak) | A container in the identity provider holding one tenant's members, identity providers and groups. An organisation may have several identity providers; a provider belongs to exactly one organisation. → |
| Policy | A rule the platform holds: naming, lifecycle, access, requirements. Naming is enforced today; the other three are recorded, and enforcement is on the roadmap. → |
| Naming policy | The rule fixing what a new external id may look like, checked on every write. Set per tenant, overridable on one data set, and never applied to events. → |
| Near-duplicate guard | The always-on check that refuses pump-a-01 beside an existing pump_a_01, anywhere in the tenant. → |
| Finding | The record left when a policy warns rather than refuses: the write went through and a data steward is expected to look at it. → |
DATAHUB_ACCESS | The baseline role required to use the platform at all. → |
DATAHUB_ADMIN | The escape-hatch role granting read and write on every data set in every tenant. → |
| Wildcard data set group | /datasets/*/read or /datasets/*/write: an organization group granting read or write on all of that tenant's data sets. → |
| Rate limit | The per-minute request budget a tenant, and each identity inside it, may spend. Over it is refused until the next minute. → |
| Ingest quota | The daily allowance for events, entities, relationships, measurements and bytes written, reset at midnight UTC. → |
| Lifetime ceiling | How large a tenant may grow in total rather than how fast. Off unless a deployment switches it on, and raised on request rather than by waiting. → |
tenant_limits | The one-row table in each tenant's schema that overrides any deployment default limit, live and without a restart. → |
| Identity provider | The external system holding user accounts and issuing tokens. → |
| Air-gapped | Running with no internet connection at all, fully supported. → |
Industrial standards
| Term | In one line |
|---|---|
| ISA-5.1 | Instrument tagging: PT is a pressure transmitter, FIC a flow indicating controller. → |
| IEC/ISO 81346 | Reference designations separating aspects with prefix signs: = function, - product, + location. → |
| ISO 15926 | The life-cycle data model for process plants; Part 8 maps it onto RDF and OWL. → |
| NORSOK Z-DP-002 | The Norwegian shelf coding system, breaking an installation into systems, subsystems and item tags. → |
| CFIHOS | Fixes the equipment classes and attribute names handed over from project to operations. → |
Roles in a rollout
| Term | In one line |
|---|---|
| Sponsor | Chooses the first question and unblocks access to source systems. → |
| Domain expert | Says what the things in your operation are and how they relate. The critical role. → |
| Data steward | Owns naming conventions, data set boundaries and model quality over time. → |
| Administrator | Runs the platform: users, permissions, retention, availability. → |
| Integrator | Gets data out of source systems and into DataHub. → |
Go deeper
- The four building blocks: the vocabulary in context
- What is a digital twin?: what these pieces add up to
- What is an ontology?: the model concepts explained properly
- Frequently asked questions