> For the complete documentation index, see [llms.txt](https://v2.dataos.info/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://v2.dataos.info/concepts/glossary.md). # Glossary ### A

Term	Definition
AI Activation	The MCP tools and endpoints that let AI assistants discover, inspect, and query governed Data Products through approved metadata and runtime APIs. See AI Activation.
Architecture	DataOS's layered design: a hosted Control Plane, a Data Plane in your cloud, declarative Resources, and the Data Product model built on top.
App	A Resource for deploying containerized applications (for example, Streamlit) inside DataOS with platform authentication, ingress, and observability. See Deploy a data application.
Activation	The process of connecting downstream tools such as BI, APIs, and notebooks to a Data Product's semantic layer so every surface reads the same governed definitions. See Activation.
Assertions	Blocking quality rules that stop bad data from moving forward during build or execution. See Assertions.

### B

Term	Definition
Bento	A Stack for long-running stream processing Workers that read, transform, and write event streams between sources and sinks.
Bifrost	DataOS's access-control component for governing who can access Data Products and platform resources.
Bundle	A Resource that deploys multiple related Resources together in dependency order as a single unit. See Bundle.
BYOC	Bring Your Own Cloud. A deployment model where you provision and operate the Kubernetes cluster and cloud infrastructure; DataOS installs the Data Plane on top.

### C

Term	Definition
CDC	Change Data Capture. An ingestion pattern that streams inserts, updates, and deletes from a source in near real time via binlog or WAL, rather than re-reading full tables on a schedule. See CDC ingestion.
Compute	A Resource that defines Kubernetes node pools or capacity profiles that workloads run on. Tenants attach Compute to a Data Plane so Stacks, Workflows, and Services have somewhere to execute. See Compute.
Consumer	A user, application, or workflow that discovers, queries, and uses a published Data Product. See Consumer journey.
Consumer journey	The end-to-end path a user follows to discover, understand, query, and consume a Data Product. See Walkthrough.
Contract	The published guarantees a Data Product makes about its data: schema, semantics, quality rules, SLAs, and versioning policy. See Define the contract.
Control Plane	The centralized hosted layer that manages authentication, orchestration, metadata, policies, and platform apps. Customer data and compute stay in the customer cloud.

### D

Term	Definition
Data Admin	A Tenant role focused on managing data source access: creating Depots, managing Secrets, and granting connection permissions to builders. See Tenant roles.
Data Consumer	The lowest-privileged Tenant role for discovering and querying published Data Products via Hub, Studio, BI, APIs, or AI. See Tenant roles.
Data Developer	A Tenant role for building and operating Data Products using platform Resources such as Vulcan, Nilus, and Workflows. See Tenant roles.
Data Plane	The execution environment that runs inside your cloud, typically on Kubernetes. It hosts Tenants, workloads, engines, and data processing while keeping enterprise data within your infrastructure boundary. See Provision a Data Plane.
Data Product	A composable unit that delivers data, interfaces, or APIs for a specific business outcome. See Build your first Data Product.
Data Product Hub	The catalog UI where consumers browse, search, filter, and open Data Products, Metrics, and Perspectives for a Tenant. See Discover.
Data Product Lead	A named individual on the Data Product Team who owns the contract externally: SLAs, change approval, deprecation, and consumer relationships.
Data Product Team	The team accountable end-to-end for a Data Product: build, run, support, evolve, and deprecate: without owning underlying platform infrastructure.
DataOS CLI	The command-line tool (`dataos-ctl`) for applying, inspecting, deleting, and debugging Resources, Domains, Stacks, and Lakehouse operations across environments. See CLI reference.
Depot	A declarative connection to an external or internal data source (database, object store, stream, or lake table). Data stays at the source; the Depot registers access and exposes a uniform address for queries and pipelines. See Depot.
Desired state	The target configuration declared in a manifest. The platform works to keep the live Resource aligned with that definition. See Resources overview.
Dimension	A categorical or descriptive field used to filter, group, or slice analysis in a semantic model (for example, customer segment or region). See Semantic models.

### E

Term	Definition
Engine	Where a Data Product's data physically lives and is queried, for example DataOS Lakehouse, Snowflake, BigQuery, Databricks, or Postgres. DataOS is the control layer above the engine; the engine holds the data. See Choose your engine.
External Models	References to existing external or upstream models or assets that a Data Product can use without redefining them. See External models.

### F

Term	Definition
Flare	A DataOS Stack used for Spark-based query and maintenance operations on Lakehouse and Delta/Iceberg data including compaction, reads, and writes.
Freshness	How recently a Data Product's data reflects its sources. Shown in Hub trust signals and part of the product SLA. See Quality.
Full-stack	An adoption pattern where DataOS provides end-to-end capabilities including the native Lakehouse, ingestion via Nilus, build via Vulcan, and DataOS-native consumption surfaces.

### G

Term	Definition
Gateway	The serving interface used to connect to a Data Product or one of its consumption endpoints.
Grant	A governance Resource that assigns permissions to users or groups over specific Data Product assets or platform resources. See Grant.
Governance	The cross-product function that sets classification, access, retention, tier rules, and compliance policies that all Data Product Teams operate within. See Governance.

### H

Term	Definition
Heimdall	DataOS's authentication service. APIs delegate token validation to Heimdall; authentication errors in build and runtime often reference it.

### I

Term	Definition
In-band Validation	Quality validation that runs inside the build or deploy flow and can block bad data before publication. See Data quality.
Instance	A single deployed DataOS environment with its own Control Plane connection, shared Data Plane, and one or more Tenants. The top-level platform boundary for an organization. See Instance vs Tenant.
Instance Admin	The web application Operators use to create Tenants, attach Data Planes and Compute, invite users, and manage Instance-level configuration.

### L

Term	Definition
Lakehouse	A managed storage Resource combining Apache Iceberg tables on cloud object storage with catalog, query, and maintenance services including REST Catalog, Spark, Sherpa, and Flare. See Lakehouse.
Layer	A manifest metadata field (`user` or `system`) indicating whether a Resource is Tenant-scoped or platform-scoped, affecting access and visibility. See Resources overview.
Lifecycle management	How a Resource is created, updated, monitored, and retired. See Resources overview.
Lineage	Shows how data flows from source assets through transformations, outputs, semantic models, metrics, and downstream consumers. See Assets.
Link	A Resource that registers an externally hosted application in DataOS with metadata and deep links so it appears alongside native apps. See Link an external application.

### M

Term	Definition
Manifest	The YAML file that defines a Resource. It declares the metadata, type, and spec the platform should apply. See Resources overview.
MCP	Model Context Protocol. Lets an AI assistant connect to external tools and data sources. In DataOS, the Data Product MCP lets supported AI clients search, inspect, and query governed Data Products through approved metadata and runtime APIs. See MCP overview.
MCP Tools / MCP Endpoints	Interfaces that let AI agents access Data Product data, metadata, schema, quality, runs, and trust context. See AI Activation.
Medallion architecture	A layered modeling pattern (commonly Bronze → Silver → Gold) where raw data is progressively cleaned and made business-ready before exposure in semantic models.
Measure	An aggregatable numeric field in a semantic model, for example total revenue or order count. Metrics and dashboards bind to measures rather than ad-hoc SQL. See Semantic models.
Metadata Extraction	The process of scanning or collecting metadata from source systems so assets become searchable, understandable, and usable in DataOS. See Understanding Metadata Pipelines.
Metis	DataOS's metadata and catalog system. Stores technical, operational, and business metadata and powers search, lineage, and the Data Product Hub.
Metric	A governed time-series definition built from a measure, time column, granularity, and optional dimensions or segments. Defined once and reused consistently; discoverable in the Hub and queryable through Studio, APIs, and MCP. See Business metrics.
Minerva	A dynamic federated SQL query Resource (Trino-based) that exposes configured Depots and Lakehouse data through a single SQL endpoint for exploration and BI. See Minerva.
Models	Transformation logic that turns source data into business-ready data. See Models.
Modern Data Cloud	The hosted SaaS environment where the DataOS Control Plane runs. Your Data Plane connects to it for governance, orchestration, and administration.

### N

Term	Definition
Nilus	DataOS's data movement stack for batch and CDC ingestion from operational systems, SaaS APIs, and warehouses into destinations such as the Lakehouse. See Nilus.

### O

Term	Definition
Operator	The highest-privileged Instance-level role. Provisions Data Planes, creates Tenants, attaches Compute, installs Stacks, and manages platform-wide configuration. See Operator role.
Overlay	An adoption pattern where DataOS layers on top of an existing warehouse or lake you already use. The existing engine stays; DataOS adds Data Product discipline, semantics, and consumption interfaces.

### P

Term	Definition
Perspective	A saved, reusable result created from a validated Data Product query. A first-class consumable view with its own API for business users, applications, and downstream consumption. See Perspectives.
Plan	A preview or dry-run step (via `vulcan plan`) that shows what would change before applying a Vulcan project: dependency graph, diffs, and impact of model or semantic updates. See Plan guide.
Policy	A governance Resource defining access controls, data masking, row and column rules, and compliance constraints scoped to data or products. See Policy.
Policy Engine	The enforcement layer that applies access, masking, and governance rules declared in policies across discovery, build, and consumption surfaces.
Poros	The DataOS operator/controller that provisions and manages dynamic Resources such as Minerva from declarative specs.
Projections	A templating system that injects runtime values (secrets, Depot connection details, Tenant context, environment variables) into container config at deploy time so YAML stays environment-agnostic. See Projections.

### R

Term	Definition
Reconciliation	The continuous process that compares actual state with desired state. When drift appears, the platform applies changes to restore the declared state. See Resources overview.
Resource	A declarative DataOS object that defines connectivity, storage, compute, governance, or workload behavior. Resources are the building blocks of a Data Product. See Resources overview.
REST Catalog	The Iceberg metastore service in a Lakehouse that stores table metadata (schema, snapshots, partitions, file locations) backed by PostgreSQL. See Lakehouse.
RunAsUser	The runtime identity a workload executes as inside DataOS. When set in a resource manifest, the workload runs under the specified user's permissions instead of the creator's. Requires an explicit grant from an Operator. See RunAsUser permissions.

### S

Term	Definition
Secret	A secure Resource for storing credentials (passwords, API keys, certificates) separately from workload YAML. Workloads reference Secrets by name; values are injected at runtime. See Secret.
Segment	A predefined filter on a semantic model (for example, "corporate customers only") that consumers can apply without writing SQL predicates. See Semantic models.
Semantic layer	The tier above individual Data Products where semantic models are registered. Holds meaning and governed definitions, not storage. See Define semantic model.
Semantic model	The business-facing interface for one physical model: dimensions, measures, segments, joins, and descriptions. Consumers query semantic models instead of raw tables. See Define semantic model.
Service	A long-running Resource that stays up and exposes network ports (HTTP/TCP). Used for APIs, databases, streaming endpoints, and other always-on applications. See Service.
Sherpa	A Lakehouse component that translates SQL requests and orchestrates maintenance jobs against the embedded Spark cluster.
SLA	Service Level Agreement. A Data Product's published, measurable commitments on freshness, availability, completeness, and quality. See Quality.
Spec	The `spec:` section in a Resource manifest YAML that holds type-specific configuration such as connection details, compute, stack reference, and DAG steps. See Resources overview.
Stack	A reusable runtime template that defines how a workload step runs (container image, config projection, and stack-specific settings). Workflows and Services reference a Stack. See Stacks.
Studio	The semantic query UI on a Data Product page. Consumers select dimensions and measures, run governed queries, export results, and save Perspectives. See Studio.

### T

Term	Definition
Tag	A flexible label attached to a Resource. Use tags to group, filter, and discover assets across domains, teams, and environments. See Resources overview.
Tenant	An authorization and governance boundary inside an Instance. Use Tenants to separate teams, domains, or environments that share the same platform. See Tenant.
Tenant Admin	The highest-privileged role within a Tenant. Manages users, access, and Tenant resources day to day after an Operator provisions the Tenant. See Tenant roles.
Tunnel	An outbound encrypted connection from the Data Plane to Modern Data Cloud so the platform can be managed without exposing inbound ports on your cluster.

### U

Term	Definition
Unit Tests	Developer tests that validate whether model logic in SQL or Python works correctly before the Data Product is published. See Tests.
Uniform Data Link (UDL)	A stable DataOS address for data in a Depot, formatted as `dataos://[depot]/[source-path]`, used across Stacks, models, and pipelines without repeating credentials. See Depot.

### V

Term	Definition
Volume	A Resource for persistent disk storage attached to workloads. Used when a Service or Worker needs durable local state beyond object storage. See Volume.
Vulcan	DataOS's declarative build stack for Data Products: SQL and Python models, validation, semantic layer, materialization, APIs, and CI/CD-style plan/apply workflows. See Vulcan.

### W

Term	Definition
Worker	A long-running background Resource that processes work continuously without exposing a service endpoint. Used for stream consumers, sinkers, and background loops. See Worker.
Workflow	A batch Resource that runs one or more steps as a directed acyclic graph (DAG), on a schedule or on demand. Used for ingestion jobs, transformations, validations, and backfills. See Workflow.
Workbench	The general SQL exploration UI for querying physical tables across connected sources with governed access and captured lineage. See Workbench.

--- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://v2.dataos.info/concepts/glossary.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.