Key concepts

This page defines the frequently used terms and concepts in Holistics, organized by where they sit in the architecture. They map to the structural pieces described in Why Holistics: the expressive semantic layer where business logic lives, and the analytics-as-code infrastructure that keeps it durable.

Source: your data warehouse

The system of record. Holistics is a thin client on top of your warehouse; your raw data never leaves your system.

Concept	Definition
Data Source	A connection to your SQL database, typically a data warehouse like Snowflake, BigQuery, Redshift, or Databricks. Holistics queries it directly; nothing is copied or stored on our side.

Modeling: the semantic layer

Where business logic lives as code. Models, datasets, metrics, and relationships are defined declaratively and queried compositionally.

Concept	Definition
Semantic Layer	The governed layer where business logic lives. Models, metrics, dimensions, datasets, and relationships are defined here as composable code objects. AI, dashboards, drag-and-drop exploration, and embedded analytics all reason from it.
Data Model	An abstract object that sits on top of a database table or query, where business logic is added. Types: Table, Query, Import.
Relationship	The link between data models. Similar to joins or foreign-key relationships in a database, but defined declaratively at the model level.
Dimension	A non-aggregate field that references an underlying column or is derived using non-aggregate functions.
Measure	An aggregating field created with aggregate functions (SUM, COUNT, etc.). Defined inside a model.
Dataset	A curated collection of data models and their relationships. Datasets are the unit of self-service; business users explore datasets, and dashboards and charts build on top of them.
Metric	A full aggregation query written in AQL: composable, reusable, and independent of any single dataset or visualization. Metrics are first-class objects you can combine, transform, and reuse.
AMQL	The analytics-as-code language behind Holistics' expressive semantic layer. Two parts: AML for modeling the semantic layer; AQL for querying it.
AML	Analytics Modeling Language. The typed declarative language the semantic layer itself is written in: models, datasets, dashboards, and relationships. AML is what makes the semantic layer programmable (vs YAML configs).
AQL	Analytics Query Language. The composable query language layered on top of the AML-defined semantic layer. Metrics are first-class objects, not SQL strings. Compiles deterministically to SQL.

info

Metric vs. measure. In the BI world, "metric" and "measure" are often used interchangeably. In Holistics specifically, a measure is defined in a model, while a metric is defined at the dataset level and written in AQL.

Execution: compile and run

How questions become answers. AML and AQL compile to native SQL, executed against your warehouse.

Concept	Definition
Query Engine	The compiler and executor. AQL and AML compile deterministically to warehouse-native SQL. Every query is pushed down to your warehouse, so the warehouse does the heavy lifting.
Inspectable Compiled SQL	The SQL that AQL compiles to is visible. Useful for debugging, performance tuning, and building trust in AI-generated queries.

Consumption: output surfaces

How people and AI consume the same governed metrics.

Concept	Definition
Data Exploration	The interface for exploring and visualizing data: drag-and-drop, SQL editor, or natural-language conversation with AI. All three reason from the same governed semantic layer.
Dashboard	A collection of charts and content blocks that present data to business users. Supports filters, drill-throughs, and exploration. Dashboards themselves are also code, so they're version-controllable.
Holistics AI	Natural-language AI that operates on top of the governed semantic layer. Generates AQL against your governed metric definitions instead of raw SQL against schema, which is why answers stay reliable across follow-up questions. See Why Holistics AI is reliable.
Embedded Analytics	Embed dashboards, self-serve, and AI inside your product. The same governed semantic layer powers customer-facing analytics.

Foundation: analytics-as-code

The cross-cutting substrate. Everything above (models, datasets, metrics, dashboards, permissions) is code, governed by the same engineering practices used for production software.

Concept	Definition
Analytics-as-Code	Every definition is stored as code. This makes the semantic layer durable rather than mutable: business logic gets history, review, branches, environments, inspectable compiled output, and rollback.
Git Integration	Your Holistics code base is powered by Git: branches, pull requests, code review, history, rollback. Use Holistics' built-in repository or connect your own external Git repo.
Environments	Develop in dev, validate in staging, deploy to prod through a real promotion workflow. Multi-environment setups, dynamic data sources, and dynamic schemas keep production stable while you iterate.