The Semantic Layer
The semantic layer is where Colrows keeps the meaning of your data - not another copy of the data itself, but a governed graph of what every table, column, metric, and business term means and how they connect. Every governed query and every agent answer compiles against it. It is the heart of the platform, and this part of the docs tells its story end to end.
Every Colrows request runs intent → context resolution → constrained planning → governed execution. The semantic layer is the context that makes the middle two steps possible - so a plan can only ever be a governed, provable one.
The story in three chapters
Read these in order for the whole picture, or jump to the one you need.
What the semantic layer is: a living, governed graph of business meaning - entities, metrics-as-state, events, constraints, and policies - that AI agents and analysts reason over before any data is touched.
2 How it's built Semantic PipelinesHow the graph comes to exist and stays true: scanners gather context from your databases, analytics platforms, data catalogs, dbt, and source code; knowledge, value-resolution, and publishing pipelines turn it into the graph; and change records keep it current as your sources drift.
3 Govern it CatalogThe human's seat at the table: the curator workspace to browse every asset, review what the pipelines proposed, correct a definition, and approve it - with a full, attributed version history behind every edit.
Start with what the semantic layer actually is.
Read Chapter 1: Consensus →