Semantic Catalog
The Semantic Catalog is currently in development. This page previews the planned functionality. We'll update the docs when it launches.
The Semantic Catalog is where you'll define the "truth" of your data — metrics, dimensions, and relationships that ground the AI agent for accurate, consistent answers.
Why Semantic Layers?
Without semantic definitions, the AI has to guess how to calculate metrics. With a semantic layer:
- Consistent metrics — "Revenue" means the same thing everywhere
- Accurate joins — The AI knows how tables connect
- Business context — Terms like "active user" have clear definitions
- Reduced hallucinations — Grounded answers, not guesses
Planned Features
Metric Definitions
Define quantitative measures like revenue, churn rate, and active users using a YAML-based specification.
Dimensions
Specify how to slice and dice your data — by region, time period, segment, and more.
Relationships
Declare how your tables connect so the AI can automatically generate correct joins.
Versioning
Semantic layers will be versioned. Pin notebooks to specific versions for reproducibility.
Agent Integration
When a semantic layer is pinned, the AI agent will use your definitions to write accurate SQL — reducing guesswork and hallucinations.
What You Can Do Today
While the Semantic Catalog is in development, you can still get accurate results by:
- Being specific in prompts — Reference table and column names directly
- Using Conversations — The agent introspects your schema and writes SQL based on actual table structures
- Building Notebooks — Create reusable SQL cells with explicit dependencies
Preview: YAML Specification
The semantic layer will use a YAML format. See the YAML Spec Reference for the planned schema.