hyperlakeDiscuss a deployment ↗
Videos · 2:19

Semantic Layer Architecture   One Definition for Every Question

Semantic layer architecture defines metrics, dimensions, and joins once, giving BI tools, data apps, and AI agents consistent answers across changing data.

In the full article

  1. Should a semantic layer be physical or virtual?
  2. How does a semantic layer help AI agents?
  3. What is a semantic layer, and how does semantic layer architecture work?
  4. What does a semantic layer define?
  5. How Hyperlake helps
  6. Frequently asked questions
Read the transcript

Auto-generated captions; wording may contain recognition errors.

[music] >> Every analytical data system eventually develops the same problem. The business asks a question and the data team writes SQL. [music] Definitions drift between analysts. Revenue calculated by finance does not match revenue calculated by sales. >> [music] >> The semantic layer exists to solve this once at the right architectural layer rather than repeatedly in every [music] query. A semantic layer is a logical abstraction that sits between raw data and data consumers, translating business concepts [music] into the technical definitions that compute against underlying tables. Instead of every analyst writing their own interpretation of active monthly revenue, >> [music] >> the semantic layer defines it once. Which tables to join, which filters [music] to apply, which calculation to perform. Every downstream consumer, [music] whether a BI tool, a data application, or an AI agent, queries through that definition and gets the same [music] answer. The semantic layer has two architectural modes. A physical semantic layer materializes defined [music] metrics into precomputed tables that downstream tools query directly, producing fast performance but requiring storage. A virtual semantic layer defines the business logic and computes [music] it at query time against the underlying sources, keeping the logic always current at the cost [music] of query time computation. The definitions stored in a semantic layer cover three categories. Metrics are the calculated values the business tracks, revenue, [music] churn rate, and customer lifetime value. Dimensions are the attributes used to filter and slice those metrics, >> [music] >> region, product category, and time period. Relationships are the join [music] paths between underlying tables that produce correct results when combined. For AI systems, >> [music] >> the semantic layer translates natural language questions into technically correct data queries. [music] An agent that queries through a semantic layer works with stable governed definitions rather than raw schemas. [music] When the underlying schema changes, only the semantic layer needs updating and every downstream agent query continues [music] producing correct results.

Start with a workload. Build the environment around it.

Explore example deployments, or see how the platform assembles, deploys, governs and operates the stack.