Video explainers

The semantic layer, in under a minute each

8 short animated explainers covering the questions that come up first when a team puts a semantic layer over its lakehouse: what the layer is, what a semantic model is made of, how a query resolves through it, and why agents change the stakes. There is no narration and no audio track at all, so every point is on screen and the clips read the same muted, in a meeting, or embedded in your own docs.

• 0:44

  01 What Is a Semantic Layer

  A semantic layer translates physical tables into the metrics, dimensions and relationships a business actually asks about, and the definitions run rather than sit in a wiki.

  • Tables on one side, questions on the other
  • One machine-readable layer in between
  • Definitions, plus the machinery around them
  • The definitions execute, they are not documentation
• 0:44

  02 Metrics, Dimensions, Relationships

  Inside a semantic model: metrics carry their own filters and time logic, dimensions are the ways you slice them, and declared relationships are what keep generated joins from double counting.

  • A metric is a number with its rules attached
  • Dimensions are the ways a metric is sliced
  • Declare the entities once, join them correctly forever
  • Undeclared joins are how totals silently inflate
• 0:44

  03 Governed Metrics

  Metric logic scattered across tools drifts until meetings become arguments about whose number is real. A governed semantic layer makes the definition singular, reviewable and enforceable.

  • One metric name, six different answers
  • Logic copied forward, edited under deadline
  • Treat a definition like code, because it is
  • Governance you can audit, not hope for
• 0:44

  04 How a Semantic Query Resolves

  Resolve, generate, optimize, enforce. Understanding those four steps is enough to read any semantic layer's architecture page, because the products differ in where the steps run, not in what they are.

  • Turn names into a plan
  • Compile the plan into engine SQL
  • Answer from a materialization when one fits
  • Apply the rules before results leave
• 0:44

  05 Headless BI

  When the semantic model lives inside one BI tool, only that tool is governed. Headless BI runs the model as a service so dashboards, notebooks, apps and agents all resolve against the same definitions.

  • Semantics trapped inside one tool
  • The model becomes a service, not a feature
  • Whatever the consumer already speaks
  • Swap the front end, keep the meaning
• 0:44

  06 Semantic Layers and AI Agents

  MCP can connect an agent to your data in an afternoon, and what it finds on the other side decides whether the answer is trustworthy. Raw schemas produce confident guesses; governed metrics produce endorsed numbers.

  • The protocol connects, it does not explain
  • What the agent finds decides the answer
  • Grounding is what closes the accuracy gap
  • Your semantic layer is a knowledge interface
• 0:44

  07 Apache Ossie

  Apache Ossie, incubating at the ASF and formerly Open Semantic Interchange, is a vendor-neutral specification for metrics, dimensions and relationships, so semantic definitions can move between BI, analytics and AI platforms.

  • Lock-in has been unwinding one layer at a time
  • Years of business logic in a private format
  • A neutral spec for metrics and dimensions
  • From an industry initiative to an Apache project
• 0:44

  08 The Semantic Lakehouse

  The lakehouse opened up files, tables and catalogs. Putting the semantic layer next to that open data is what turns storage everyone can read into answers everyone can trust.

  • The lakehouse opened the stack underneath
  • An open table is still not a definition
  • Put the semantics next to the data
  • Open underneath, governed on top

Silent by design: no narration, no music. Hover to preview, click to play, or download any MP4 and reuse it. Terms used in these clips are defined in the knowledge base.