What changed, and how to install it.

Install it from the VS Code Marketplace. See the install steps.

Semanticus 1.6.0 is about trusting the numbers. Ask your model a question and keep the answer, check a number against its source, and see how Copilot will answer your model before your users do.

• The Query notebook: Explore > Query is a notebook. Each question is one row with its number, one line on how it was read, and one chart. DAX, Table, Speed, Steps, Compare, Save and Explain open when you want them. The notebook is free; asking in plain words is Pro.
• Validation: check a measure against SQL or a number you trust, with both numbers side by side. Accept an expected difference with a note: it stays visible as accepted, never as a pass.
• The Copilot test: your assistant writes the questions a person would ask, a small model answers from only what Copilot sees, and the engine grades each answer by its number. The misses are grouped into a few causes, with the fixes as Plan rows. Every fifth question is held back, so the score cannot be tuned to the test.
• Explain this number: “What makes it up” splits a number by the column that explains it, and a Split by row offers the other columns that also explain most of it.
• Publish that tells the truth: one press sends a rename with every measure, relationship, role and hierarchy that uses it. The page lists no false “new” or “remove it there” rows, and after the press it says “Nothing to publish”.
• Saving and closing: your assistant saves to the model’s own file, and closing VS Code no longer waits for work that is already saved or published.

The full list of changes and fixes is in the changelog under 1.6.0.

The easiest way is the VS Code Marketplace: search Semanticus in the Extensions view. After installing over an older version, use Help, then Restart engine, once. See the install steps.

The product pictures on this site are real captures from the app. Pictures show model states and the work you can review. A picture of fixture data proves presentation, not a live model outcome. We publish no benchmark number or model-grade promise here.