Find a semantic layer

Data & AI → Semantic Layers lists every semantic layer you can see. Once a workspace has more than a handful, the search box, the filters and the columns are how you find the one you want. To start a new one, select Create; see Create a semantic layer.

Search and filter

The search box matches a layer’s name, its data connection and its description, so you can look for “sales” whether that word is in the title or only in the description.

Three filters sit beside it:

• All data connections — narrow the list to layers built on one connection.
• All creators — narrow the list to layers created by one person. Handy when you know a colleague built the model but not what they called it.
• Sort order — newest first by default.

Filters combine with the search box rather than replacing it.

What the columns tell you

Contents — what the layer holds, as two pairs: dimensions and measures on one line, tables and fields on the other. A layer showing Empty has been created but not yet built out.

Published Workbooks — how many published workbooks are built on this layer. See Published workbook counts below.

Data connection — the connection the layer reads from. Select it to open that connection.

Version — the layer’s current version, or Not versioned if no version has been taken yet. See How versioning works.

Last updated and Created — when, and by whom.

You can sort by name, published workbooks, data connection and last updated, either from the toolbar or by selecting the column heading.

Check a layer

Check, on the row of a layer you can edit, checks it without opening it. It compares the layer’s tables and fields with the warehouse, and looks for errors in its own measures, dimensions and questions. It never publishes anything.

What it finds appears as a mark beside the layer’s name: Healthy when its sources and its model are both clean, a count of errors or warnings, or 1 missing for a table the warehouse no longer has. The day of the last check sits under Check. Hover over the mark for the findings, and for when the layer was checked and by whom. Select it to open the layer, where the Issues panel lists every finding with its fix.

The last check of each layer is kept, so the mark is still there the next time you open the list. That includes a check a colleague ran, and a fresh source check run by Nash or another MCP client. If the layer has changed since, the mark is drawn as an outline rather than filled in, and its hover says so, because what it found may already be fixed. Select Check again to see where the layer stands now.

Only people who can edit a layer see its mark and its Check.

Nash and other MCP clients read the same last check as lastCheck, so a routine that looks after your layers can see which ones need attention, and when each was last checked, without checking them all again. See Nash MCP setup.

Published workbook counts

The Published Workbooks column counts the published workbooks that build on each layer. It is worth knowing exactly what that number means before you act on it.

It counts published workbooks only. A workbook someone is still drafting against the layer is not included. A workbook with two published versions that both read this layer counts once, not twice.

Zero means zero. A layer showing 0 has no published workbooks on it — usually a safe sign if you are thinking about deleting or restructuring it.

An em dash (—) means unknown, not zero. It appears when the count could not be worked out: the list could not be read, or the workspace holds too many published workbooks to total exactly. When you see it, sorting by this column is switched off rather than sorting on a partial answer. Treat a dash as “check before you change anything”.

The number can lag by a few minutes. Counts are worked out once and reused for about five minutes, so a workbook published moments ago may not be reflected yet. Reload the page after that to see it.

Who sees this column

The Published Workbooks column is shown to workspace owners and administrators. If you hold another role the column is not there at all, rather than empty.

That is deliberate. The count is worked out using your own access, so if you cannot see every workbook in the workspace, the total would be of the workbooks you can see while looking like the workspace’s full total. Rather than show you a number that reads as something it is not, the column is left out. If you need the figure, ask an administrator.

Related

• Create a semantic layer — an empty layer, a copy, a Snowflake semantic view or a YAML file, from the Create menu.
• Working in the Semantic Layer Editor — what to do once you have opened one.
• Data connections — the Semantic layers column there answers the same question from the other direction.