If your semantic layer uses a Snowflake connection, you can download the SQL that creates the same model as a semantic view in your Snowflake account. Astrato does not run anything in Snowflake: you review the SQL and run it yourself. Downloading the SQL does not use Nash or any AI credits.
Before you start
The semantic layer must use a Snowflake connection. The export does not appear for other connections or for local files.
Resolve any validation errors in the semantic layer. You do not need to publish: the SQL is generated from the current draft.
In Developer View, apply or reset any edits you have typed. The export fails while the editor has unapplied changes.
Download the SQL
Open the semantic layer and go to Developer View.
Open Export in the toolbar and choose Export Snowflake Semantic View SQL.
Your browser downloads a
.sqlfile named after your semantic layer.
By default the semantic view is created in your connection's default database and schema and named after your semantic layer. If the connection has no default database or schema, the semantic view uses the one of the first table in your semantic layer. To use a different database, schema or name, edit the CREATE SEMANTIC VIEW statement in the script before you run it, and change any other statement that refers to the same database, schema or name. Some scripts create supporting views before that statement.
Definitions travel with the view
Descriptions you wrote in Astrato become comments on the matching tables, dimensions, and measures in the semantic view. The script also lists every described dimension and measure under Definitions in the semantic view's own comment, after the semantic layer's description. The definitions are then available in Snowflake as part of the view, and nothing extra is created in your schema. If the glossary is very long, the last entries are left out of the comment; the per-item comments are always kept.
Run it in Snowflake
The script's header comment includes a link to your Snowflake account when the connection records one.
Open a worksheet and paste or open the script.
Run every statement in order, for example with Run All in Snowsight. Some scripts create supporting views first, and the semantic view depends on them.
Use a role that can create semantic views in the target schema and read the tables the script lists.
What is left out
Snowflake semantic views cannot express everything a semantic layer can, such as period comparisons, bucket dimensions, row filters on tables and some advanced measures. The script includes everything that has an exact Snowflake form. Anything that does not is left out, and the script's header comment lists each item that was left out.
Business questions are left out too: they would become the view's verified queries, which need SQL written against the view's own tables. They appear in the header comment. Add them yourself, or ask Nash to write them from the semantic layer.
To keep the semantic view in step with your semantic layer, download the SQL again after you make changes and run the new script.