Overview: semantic views in the studio
A semantic view is a Snowflake object built from governed metrics. Once it exists, it can be queried through Cortex Analyst, Cortex Agents, and standard Snowflake tools. This section covers what a semantic view is and how you query it through Cortex Analyst.
Where a semantic view comes from
A semantic view is built from the metrics of a deployed data product. The Data Product Studio generates metrics for a data product from its Facts and Dimensions, you review and edit them, and those governed metrics build the semantic view. The semantic view carries the same process context and lineage as the data product it came from, so a question answered against it traces back to the process it measures.
The articles that walk through generating metrics and building a semantic view are being written and are not part of this set yet. The high-level picture is enough for this overview: a deployed data product produces metrics, and those metrics produce a semantic view you can query.
What you can do with it
Once a semantic view is deployed, you query it in natural language through Cortex Analyst. See Query a semantic view with Cortex Analyst.
A semantic view can also be refined so it answers questions more reliably. Synonyms map alternative phrasings to a metric, verified queries supply known-good examples, and custom instructions tell Cortex how to read ambiguous prompts. Articles on each of those are coming.
Note: A semantic view is a Snowflake object. Querying it needs access to the deployed data product and to Cortex in your Snowflake account.
See also
- Sibling: Query a semantic view with Cortex Analyst
- Reference: Semantic view in the Glossary