LLMs.txt directory

Integrate with Snowflake Semantic Views

ThoughtSpot integrates with Snowflake Semantic Views, allowing you to leverage semantic models defined in Snowflake in ThoughtSpot. This enables centralized metric and business logic management in Snowflake, while providing a seamless analytics experience in ThoughtSpot, by leveraging Snowflake’s semantic layer. The integration is designed for flexibility, efficiency, and to improve data governance and discoverability.

The integration supports both direct connection to Snowflake and YAML file uploads for flexibility.

Administrators can manage and resync Semantic Views as needed to reflect changes made in Snowflake.

The Snowflake Semantic Views is in Early Access. To enable this feature, contact ThoughtSpot Support.

Features:

  • Connect directly to Snowflake Semantic Views from ThoughtSpot.

  • Import semantic models, including metrics, formulas, and column descriptions, into ThoughtSpot.

  • Use imported semantic views as data models for building dashboards, SpotIQ queries, and analytics.

  • Column-level metadata and documentation from Snowflake are automatically pulled into ThoughtSpot.

This integration is separate from dbt integration; you can choose to use either or both, depending on your data modeling strategy.

To edit a Snowflake semantic integration, do the following:

  1. Go to the Data workspace tab and select Semantic integrations.

  2. Expand the more menu for the semantic integration that you want to edit and select View Model.

  3. Click Edit model to open the Data model editor.

Prerequisites

Please confirm that the user has access to all underlying tables referenced by the semantic view.

Integrating with Snowflake

  1. Define Semantic Views in Snowflake:

    1. Create semantic views in Snowflake using their native tools.

    2. Add business logic, metrics, and column descriptions as needed.

  2. Connect ThoughtSpot to Snowflake:

    1. In ThoughtSpot, select Snowflake as your data source.

    2. Choose the relevant database and select the semantic view you want to use.

    3. ThoughtSpot will import the semantic view, including all defined metrics and metadata.

  3. Use in ThoughtSpot

    1. The imported semantic view appears as a data model in ThoughtSpot.

    2. Build dashboards, run queries, and use SpotIQ on top of the imported model.

    3. All business logic and metrics defined in Snowflake are available for analytics in ThoughtSpot.

When a semantic view is imported, tables are imported into ThoughtSpot using the aliases defined in the semantic view. If a table already exists in the connection with a different name, it is automatically renamed to match the alias during the semantic view import.

Import Spotter memory from semantic views Early Access

A Snowflake semantic view can include verified_queries, which pair example questions with the SQL that answers them, and ai_sql_generation instructions, which guide how SQL is generated for the view. When this feature is enabled, ThoughtSpot imports these fields with the semantic view. It uses AI to translate them into Spotter memory on the resulting Model. Spotter then applies this memory when answering questions on that Model.

Importing Spotter memory from semantic views is in Early Access and is off by default. To enable it, contact ThoughtSpot Support. When this feature is off, ThoughtSpot imports the semantic view but omits the verified_queries and ai_sql_generation fields.

ThoughtSpot imports verified_queries, ai_sql_generation instructions, or both, depending on what the semantic view contains.

ThoughtSpot imports memory whether you connect to the semantic view directly or upload a semantic view YAML file.

You can import memory even if Spotter 3 is not enabled for your Org. However, Spotter uses the imported memory to answer questions only after Spotter 3 is enabled. After import, you can review the imported memory the same way as other memory on the Model. For more information, see Manage memory.

Any changes you make to imported memory in ThoughtSpot are lost when you re-import the semantic view.

Limitations

If the semantic model has two or more tables, then the tables must have joins defined.

Snowflake window functions are not supported in ThoughtSpot.

If you do a resync, you will lose your locally-made edits and Snowflake will be considered the source of truth.