When you connect Snowflake or Databricks to Perplexity, you can choose to have Computer generate a Data Map of your data warehouse or lakehouse. The Data Map captures important tables and columns, common query patterns, and relationships between objects so Computer can translate natural-language questions into accurate queries.

​
How it works

Computer explores your data model using your connected account’s permissions. It examines schemas, tables, views, and, where available, historical usage patterns to learn which data matters and how it is used. The resulting Data Map captures:
  • Tables and relationships: What key tables contain, how they relate, and which columns to use when joining them.
  • Business definitions: How your organization defines metrics and concepts, including relevant calculations and filters.
  • Query guidance: Common query patterns, recommended tables, and caveats that help avoid incorrect results.
The Data Map can be saved as a skill: reusable knowledge that Computer reads when answering relevant questions. An organization-wide skill makes that knowledge available to everyone in your Perplexity organization; a personal skill makes it available only to you. A shared Data Map does not grant access to the underlying warehouse. Queries still use the permissions of each user’s connected account, or the shared service account if your organization uses one. Because the map itself can contain table descriptions and business definitions, review its contents before sharing it organization-wide.

​
Generating the Data Map

A Perplexity admin can start Data Map generation while setting up the Snowflake or Databricks connector. Generation runs in a standard Computer thread, where you can follow progress, answer questions, and provide context. Computer uses the warehouse identity connected for that run. With OAuth, this is the initiating admin’s Snowflake or Databricks account; with a Snowflake service-account connection, it is the configured service account. The map’s coverage depends on the connected account’s permissions and the scope you choose. Access to query history helps Computer identify common usage patterns. After generation, review the map and choose where to save it:
  • Organization-wide skill: Recommended for shared warehouse knowledge. Members of your Perplexity organization can use it in their Computer tasks.
  • Personal skill: Keep the map in your own skills for individual use or review before sharing.
  • Download: Save the map as a file on your computer without adding it to your skills.
For connection setup, see Connecting Perplexity with Snowflake or Connecting Perplexity with Databricks.

​
Editing and regenerating the Data Map

Admins can edit the saved Data Map skill directly or ask Computer to update it. Use targeted edits to correct a definition, change a recommended table, add a required filter, or document a known issue. When asking Computer to make a change, name the saved skill and explain what should change. For example:
  • Update our Data Map skill to use the new orders table for revenue queries. Verify its columns and joins before replacing the old guidance.
  • Add this definition of an active customer to our Data Map skill, and preserve the existing definitions for other metrics.
If the warehouse has changed substantially, ask Computer to regenerate the map using the connected warehouse and your latest business context. Specify whether you want to refresh an existing area or rebuild the full map.

​
Need more help?

Reach out to your account team or contact us via Help > Contact Support.