Crawl Snowflake AI models
Atlan can discover and catalog AI models—and their registered versions—stored in the Snowflake Model Registry. Once crawled, model assets appear in Atlan alongside your other Snowflake assets, with upstream lineage built automatically using Snowflake's native lineage function. Model crawling requires the Direct extraction strategy with the Information Schema method and isn't supported with the Account Usage method or any offline extraction strategy.
Prerequisites
Before crawling AI models, make sure you have:
- A Snowflake account with models registered in the Snowflake Model Registry
- Set up the Snowflake connector in Atlan
- Crawled Snowflake assets at least once
Permissions required
In addition to the standard Snowflake connector permissions, the Atlan service account requires:
USAGEon the warehouse, and on each database and schema containing modelsUSAGEonINFORMATION_SCHEMAfor the Information Schema extraction methodSELECTon tables and views used for lineage relationshipsUSAGEon each model—required both forINFORMATION_SCHEMA.MODEL_VERSIONSto list models and forSNOWFLAKE.CORE.GET_LINEAGE()to retrieve their upstream dependencies
Grant model usage permissions with:
-- Grant USAGE on all existing models in a schema
GRANT USAGE ON ALL MODELS IN SCHEMA <database_name>.<schema_name> TO ROLE <atlan_role>;
-- Grant USAGE on future models (recommended)
GRANT USAGE ON FUTURE MODELS IN SCHEMA <database_name>.<schema_name> TO ROLE <atlan_role>;
Repeat these grants for each database and schema containing models you want to crawl. Without the FUTURE grant, every newly registered model requires a manual re-grant before Atlan can crawl it.
For the full breakdown of each privilege and what it enables, see Permissions for Snowflake AI models.
Configure crawler
To configure the crawler for AI models:
- Follow the standard Crawl Snowflake steps.
- When selecting the extraction method, choose Information Schema.
- Under asset filters, specify which databases or schemas to crawl:
- To include specific databases or schemas, configure Include Metadata filters.
- To exclude specific databases or schemas, configure Exclude Metadata filters.
- If no filters are set, Atlan crawls all databases and schemas accessible to the service account, including any models within them.
- Run the workflow.
After the workflow completes, AI Model and AI Model Version assets appear in Atlan under the crawled schema. Navigate to any AI Model Version asset to see its upstream lineage, resolved by GET_LINEAGE() up to 5 levels deep.
See also
- How Atlan builds lineage for Snowflake AI models: Understand the lineage function used, depth, and upstream asset types resolved.
- Permissions for Snowflake AI models: Full breakdown of each privilege required and what it enables.