Q: If someone’s AWS Resources Manager is not working due to a change in their default role policies, how could they enable this feature?
A: The default role is configured for demo purposes, but you can configure a specific OAuth scope role with defined API permissions, which can also be configured on the Snowflake side. When going to a more advanced level of configuration, you can bring in the exact role from the Snowflake side into Entra and provide only the permissions for that particular role within AWS Resources Manager.
Q: Is the semantic text file an index for the AI?
A: The semantic text (asymmetric keys) is not an index — it is predominantly used for doing the token exchange during the handshake with Snowflake. The Amazon Quick index (SPICE index) is totally different: it persists data and is used for getting analytical insights from your data. The asymmetric keys are for establishing the connection with Snowflake and are not related to the index itself.
Q: What is the cost implication of querying Snowflake data for a chat agent — are there costs incurred in Snowflake, AWS, or both?
A: On the Amazon Quick side, there is no additional cost beyond the Author Pro license, which includes the ability to create MCP connections using action connectors. On the Snowflake side, when you ask questions directly to Snowflake via the MCP connector, Snowflake’s Cortex agents perform the analytics and search on your database tables and schemas, which will incur its own cost. The complete details of Snowflake’s Cortex agent pricing were not available at the time of the session.
Q: What does the semantic text file accompanying the Snowflake datasets actually do in the context of the chat agent?
A: When you upload a semantic text file (containing the semantic schema/YAML) to a Space, the chat agent uses it to gain semantic insights from the schema. When the agent tries to understand your question, it references this file to determine which data sources to query — whether that’s a structured dataset or an unstructured knowledge base. Based on your question, the agent uses the uploaded file to derive the information needed before sending it to the RAG pipeline to generate a chat response.