SnowPro Specialty: Gen AI Question 270
Single answerCORTEX_FUNCTIONS_QUERY_USAGE_HISTORY viewA Snowflake administrator needs to analyze which teams are driving the highest usage of Cortex AI functions over the last 7 days so the company can allocate chargeback costs. The teams are mapped to Snowflake roles, and leadership wants a report based on actual Cortex function invocations rather than general warehouse activity. Which approach best meets this requirement using the CORTEX_FUNCTIONS_QUERY_USAGE_HISTORY view?
- A
Query the CORTEX_FUNCTIONS_QUERY_USAGE_HISTORY view for the last 7 days, aggregate usage by the role associated with each query, and summarize the Cortex function activity from that result set.
- B
Query QUERY_HISTORY for the last 7 days, group by warehouse name, and use total execution time as a proxy for Cortex function usage by team.
- C
Query the CORTEX_FUNCTIONS_QUERY_USAGE_HISTORY view and group only by user name, because roles are not relevant when tracking Cortex function usage for chargeback.
- D
Use WAREHOUSE_METERING_HISTORY to identify the warehouses with the highest credits in the last 7 days and attribute those warehouse costs directly to Cortex AI function usage.
Show answer and explanation
Correct answer: A
Explanation
For Cortex-specific monitoring and chargeback, the best practice is to use the dedicated usage surface that records Cortex AI function activity rather than relying on general-purpose query or warehouse history. The CORTEX_FUNCTIONS_QUERY_USAGE_HISTORY view is intended for analyzing Cortex function usage at the query level, making it the right source when the goal is to determine who used Cortex features and how much. In this scenario, because the business maps teams to roles, summarizing activity by role is more appropriate than summarizing only by user or warehouse. By contrast, QUERY_HISTORY and WAREHOUSE_METERING_HISTORY are broader account-usage sources and can support complementary analysis, but they do not provide the most direct basis for measuring actual Cortex function invocations. This reflects Snowflake best practice: use the most specific telemetry source available for the workload you need to audit or charge back.
- A. Correct.
Correct. The CORTEX_FUNCTIONS_QUERY_USAGE_HISTORY view is designed to analyze usage of Cortex AI functions at the query level. For a chargeback scenario where teams are mapped to roles, aggregating the relevant records over the last 7 days by the role tied to the query is the most direct way to report actual Cortex function activity. This aligns with the requirement to measure real Cortex function invocations instead of inferring usage from broader compute metrics.
- B. Incorrect.
Incorrect. QUERY_HISTORY contains general query execution metadata, but grouping by warehouse and execution time only gives an indirect estimate of activity. It does not specifically isolate Cortex function usage, which is the core requirement in this scenario. This is a common mistake when teams rely on generic query or compute telemetry instead of Cortex-specific usage views.
- C. Incorrect.
Incorrect. Grouping only by user name may be useful for some investigations, but it does not satisfy the stated business requirement because the company maps teams to roles for chargeback. Ignoring role context can produce a report that is misaligned with the organization's allocation model.
- D. Incorrect.
Incorrect. WAREHOUSE_METERING_HISTORY is useful for analyzing warehouse credit consumption, but Cortex AI function usage should not be attributed solely from warehouse metering data. The question explicitly asks for actual Cortex function invocations, and warehouse-level cost data is too broad for that purpose.