SnowPro Specialty: Gen AI Question 269
Single answerCORTEX_FUNCTIONS_QUERY_USAGE_HISTORY viewA GenAI platform team uses Snowflake Cortex functions in production and wants to understand which workloads are driving the highest AI-related consumption over the last 7 days. They need a queryable historical source that helps them attribute usage by model and function invocation patterns rather than relying only on warehouse-level metering. Which approach best meets this requirement?
- A
Query the CORTEX_FUNCTIONS_QUERY_USAGE_HISTORY view and aggregate usage by time period, function, and model-related attributes to analyze historical Cortex function activity.
- B
Use QUERY_HISTORY only, because it provides complete Cortex-specific usage attribution including token-level and model-level cost details for every function call.
- C
Use WAREHOUSE_METERING_HISTORY, because Cortex function usage is billed entirely through virtual warehouses and does not require a Cortex-specific history view.
- D
Use TASK_HISTORY, because Cortex function invocations are recorded only when triggered through scheduled pipelines and can then be summarized for all GenAI workloads.
Show answer and explanation
Correct answer: A
Explanation
The best answer is to use the CORTEX_FUNCTIONS_QUERY_USAGE_HISTORY view because the scenario specifically requires a queryable historical source for analyzing Snowflake Cortex function usage over time. This aligns with Snowflake best practice: use purpose-built Account Usage or related history views when you need governance, attribution, and trend analysis for a specific service domain rather than inferring usage indirectly from generic operational views. QUERY_HISTORY can complement the analysis, but it is not the primary Cortex-specific usage history source. Likewise, WAREHOUSE_METERING_HISTORY focuses on warehouse compute consumption, and TASK_HISTORY is limited to task executions. For exam purposes, candidates should recognize when a Snowflake-provided specialized history view is the most appropriate tool for operational reporting and cost attribution of Cortex function usage.
- A. Correct.
Correct. The CORTEX_FUNCTIONS_QUERY_USAGE_HISTORY view is intended for analyzing historical usage of Cortex function queries. In a real-world cost and operations scenario, this is the best starting point when the team wants to inspect Cortex-specific activity patterns over time instead of depending solely on generic warehouse consumption views. It supports workload analysis focused on Cortex function usage history, which is the key need described in the scenario.
- B. Incorrect.
Incorrect. QUERY_HISTORY is useful for reviewing executed SQL statements and can help identify statements that invoked Cortex functions, but it is not the dedicated Cortex usage history source described in the scenario. A common misconception is that generic query history alone provides the best attribution for Cortex usage analysis. For Cortex-specific historical analysis, the dedicated usage history view is more appropriate.
- C. Incorrect.
Incorrect. WAREHOUSE_METERING_HISTORY is for warehouse credit consumption, not for dedicated historical analysis of Cortex function usage. This distractor reflects the common mistake of assuming all AI-related usage must be analyzed through warehouse metering. Even if warehouse usage matters in broader cost governance, it does not replace the Cortex-specific historical view needed here.
- D. Incorrect.
Incorrect. TASK_HISTORY only tracks execution details for tasks. It would not provide comprehensive historical visibility into all Cortex function usage across interactive queries, applications, notebooks, or other invocation paths. This option is attractive if a team uses scheduled pipelines, but it is too narrow for the scenario's requirement.