SnowPro Specialty: Gen AI Question 153
Single answerCortex Analyst Verified Query Repository (VQR)A retail analytics team is using Cortex Analyst to let business users ask natural-language questions about sales performance. The team notices that for the prompt, "What were online sales last quarter by region?", analysts consistently want Cortex Analyst to generate the same approved SQL pattern because several similar phrasings have previously produced inconsistent SQL. They want to improve reliability for this class of questions without changing the underlying semantic model. Which approach best addresses this requirement?
- A
Add the approved natural-language question and its validated SQL to the Verified Query Repository (VQR) so Cortex Analyst can use verified examples to guide SQL generation for similar requests
- B
Create a new warehouse dedicated to Cortex Analyst so the generated SQL runs with more consistent compute resources
- C
Replace the semantic model with a larger model context window so Cortex Analyst can memorize more possible question phrasings
- D
Grant business users broader table-level privileges so Cortex Analyst has direct access to more source data when interpreting prompts
Show answer and explanation
Correct answer: A
Explanation
The best answer is to use the Verified Query Repository (VQR) to store approved question-SQL pairs. VQR helps Cortex Analyst produce more reliable SQL for recurring or business-critical question patterns by grounding generation in validated examples. This is especially useful when multiple natural-language variants should map to a known-good SQL structure. The scenario explicitly says the team does not want to change the semantic model, which rules out model redesign as the primary fix. Best practice is to use the semantic model for business definitions and governed data relationships, and use VQR to reinforce trusted query behavior for common or sensitive analytical requests. This aligns with Snowflake guidance for improving Cortex Analyst accuracy through curated semantic context and verified query examples rather than infrastructure changes or broader access grants.
- A. Correct.
Correct. The Verified Query Repository (VQR) is designed to store validated natural-language question and SQL pairs that Cortex Analyst can use to improve generation quality and consistency for similar user requests. In this scenario, the team has an approved SQL pattern and wants Cortex Analyst to prefer that pattern for comparable phrasings without redesigning the semantic model. Adding verified examples to VQR is the most direct and practical solution.
- B. Incorrect.
Incorrect. Warehouse configuration affects execution resources, cost, and performance, but it does not solve the core problem of inconsistent SQL generation for semantically similar prompts. VQR addresses generation guidance; warehouses address compute.
- C. Incorrect.
Incorrect. This reflects a common misconception that generation consistency is primarily solved by increasing model capacity. In Cortex Analyst, reliability for recurring business questions is better improved through governed semantic modeling and verified query examples, not by swapping in a larger context window to "memorize" phrasings.
- D. Incorrect.
Incorrect. Additional privileges do not improve the quality or consistency of SQL generation for a known question pattern. In fact, broader access may increase governance risk. The requirement is to guide Analyst toward an approved SQL formulation, which is what VQR is intended to support.