SnowPro Associate: Platform Question 313
Single answer3.5 Explain how to use Snowflake Cortex LLM functions.A data engineering team stores customer support case text in a Snowflake table and wants to generate short summaries directly in SQL for analyst review. The team wants the simplest approach that uses Snowflake-managed large language model capabilities without building an external integration or moving data out of Snowflake. Which solution best meets this requirement?
- A
Use a Snowflake Cortex COMPLETE function call in a SQL query, passing a prompt that includes the case text to generate the summary.
- B
Create an external function that sends each case to a third-party LLM endpoint, because Snowflake SQL cannot invoke LLMs directly.
- C
Train a custom machine learning model inside a Snowpark Container and call it from SQL, because Cortex functions require custom model training before use.
- D
Use a materialized view with a masking policy to automatically convert case text into summaries, because governance features include built-in LLM summarization.
Show answer and explanation
Correct answer: A
Explanation
The best answer is to use a Snowflake Cortex LLM function directly in SQL. For an Associate-level understanding, candidates should know that Snowflake Cortex exposes managed AI functions inside Snowflake so users can perform tasks such as text generation and summarization without exporting data or creating external service integrations. In this scenario, the COMPLETE function is the most appropriate choice because it accepts a prompt and returns generated text, making it suitable for creating short case summaries from support ticket content. By contrast, external functions, custom model training, and governance features like masking policies are either unnecessary or unrelated. This aligns with Snowflake documentation and best practices emphasizing in-platform AI processing when managed Cortex functionality meets the business requirement.
- A. Correct.
Correct. Snowflake Cortex provides SQL-accessible LLM functions, including COMPLETE, that can be used to send prompts and generate text such as summaries directly within Snowflake. This matches the requirement to keep processing in Snowflake and avoid building external integrations.
- B. Incorrect.
Incorrect. External functions can call remote services, but this is not the simplest solution and is unnecessary when Snowflake Cortex LLM functions are available directly in SQL. This option reflects the misconception that Snowflake cannot natively access LLM capabilities.
- C. Incorrect.
Incorrect. Snowpark Container Services and custom model workflows are for more advanced or customized use cases. Snowflake Cortex LLM functions do not require users to train a custom model before generating completions or summaries. This adds complexity that the scenario specifically wants to avoid.
- D. Incorrect.
Incorrect. Materialized views and masking policies are not designed to perform LLM-based text generation. Masking policies are governance controls for data protection, not summarization tools. This distractor targets confusion between governance features and AI features.