SnowPro Specialty: Gen AI Question 32
Single answerDifferent interfacesA data engineering team has built a document-question-answering prototype on Snowflake using Cortex AISQL functions. Business analysts want to test prompts interactively before the team embeds the workflow into an application, while developers want a repeatable way to call the same capability from code and automate regression tests. Which combination of Snowflake interfaces best meets both needs?
- A
Use Snowsight for interactive prompt testing and SQL worksheets, then use the Snowflake Python API or Snowpark to call the same SQL/Cortex functionality programmatically from application code or test automation.
- B
Use only SnowSQL because it provides the primary interface for prompt experimentation and built-in conversational testing for analysts and developers alike.
- C
Use the Classic Console for interactive LLM experimentation, then migrate to Streams and Tasks for application integration.
- D
Use external notebooks only, because Cortex AISQL functions are available exclusively through notebook environments and not through SQL interfaces such as Snowsight.
Show answer and explanation
Correct answer: A
Explanation
The best answer is to combine an interactive interface with a programmatic one. For Snowflake Gen AI workflows, Snowsight is well suited for exploratory work such as prompt tuning, running SQL, and reviewing outputs. When the same logic needs to be embedded into applications or test suites, developers can call Snowflake through supported APIs, drivers, or Snowpark to execute the same Cortex AISQL-based SQL logic programmatically. This separation of concerns is aligned with Snowflake best practices: use Snowsight for development and validation, then operationalize with code-based interfaces for repeatability and integration. Documentation to review includes Snowflake Cortex AISQL function documentation, Snowsight documentation for worksheets, and Snowpark / Snowflake API documentation for programmatic access.
- A. Correct.
Correct. Snowsight is the practical interface for analysts to run SQL interactively, iterate on prompts, and inspect results. The same Cortex AISQL functions can then be invoked programmatically through code using Snowflake-supported programmatic interfaces such as Snowpark or the Snowflake Python APIs/drivers, which is appropriate for application integration and automated testing. This matches a common real-world workflow: prototype in an interactive UI, operationalize through code.
- B. Incorrect.
Incorrect. SnowSQL is a command-line client for executing SQL, but it is not the primary analyst-friendly interface for interactive Gen AI prompt experimentation. It lacks the collaborative worksheet experience and convenience that Snowsight provides for prompt iteration and result inspection. Teams may use SnowSQL in automation scenarios, but relying on it alone does not best satisfy the analysts' interactive testing requirement.
- C. Incorrect.
Incorrect. The Classic Console is not the recommended interface for modern interactive experimentation with Cortex AISQL. Streams and Tasks are orchestration features for change data capture and scheduled execution, not the primary interface for embedding Gen AI features into an application. This option confuses operational data pipeline features with user and developer interfaces.
- D. Incorrect.
Incorrect. Cortex AISQL functions are designed to be called from SQL, including within Snowsight worksheets, and are not limited to notebooks. External notebooks can be used in some development workflows, but saying the functions are available exclusively through notebooks is factually wrong. This distractor reflects the misconception that Gen AI capabilities in Snowflake require notebook-centric development.