SnowPro Specialty: Gen AI Question 130
Single answerHelper functionsA data engineering team is building a customer-support assistant in Snowflake that uses Cortex LLM functions to draft responses from case notes. Before sending the prompt to the model, they need to insert customer-provided text safely so that quotes, line breaks, and other characters do not break the SQL statement or corrupt the prompt structure. They also want a solution that is easy to reuse across multiple prompt templates. Which approach is the most appropriate?
- A
Use a Snowflake helper function to safely embed the dynamic text into the prompt template before passing it to the Cortex function.
- B
Concatenate the raw customer text directly into the SQL string because Cortex functions automatically sanitize all prompt content.
- C
Convert the customer text to uppercase before concatenation so special characters cannot affect the SQL or prompt formatting.
- D
Store the customer text in a temporary table first, because helper functions can only be used with structured JSON inputs and not plain text.
Show answer and explanation
Correct answer: A
Explanation
The best answer is to use a Snowflake helper function for prompt construction. In real-world GenAI workflows, prompts often include dynamic user or application data, and directly concatenating those values into SQL strings is error-prone. Helper functions exist to support safer and cleaner prompt assembly, improving maintainability across templates and reducing issues caused by embedded quotes, special characters, and formatting. This aligns with Snowflake best practices for working with Cortex AI SQL functions: build prompts in a structured, reusable way rather than relying on fragile manual concatenation. Candidates should recognize that the helper function's role is not model inference itself, but safe and practical prompt preparation for downstream Cortex calls.
- A. Correct.
Correct. Snowflake helper functions are designed to make prompt construction safer and more reusable, especially when inserting dynamic values into prompt templates. In this scenario, the key issue is preserving text correctly and avoiding malformed SQL or prompt content when user-supplied values contain quotes, newlines, or similar characters. Using the appropriate helper function is the best practice for reliable prompt assembly before calling a Cortex LLM function.
- B. Incorrect.
Incorrect. This reflects a common misconception: Cortex LLM functions process the prompt you provide, but they do not automatically fix unsafe SQL string construction in your query text. If raw text is concatenated directly into SQL, embedded quotes or control characters can break the statement or alter the intended prompt format. Sanitization and safe construction must happen before the Cortex function is invoked.
- C. Incorrect.
Incorrect. Converting text to uppercase does not solve escaping or safe string construction problems. Quotes, backslashes, line breaks, and other special characters can still interfere with SQL literals or prompt structure regardless of case. This option may seem attractive because it transforms the input, but it does not address the actual safety and templating requirement.
- D. Incorrect.
Incorrect. Temporary tables can be useful in some workflows, but they are not required just to safely inject plain text into a prompt. The statement that helper functions only work with structured JSON inputs is inaccurate. Helper functions are intended to simplify prompt handling and dynamic value insertion, including common text-based use cases.