Snowflake Cortex's COMPLETE function is a general-purpose large language model function that accepts a custom prompt and generates free-form text output, making it suitable for interpreting restriction descriptions (such as 'allergic to nuts' or 'no gluten') and producing a derived list of foods to avoid. In contrast, SENTIMENT only returns a numeric sentiment score, CLASSIFY_TEXT assigns text to a predefined fixed set of category labels rather than generating new content, and SUMMARIZE condenses existing text rather than inferring and listing new information not explicitly present verbatim. Only the COMPLETE-based query correctly asks the model to synthesize a list of foods that must be avoided based on the customer-provided restrictions.
Community Discussion