LLM Based SQL Template Matcher

[anisbhsl]

Description

Implement a LLMBasedSQLTemplateMatcher class that takes in a user query text and provides top_k matching templates by substituting the entities.

class LLMBasedSQLTemplateMatcher(SQLTemplateMatcher):
    """
    LLMBasedSQLTemplateMatcher uses LLM to find the best matching template.
    Given a query, it will extract the key entities and use LLM to find best SQL template
    and generates a subsituted SQL query.

    Args:
        templates: A list of SQL templates.
        ddl_schema: The DDL schema for available tables.
        similarity_threshold: The similarity threshold to be used for fuzzy matching.
    """
    def __init__(self, templates: List[SQLTemplate],
                 ddl_schema: Optional[str] = None,
                 similarity_threshold: float = 0.4,
                 debug: bool = False) -> None:
        super().__init__(debug = debug)

        self.templates = templates
        self.ddl_schema = ddl_schema
        self.similarity_threshold = similarity_threshold

    def match(self, query: str, top_k = 1, **kwargs) -> List[str]:
        pass
        

[NISH1001]

More discussion at:
#5