Overview

The Text2SQLQuestionGenerator operator is designed to generate natural language questions corresponding to given SQL queries. If an entry lacks a natural language question, this operator generates multiple candidate questions, evaluates them, and selects the best one to complete the data entry. It leverages language models for generation and embedding models for selection, interacting with a database manager to retrieve schema information.

__init__

def __init__(self, 
            llm_serving: LLMServingABC, 
            embedding_serving: LLMServingABC, 
            database_manager: DatabaseManager, 
            question_candidates_num: int = 5,
            prompt_template = None
            )

Parameter	Type	Default	Description
llm_serving	LLMServingABC	Required	An instance of a large language model serving class, used for generating question candidates.
embedding_serving	LLMServingABC	Required	An instance of an embedding model serving class, used to create embeddings for selecting the best question.
database_manager	DatabaseManager	Required	A manager to interact with the database and retrieve schema information.
question_candidates_num	int	5	The number of candidate questions to generate for each SQL query.
prompt_template	PromptABC	None	The prompt template object used to construct the input for the LLM. Defaults to Text2SQLQuestionGeneratorPrompt if not provided.

Prompt Template Descriptions

Prompt Template Name	Primary Purpose	Applicable Scenarios	Feature Description

run

def run(self, 
        storage: DataFlowStorage,
        input_sql_key: str = "sql",
        input_db_id_key: str = "db_id",
        output_question_key: str = "question",
        output_evidence_key: str = "evidence"
        )

Parameter	Type	Default	Description
storage	DataFlowStorage	Required	The data flow storage instance for reading and writing data.
input_sql_key	str	"sql"	The column name in the input data that contains the SQL query.
input_db_id_key	str	"db_id"	The column name in the input data that contains the database ID.
output_question_key	str	"question"	The column name for storing the generated natural language question in the output data.
output_evidence_key	str	"evidence"	The column name for storing the generated evidence or external knowledge in the output data.

🧠 Example Usage