yipao.agent.core.Agent¶

class yipao.agent.core.Agent(llm: CustomLLM, database, vectorstore, name_collection='ddl')[source]¶

Agent class for managing interactions between language models and SQL databases. It uses Retrieval-Augmented Generation to construct and execute SQL queries based on natural language input.

llm¶

A language model for generating SQL queries.

Type:: CustomLLM

database¶: The database object to execute SQL queries.

vectorstore¶: The storage for vector embeddings of SQL statements.

retriever¶

A retriever for related data definitions.

Type:: VectorStoreRetriever

promptBuilder¶

A builder for creating SQL prompts.

Type:: PromptBuilder

history¶

A record of executed SQL queries.

Type:: str

__init__(llm: CustomLLM, database, vectorstore, name_collection='ddl')[source]¶

Initializes the Agent with a language model, database, and vector store.

Parameters:

llm (CustomLLM) – The language model used to generate SQL queries.
database – The database interface to execute SQL queries.
vectorstore – The vector store interface for saving SQL statement embeddings.

Methods

`__init__`(llm, database, vectorstore[, ...])	Initializes the Agent with a language model, database, and vector store.
`document_database`(database[, force_update])	Documents the database by extracting table descriptions and saving them into the vector store.
`execute_sql`(sql_query)	Executes an SQL query on the database.
`invoke`(question[, iterations, debug])	Processes a natural language question to generate and execute SQL queries over multiple iterations.