Skip to content

Latest commit

 

History

History
76 lines (64 loc) · 4.14 KB

2_3_1_SQLDB_WRAPPER.md

File metadata and controls

76 lines (64 loc) · 4.14 KB
Raw

SQLDB Wrapper

Based on SQLAlchemy and langchain's SQLDatabase, agentUniverse offers a SQLDBWrapper class allowing users to conveniently use most SQL-based databases like SQLite, MySQL, Oracle, etc. For details, please refer to the SQLAlchemy official website. With SQLDBWrapper, you can easily manage multiple different database connection objects at the same time.

Registration Method

Step One: Configure Scan Path

gentUniverse automatically registers the SQLDBWrapper configuration files by scanning the configured paths at application startup. The paths to scan are configured in the config.toml file:

[CORE_PACKAGE]
default = ['default_scan_path']
sqldb_wrapper = ['sqldb_wrapper_scan_path']

By default, AgentUniverse scans all paths under default or sqldb_wrapper, with paths under sqldb_wrapper having a higher priority over default.

Step Two: Configuration File

name: 'demo_sql'
description: 'demo_sql'
db_uri: "sqlite:///./demo.db"
sql_database_args:
  include_tables: ["users"]
engine_args:
  pool_size: 10
metadata:
  type: 'SQLDB_WRAPPER'
  • name: The name of the SQLDBWrapper, used to identify a unique instance.
  • description: A description of the SQLDBWrapper's functionality.
  • db_uri: A SQLAlchemy-style database URI, used to create a SQLAlchemy engine.
  • sql_database_args: Optional parameters for the SQLDatabase class in LangChain, for specific details refer to LangChain's official website.
  • engine_args: Optional parameters, used to configure SQLAlchemy engine parameters. For specific configurable contents, please refer to SQLAlchemy's webpage.
  • metadata: Indicates that this configuration is for a SQLDB_WRAPPER, no need to change.

Usage

The SQLDBWrapper class offers the following functions for users:

run

run accepts a str type parameter command, representing an SQL statement, and executes this SQL via SQLAlchemy, returning the results. The result is a list, if command is a query statement then the list will contain the query results, each element is a dictionary representing a row in the database, with the key as the column name and the value as the corresponding value.

from agentuniverse.database.sqldb_wrapper import SQLDBWrapper
from agentuniverse.database.sqldb_wrapper_manager import SQLDBWrapperManager

demo_sqldb_wrapper:SQLDBWrapper = SQLDBWrapperManager().get_instance_obj("demo_sqldb_wrapper")
demo_sqldb_wrapper.run("select * from USERS")

Example of returned result:

[{'age': 30, 'id': 1, 'name': 'Alice'}, {'age': 25, 'id': 2, 'name': 'Bob'}]

run_with_str_return

run_with_str_return accepts a str type parameter command, representing an SQL statement, and executes this SQL via SQLAlchemy, returning the results. The result is a str, if command is a query statement then the str consists of all the query results concatenated together, and if the max_string_length parameter was configured in sql_database_args, the result will be truncated. This function is more convenient for using the results as input for llm.

from agentuniverse.database.sqldb_wrapper import SQLDBWrapper
from agentuniverse.database.sqldb_wrapper_manager import SQLDBWrapperManager

demo_sqldb_wrapper:SQLDBWrapper = SQLDBWrapperManager().get_instance_obj("demo_sqldb_wrapper")
demo_sqldb_wrapper.run_with_str_return("select * from USERS")

Example of returned result:

"[(1, 'Alice', 30), (2, 'Bob', 25)]"

sql_database

If you wish to use some of the native features of SQLAlchemy or SQLDatabase, you can access the corresponding object through the sql_database property of SQLDBWrapper:

from agentuniverse.database.sqldb_wrapper import SQLDBWrapper
from agentuniverse.database.sqldb_wrapper_manager import SQLDBWrapperManager

demo_sqldb_wrapper:SQLDBWrapper = SQLDBWrapperManager().get_instance_obj("demo_sqldb_wrapper")
sql_database_ins = demo_sqldb_wrapper.sql_database
sql_alchemy_engine = sql_database_ins._engine