用于文本到 SQL 的智能体，带有自动错误修正功能

在本教程中，我们将学习如何实现一个利用 SQL 的智能体，使用 transformers.agents。

与标准文本到 SQL pipeline 相比，它有什么优势？

标准的文本到 SQL pipeline 是脆弱的，因为生成的 SQL 查询可能是错误的。更糟糕的是，查询可能是错误的，但并不会引发错误，而是返回一些错误的/无用的输出，且不会发出警报。

👉 相比之下，智能体系统能够批判性地检查输出，并决定是否需要更改查询，从而大大提高其性能。

让我们开始构建这个智能体吧！💪

设置 SQL 表

from sqlalchemy import (
    create_engine,
    MetaData,
    Table,
    Column,
    String,
    Integer,
    Float,
    insert,
    inspect,
    text,
)

engine = create_engine("sqlite:///:memory:")
metadata_obj = MetaData()

# create city SQL table
table_name = "receipts"
receipts = Table(
    table_name,
    metadata_obj,
    Column("receipt_id", Integer, primary_key=True),
    Column("customer_name", String(16), primary_key=True),
    Column("price", Float),
    Column("tip", Float),
)
metadata_obj.create_all(engine)

rows = [
    {"receipt_id": 1, "customer_name": "Alan Payne", "price": 12.06, "tip": 1.20},
    {"receipt_id": 2, "customer_name": "Alex Mason", "price": 23.86, "tip": 0.24},
    {"receipt_id": 3, "customer_name": "Woodrow Wilson", "price": 53.43, "tip": 5.43},
    {"receipt_id": 4, "customer_name": "Margaret James", "price": 21.11, "tip": 1.00},
]
for row in rows:
    stmt = insert(receipts).values(**row)
    with engine.begin() as connection:
        cursor = connection.execute(stmt)

让我们检查系统是否能通过一个基本查询正常工作：

>>> with engine.connect() as con:
...     rows = con.execute(text("""SELECT * from receipts"""))
...     for row in rows:
...         print(row)

(1, 'Alan Payne', 12.06, 1.2)
(2, 'Alex Mason', 23.86, 0.24)
(3, 'Woodrow Wilson', 53.43, 5.43)
(4, 'Margaret James', 21.11, 1.0)

构建我们的智能体

现在，让我们将 SQL 表格使其可由智能体（工具）检索。

智能体的 description 属性将被嵌入到大语言模型（LLM）的提示中，这样它就能了解如何使用这个工具。在这个步骤中，我们需要描述 SQL 表格的结构，以便让智能体能够正确地执行查询并与数据库交互。

>>> inspector = inspect(engine)
>>> columns_info = [(col["name"], col["type"]) for col in inspector.get_columns("receipts")]

>>> table_description = "Columns:\n" + "\n".join([f"  - {name}: {col_type}" for name, col_type in columns_info])
>>> print(table_description)

Columns:
  - receipt_id: INTEGER
  - customer_name: VARCHAR(16)
  - price: FLOAT
  - tip: FLOAT

现在，让我们构建我们的工具。它需要以下内容：（详细信息请参阅文档）

带有 Args: 部分的文档字符串
类型提示

from transformers.agents import tool


@tool
def sql_engine(query: str) -> str:
    """
    Allows you to perform SQL queries on the table. Returns a string representation of the result.
    The table is named 'receipts'. Its description is as follows:
        Columns:
        - receipt_id: INTEGER
        - customer_name: VARCHAR(16)
        - price: FLOAT
        - tip: FLOAT

    Args:
        query: The query to perform. This should be correct SQL.
    """
    output = ""
    with engine.connect() as con:
        rows = con.execute(text(query))
        for row in rows:
            output += "\n" + str(row)
    return output

现在让我们创建一个利用这个工具的智能体。

我们将使用 ReactCodeAgent，它是 transformers.agents 的主要智能体类：一个根据 ReAct 框架编写代码并能迭代先前输出的智能体。

llm_engine 是驱动智能体系统的 LLM。HfEngine 允许你通过 HF 的推理 API 调用 LLM，无论是通过无服务器或专用端点，但你也可以使用任何专有的 API：查看这个指南以了解如何进行适配。

from transformers.agents import ReactCodeAgent, HfApiEngine

agent = ReactCodeAgent(
    tools=[sql_engine],
    llm_engine=HfApiEngine("meta-llama/Meta-Llama-3-8B-Instruct"),
)

agent.run("Can you give me the name of the client who got the most expensive receipt?")

提高难度：表格连接

现在让我们增加一点难度！我们希望智能体能够处理跨多个表的连接查询。

因此，让我们创建一个第二个表，用于记录每个 receipt_id 对应的服务员姓名！

table_name = "waiters"
receipts = Table(
    table_name,
    metadata_obj,
    Column("receipt_id", Integer, primary_key=True),
    Column("waiter_name", String(16), primary_key=True),
)
metadata_obj.create_all(engine)

rows = [
    {"receipt_id": 1, "waiter_name": "Corey Johnson"},
    {"receipt_id": 2, "waiter_name": "Michael Watts"},
    {"receipt_id": 3, "waiter_name": "Michael Watts"},
    {"receipt_id": 4, "waiter_name": "Margaret James"},
]
for row in rows:
    stmt = insert(receipts).values(**row)
    with engine.begin() as connection:
        cursor = connection.execute(stmt)

我们需要更新 SQLExecutorTool，将这个表的描述添加进去，以便让 LLM 能够正确地利用这个表中的信息。

>>> updated_description = """Allows you to perform SQL queries on the table. Beware that this tool's output is a string representation of the execution output.
... It can use the following tables:"""

>>> inspector = inspect(engine)
>>> for table in ["receipts", "waiters"]:
...     columns_info = [(col["name"], col["type"]) for col in inspector.get_columns(table)]

...     table_description = f"Table '{table}':\n"

...     table_description += "Columns:\n" + "\n".join([f"  - {name}: {col_type}" for name, col_type in columns_info])
...     updated_description += "\n\n" + table_description

>>> print(updated_description)

Allows you to perform SQL queries on the table. Beware that this tool's output is a string representation of the execution output.
It can use the following tables:

Table 'receipts':
Columns:
  - receipt_id: INTEGER
  - customer_name: VARCHAR(16)
  - price: FLOAT
  - tip: FLOAT

Table 'waiters':
Columns:
  - receipt_id: INTEGER
  - waiter_name: VARCHAR(16)

由于这个请求比之前的更具挑战性，我们将切换 LLM 引擎，使用更强大的 Qwen/Qwen2.5-72B-Instruct！

sql_engine.description = updated_description

agent = ReactCodeAgent(
    tools=[sql_engine],
    llm_engine=HfApiEngine("Qwen/Qwen2.5-72B-Instruct"),
)

agent.run("Which waiter got more total money from tips?")

它直接就能工作！设置过程出乎意料地简单，不是吗？

✅ 现在你可以去构建你一直梦想的文本到 SQL 系统了！✨

< > Update on GitHub

Open-Source AI Cookbook

用于文本到 SQL 的智能体，带有自动错误修正功能

设置 SQL 表

构建我们的智能体

提高难度：表格连接