如何在 SQL 问答过程中进行查询验证

也许任何SQL链或代理中最容易出错的部分是编写有效且安全的SQL查询。在本指南中，我们将介绍一些验证查询和处理无效查询的策略。

我们将涵盖：

1. 在查询生成步骤中添加一个“查询验证器”步骤；
2. 提示工程以降低错误发生率。

设置

首先，获取所需的包并设置环境变量：

%pip install --upgrade --quiet  langchain langchain-community langchain-openai

# Uncomment the below to use LangSmith. Not required.
# import os
# os.environ["LANGSMITH_API_KEY"] = getpass.getpass()
# os.environ["LANGSMITH_TRACING"] = "true"

以下示例将使用与 Chinook 数据库的 SQLite 连接。按照 这些安装步骤 在此笔记本文件所在目录中创建 Chinook.db：

• 保存 此文件 为 Chinook_Sqlite.sql
• 运行 sqlite3 Chinook.db
• 运行 .read Chinook_Sqlite.sql
• 测试 SELECT * FROM Artist LIMIT 10;

现在，Chinook.db 已在我们的目录中，我们可以使用基于 SQLAlchemy 的 SQLDatabase 类与它进行交互：

from langchain_community.utilities import SQLDatabase

db = SQLDatabase.from_uri("sqlite:///Chinook.db")
print(db.dialect)
print(db.get_usable_table_names())
print(db.run("SELECT * FROM Artist LIMIT 10;"))

API 参考：SQLDatabase

sqlite
['Album', 'Artist', 'Customer', 'Employee', 'Genre', 'Invoice', 'InvoiceLine', 'MediaType', 'Playlist', 'PlaylistTrack', 'Track']
[(1, 'AC/DC'), (2, 'Accept'), (3, 'Aerosmith'), (4, 'Alanis Morissette'), (5, 'Alice In Chains'), (6, 'Antônio Carlos Jobim'), (7, 'Apocalyptica'), (8, 'Audioslave'), (9, 'BackBeat'), (10, 'Billy Cobham')]

查询检查器

也许最简单的策略是让模型本身检查原始查询中常见的错误。假设我们有以下SQL查询链：

选择 聊天模型:

OpenAI

安特罗比克

Azure

谷歌 Gemini

谷歌Vertex

AWS

Groq

Cohere

NVIDIA

Fireworks AI

Mistral AI

Together AI

IBM watsonx

Databricks

xAI

困惑度

pip install -qU "langchain[openai]"

import getpass
import os

if not os.environ.get("OPENAI_API_KEY"):
  os.environ["OPENAI_API_KEY"] = getpass.getpass("Enter API key for OpenAI: ")

from langchain.chat_models import init_chat_model

llm = init_chat_model("gpt-4o-mini", model_provider="openai")

from langchain.chains import create_sql_query_chain

chain = create_sql_query_chain(llm, db)

API 参考：create_sql_query_chain

我们希望验证其输出结果。可以通过扩展链路，添加第二个提示和模型调用来进行验证：

from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate

system = """Double check the user's {dialect} query for common mistakes, including:
- Using NOT IN with NULL values
- Using UNION when UNION ALL should have been used
- Using BETWEEN for exclusive ranges
- Data type mismatch in predicates
- Properly quoting identifiers
- Using the correct number of arguments for functions
- Casting to the correct data type
- Using the proper columns for joins

If there are any of the above mistakes, rewrite the query.
If there are no mistakes, just reproduce the original query with no further commentary.

Output the final SQL query only."""
prompt = ChatPromptTemplate.from_messages(
    [("system", system), ("human", "{query}")]
).partial(dialect=db.dialect)
validation_chain = prompt | llm | StrOutputParser()

full_chain = {"query": chain} | validation_chain

API 参考：StrOutputParser | ChatPromptTemplate

query = full_chain.invoke(
    {
        "question": "What's the average Invoice from an American customer whose Fax is missing since 2003 but before 2010"
    }
)
print(query)

SELECT AVG(i.Total) AS AverageInvoice
FROM Invoice i
JOIN Customer c ON i.CustomerId = c.CustomerId
WHERE c.Country = 'USA'
AND c.Fax IS NULL
AND i.InvoiceDate >= '2003-01-01' 
AND i.InvoiceDate < '2010-01-01'

注意我们如何在 Langsmith trace 中看到链的两个步骤。

db.run(query)

'[(6.632999999999998,)]'

这种方法的明显缺点是我们需要进行两次模型调用，而不是一次，才能生成查询。为了克服这一点，我们可以尝试在一次模型调用中完成查询生成和查询检查：

system = """You are a {dialect} expert. Given an input question, create a syntactically correct {dialect} query to run.
Unless the user specifies in the question a specific number of examples to obtain, query for at most {top_k} results using the LIMIT clause as per {dialect}. You can order the results to return the most informative data in the database.
Never query for all columns from a table. You must query only the columns that are needed to answer the question. Wrap each column name in double quotes (") to denote them as delimited identifiers.
Pay attention to use only the column names you can see in the tables below. Be careful to not query for columns that do not exist. Also, pay attention to which column is in which table.
Pay attention to use date('now') function to get the current date, if the question involves "today".

Only use the following tables:
{table_info}

Write an initial draft of the query. Then double check the {dialect} query for common mistakes, including:
- Using NOT IN with NULL values
- Using UNION when UNION ALL should have been used
- Using BETWEEN for exclusive ranges
- Data type mismatch in predicates
- Properly quoting identifiers
- Using the correct number of arguments for functions
- Casting to the correct data type
- Using the proper columns for joins

Use format:

First draft: <<FIRST_DRAFT_QUERY>>
Final answer: <<FINAL_ANSWER_QUERY>>
"""
prompt = ChatPromptTemplate.from_messages(
    [("system", system), ("human", "{input}")]
).partial(dialect=db.dialect)


def parse_final_answer(output: str) -> str:
    return output.split("Final answer: ")[1]


chain = create_sql_query_chain(llm, db, prompt=prompt) | parse_final_answer
prompt.pretty_print()

================================ System Message ================================

You are a {dialect} expert. Given an input question, create a syntactically correct {dialect} query to run.
Unless the user specifies in the question a specific number of examples to obtain, query for at most {top_k} results using the LIMIT clause as per {dialect}. You can order the results to return the most informative data in the database.
Never query for all columns from a table. You must query only the columns that are needed to answer the question. Wrap each column name in double quotes (") to denote them as delimited identifiers.
Pay attention to use only the column names you can see in the tables below. Be careful to not query for columns that do not exist. Also, pay attention to which column is in which table.
Pay attention to use date('now') function to get the current date, if the question involves "today".

Only use the following tables:
{table_info}

Write an initial draft of the query. Then double check the {dialect} query for common mistakes, including:
- Using NOT IN with NULL values
- Using UNION when UNION ALL should have been used
- Using BETWEEN for exclusive ranges
- Data type mismatch in predicates
- Properly quoting identifiers
- Using the correct number of arguments for functions
- Casting to the correct data type
- Using the proper columns for joins

Use format:

First draft: <<FIRST_DRAFT_QUERY>>
Final answer: <<FINAL_ANSWER_QUERY>>


================================ Human Message =================================

{input}

query = chain.invoke(
    {
        "question": "What's the average Invoice from an American customer whose Fax is missing since 2003 but before 2010"
    }
)
print(query)

SELECT AVG(i."Total") AS "AverageInvoice"
FROM "Invoice" i
JOIN "Customer" c ON i."CustomerId" = c."CustomerId"
WHERE c."Country" = 'USA'
AND c."Fax" IS NULL
AND i."InvoiceDate" BETWEEN '2003-01-01' AND '2010-01-01';

db.run(query)

'[(6.632999999999998,)]'

Human-in-the-loop

在某些情况下，我们的数据敏感度很高，因此我们永远不希望在没有人类批准的情况下执行SQL查询。前往 工具使用：人工介入 页面，了解如何向任何工具、链或代理添加人工介入。

错误处理

在某个时刻，模型可能会犯错并生成无效的SQL查询。或者我们的数据库会出现问题。或者模型API会宕机。我们希望为链和代理添加一些错误处理行为，以便在这些情况下能够优雅地失败，甚至自动恢复。要了解有关工具使用的错误处理信息，请前往 工具使用：错误处理 页面。