Cassandra数据库工具包

Apache Cassandra® is a widely used database for storing transactional application data. The introduction of functions and >tooling in Large Language Models has opened up some exciting use cases for existing data in Generative AI applications.

The Cassandra Database toolkit enables AI engineers to integrate agents with Cassandra data efficiently, offering the following features:

• Fast data access through optimized queries. Most queries should run in single-digit ms or less.
• Schema introspection to enhance LLM reasoning capabilities
• Compatibility with various Cassandra deployments, including Apache Cassandra®, DataStax Enterprise™, and DataStax Astra™
• Currently, the toolkit is limited to SELECT queries and schema introspection operations. (Safety first)

有关创建Cassandra数据库代理的更多信息，请参阅CQL代理示例

快速入门

• 安装 cassio 库
• 为要连接的Cassandra数据库设置环境变量
• 初始化 CassandraDatabase
• 通过 toolkit.get_tools() 将工具传递给您的代理
• 坐下来，看着它为你完成所有工作。

操作原理

Cassandra Query Language (CQL) 是与 Cassandra 数据库交互的主要 以人为中心 的方式。虽然在生成查询时提供了一些灵活性，但它需要了解 Cassandra 数据建模的最佳实践。LLM 函数调用使代理能够推理并选择工具来满足请求。使用 LLM 的代理在选择适当的工具包或工具包链时，应使用特定于 Cassandra 的逻辑进行推理。这减少了当 LLM 被迫提供自上而下的解决方案时引入的随机性。你希望 LLM 完全不受限制地访问你的数据库吗？是的。可能不是。为了实现这一点，我们在为代理构建问题时提供了一个提示：

你是一个 Apache Cassandra 专家查询分析机器人，具有以下特点和规则：

• 你将从最终用户那里获取一个关于在数据库中查找特定数据的问题。
• 你将检查数据库的模式并创建查询路径。
• 你将为用户提供正确的查询，以找到他们所需的数据，并展示查询路径提供的步骤。
• 你将使用最佳实践，通过分区键和聚簇列查询 Apache Cassandra。
• 避免在查询中使用 ALLOW FILTERING。
• 目标是找到一条查询路径，因此可能需要查询其他表才能得到最终答案。

以下是一个 JSON 格式的查询路径示例：

 {
  "query_paths": [
    {
      "description": "Direct query to users table using email",
      "steps": [
        {
          "table": "user_credentials",
          "query": 
             "SELECT userid FROM user_credentials WHERE email = 'example@example.com';"
        },
        {
          "table": "users",
          "query": "SELECT * FROM users WHERE userid = ?;"
        }
      ]
    }
  ]
}

提供的工具

cassandra_db_schema

收集连接数据库或特定模式的所有模式信息。在确定操作时对代理至关重要。

cassandra_db_select_table_data

从特定的键空间和表中选择数据。代理可以传递谓词参数以及返回记录数的限制。

cassandra_db_query

这是 cassandra_db_select_table_data 的实验性替代方案，它采用由代理完全形成的查询字符串，而不是参数。警告：这可能会导致性能不佳（甚至无法运行）的异常查询。在未来的版本中可能会移除此功能。如果它实现了某些很酷的功能，我们也希望能了解。你永远不知道！

环境设置

安装以下 Python 模块：

pip install ipykernel python-dotenv cassio langchain_openai langchain langchain-community langchainhub

.env 文件

连接是通过 cassio 使用 auto=True 参数进行的，且笔记本使用了 OpenAI。你应该相应地创建一个 .env 文件。

对于Cassandra，设置：

CASSANDRA_CONTACT_POINTS
CASSANDRA_USERNAME
CASSANDRA_PASSWORD
CASSANDRA_KEYSPACE

对于Astra，设置：

ASTRA_DB_APPLICATION_TOKEN
ASTRA_DB_DATABASE_ID
ASTRA_DB_KEYSPACE

例如：

# Connection to Astra:
ASTRA_DB_DATABASE_ID=a1b2c3d4-...
ASTRA_DB_APPLICATION_TOKEN=AstraCS:...
ASTRA_DB_KEYSPACE=notebooks

# Also set 
OPENAI_API_KEY=sk-....

(您还可以修改以下代码，直接与 cassio 连接。)

from dotenv import load_dotenv

load_dotenv(override=True)

# Import necessary libraries
import os

import cassio
from langchain import hub
from langchain.agents import AgentExecutor, create_openai_tools_agent
from langchain_community.agent_toolkits.cassandra_database.toolkit import (
    CassandraDatabaseToolkit,
)
from langchain_community.tools.cassandra_database.prompt import QUERY_PATH_PROMPT
from langchain_community.utilities.cassandra_database import CassandraDatabase
from langchain_openai import ChatOpenAI

API 参考：hub |AgentExecutor | create_openai_tools_agent | CassandraDatabaseToolkit | CassandraDatabase | ChatOpenAI

连接到Cassandra数据库

cassio.init(auto=True)
session = cassio.config.resolve_session()
if not session:
    raise Exception(
        "Check environment configuration or manually configure cassio connection parameters"
    )

# Test data pep

session = cassio.config.resolve_session()

session.execute("""DROP KEYSPACE IF EXISTS langchain_agent_test; """)

session.execute(
    """
CREATE KEYSPACE if not exists langchain_agent_test 
WITH replication = {'class': 'SimpleStrategy', 'replication_factor': 1};
"""
)

session.execute(
    """
    CREATE TABLE IF NOT EXISTS langchain_agent_test.user_credentials (
    user_email text PRIMARY KEY,
    user_id UUID,
    password TEXT
);
"""
)

session.execute(
    """
    CREATE TABLE IF NOT EXISTS langchain_agent_test.users (
    id UUID PRIMARY KEY,
    name TEXT,
    email TEXT
);"""
)

session.execute(
    """
    CREATE TABLE IF NOT EXISTS langchain_agent_test.user_videos ( 
    user_id UUID,
    video_id UUID,
    title TEXT,
    description TEXT,
    PRIMARY KEY (user_id, video_id)
);
"""
)

user_id = "522b1fe2-2e36-4cef-a667-cd4237d08b89"
video_id = "27066014-bad7-9f58-5a30-f63fe03718f6"

session.execute(
    f"""
    INSERT INTO langchain_agent_test.user_credentials (user_id, user_email) 
    VALUES ({user_id}, 'patrick@datastax.com');
"""
)

session.execute(
    f"""
    INSERT INTO langchain_agent_test.users (id, name, email) 
    VALUES ({user_id}, 'Patrick McFadin', 'patrick@datastax.com');
"""
)

session.execute(
    f"""
    INSERT INTO langchain_agent_test.user_videos (user_id, video_id, title)
    VALUES ({user_id}, {video_id}, 'Use Langflow to Build a LangChain LLM Application in 5 Minutes');
"""
)

session.set_keyspace("langchain_agent_test")

# Create a CassandraDatabase instance
# Uses the cassio session to connect to the database
db = CassandraDatabase()

# Choose the LLM that will drive the agent
# Only certain models support this
llm = ChatOpenAI(temperature=0, model="gpt-4-1106-preview")
toolkit = CassandraDatabaseToolkit(db=db)

tools = toolkit.get_tools()

print("Available tools:")
for tool in tools:
    print(tool.name + "\t- " + tool.description)

Available tools:
cassandra_db_schema	- 
    Input to this tool is a keyspace name, output is a table description 
    of Apache Cassandra tables.
    If the query is not correct, an error message will be returned.
    If an error is returned, report back to the user that the keyspace 
    doesn't exist and stop.
    
cassandra_db_query	- 
    Execute a CQL query against the database and get back the result.
    If the query is not correct, an error message will be returned.
    If an error is returned, rewrite the query, check the query, and try again.
    
cassandra_db_select_table_data	- 
    Tool for getting data from a table in an Apache Cassandra database. 
    Use the WHERE clause to specify the predicate for the query that uses the 
    primary key. A blank predicate will return all rows. Avoid this if possible. 
    Use the limit to specify the number of rows to return. A blank limit will 
    return all rows.

prompt = hub.pull("hwchase17/openai-tools-agent")

# Construct the OpenAI Tools agent
agent = create_openai_tools_agent(llm, tools, prompt)

input = (
    QUERY_PATH_PROMPT
    + "\n\nHere is your task: Find all the videos that the user with the email address 'patrick@datastax.com' has uploaded to the langchain_agent_test keyspace."
)

agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

response = agent_executor.invoke({"input": input})

print(response["output"])

> Entering new AgentExecutor chain...

Invoking: `cassandra_db_schema` with `{'keyspace': 'langchain_agent_test'}`


Table Name: user_credentials
- Keyspace: langchain_agent_test
- Columns
  - password (text)
  - user_email (text)
  - user_id (uuid)
- Partition Keys: (user_email)
- Clustering Keys: 

Table Name: user_videos
- Keyspace: langchain_agent_test
- Columns
  - description (text)
  - title (text)
  - user_id (uuid)
  - video_id (uuid)
- Partition Keys: (user_id)
- Clustering Keys: (video_id asc)


Table Name: users
- Keyspace: langchain_agent_test
- Columns
  - email (text)
  - id (uuid)
  - name (text)
- Partition Keys: (id)
- Clustering Keys: 


Invoking: `cassandra_db_select_table_data` with `{'keyspace': 'langchain_agent_test', 'table': 'user_credentials', 'predicate': "user_email = 'patrick@datastax.com'", 'limit': 1}`


Row(user_email='patrick@datastax.com', password=None, user_id=UUID('522b1fe2-2e36-4cef-a667-cd4237d08b89'))
Invoking: `cassandra_db_select_table_data` with `{'keyspace': 'langchain_agent_test', 'table': 'user_videos', 'predicate': 'user_id = 522b1fe2-2e36-4cef-a667-cd4237d08b89', 'limit': 10}`


Row(user_id=UUID('522b1fe2-2e36-4cef-a667-cd4237d08b89'), video_id=UUID('27066014-bad7-9f58-5a30-f63fe03718f6'), description='DataStax Academy is a free resource for learning Apache Cassandra.', title='DataStax Academy')To find all the videos that the user with the email address 'patrick@datastax.com' has uploaded to the `langchain_agent_test` keyspace, we can follow these steps:

1. Query the `user_credentials` table to find the `user_id` associated with the email 'patrick@datastax.com'.
2. Use the `user_id` obtained from the first step to query the `user_videos` table to retrieve all the videos uploaded by the user.

Here is the query path in JSON format:

\`\`\`json
{
  "query_paths": [
    {
      "description": "Find user_id from user_credentials and then query user_videos for all videos uploaded by the user",
      "steps": [
        {
          "table": "user_credentials",
          "query": "SELECT user_id FROM user_credentials WHERE user_email = 'patrick@datastax.com';"
        },
        {
          "table": "user_videos",
          "query": "SELECT * FROM user_videos WHERE user_id = 522b1fe2-2e36-4cef-a667-cd4237d08b89;"
        }
      ]
    }
  ]
}
\`\`\`

Following this query path, we found that the user with the user_id `522b1fe2-2e36-4cef-a667-cd4237d08b89` has uploaded at least one video with the title 'DataStax Academy' and the description 'DataStax Academy is a free resource for learning Apache Cassandra.' The video_id for this video is `27066014-bad7-9f58-5a30-f63fe03718f6`. If there are more videos, the same query can be used to retrieve them, possibly with an increased limit if necessary.

> Finished chain.
To find all the videos that the user with the email address 'patrick@datastax.com' has uploaded to the `langchain_agent_test` keyspace, we can follow these steps:

1. Query the `user_credentials` table to find the `user_id` associated with the email 'patrick@datastax.com'.
2. Use the `user_id` obtained from the first step to query the `user_videos` table to retrieve all the videos uploaded by the user.

Here is the query path in JSON format:

\`\`\`json
{
  "query_paths": [
    {
      "description": "Find user_id from user_credentials and then query user_videos for all videos uploaded by the user",
      "steps": [
        {
          "table": "user_credentials",
          "query": "SELECT user_id FROM user_credentials WHERE user_email = 'patrick@datastax.com';"
        },
        {
          "table": "user_videos",
          "query": "SELECT * FROM user_videos WHERE user_id = 522b1fe2-2e36-4cef-a667-cd4237d08b89;"
        }
      ]
    }
  ]
}
\`\`\`

Following this query path, we found that the user with the user_id `522b1fe2-2e36-4cef-a667-cd4237d08b89` has uploaded at least one video with the title 'DataStax Academy' and the description 'DataStax Academy is a free resource for learning Apache Cassandra.' The video_id for this video is `27066014-bad7-9f58-5a30-f63fe03718f6`. If there are more videos, the same query can be used to retrieve them, possibly with an increased limit if necessary.

相关

• 工具 概念指南
• 工具 如何使用指南