使用CogniSwitch Toolkit与Langchain创建知识检索应用-优快云博客

本文链接：https://blog.youkuaiyun.com/dagGAIYD/article/details/145331338

在AI驱动的开发环境中，CogniSwitch提供了一个强大的工具集，使得构建生产级应用变得更加高效和可靠。结合Langchain框架，CogniSwitch可以帮助开发者在选择存储和检索格式时提供便利，消除响应生成中的可靠性问题和虚假信息。本文将详细介绍如何使用CogniSwitch Toolkit与Langchain创建知识检索应用。

1. 技术背景介绍

CogniSwitch旨在通过先进的AI技术帮助开发者高效地管理和利用信息。它提供了一系列功能，包括知识提取、自然语言处理和情感分析，使得构建复杂的生成式AI应用变得更加简单和可靠。

2. 核心原理解析

CogniSwitch通过其工具包提供了多种功能，允许开发者轻松上传文件或URL，并通过高级的自然语言处理能力进行信息提取和分析。结合Langchain框架，可以创建一个强大的对话式检索代理，基于提供的知识库进行准确的回答和交互。

3. 代码实现演示

首先，需要注册一个CogniSwitch账号并安装必要的库：

# 安装Langchain社区版
%pip install -qU langchain-community

接下来导入必要的库：

import warnings
warnings.filterwarnings("ignore")
import os
from langchain.agents.agent_toolkits import create_conversational_retrieval_agent
from langchain_community.agent_toolkits import CogniswitchToolkit
from langchain_openai import ChatOpenAI

# 设置API密钥环境变量
cs_token = "Your CogniSwitch token"
OAI_token = "Your OpenAI API token"
oauth_token = "Your CogniSwitch authentication token"
os.environ["OPENAI_API_KEY"] = OAI_token

# 实例化CogniSwitch工具包
cogniswitch_toolkit = CogniswitchToolkit(
    cs_token=cs_token, OAI_token=OAI_token, apiKey=oauth_token
)

# 获取CogniSwitch工具列表
tool_lst = cogniswitch_toolkit.get_tools()

# 实例化语言模型
llm = ChatOpenAI(
    temperature=0,
    openai_api_key=OAI_token,
    max_tokens=1500,
    model_name="gpt-3.5-turbo-0613",
)

# 使用语言模型和工具包创建代理
agent_executor = create_conversational_retrieval_agent(llm, tool_lst, verbose=False)

上传URL

response = agent_executor.invoke("upload this url https://cogniswitch.ai/developer")
print(response["output"])
# 输出示例: The URL https://cogniswitch.ai/developer has been uploaded successfully. The status of the document is currently being processed.

上传文件

response = agent_executor.invoke("upload this file example_file.txt")
print(response["output"])
# 输出示例: The file example_file.txt has been uploaded successfully. The status of the document is currently being processed.

查询文档状态

response = agent_executor.invoke("Tell me the status of this document example_file.txt")
print(response["output"])
# 输出示例: The status of the document example_file.txt is as follows...

查询知识

response = agent_executor.invoke("How can cogniswitch help develop GenAI applications?")
print(response["output"])
# 输出示例: CogniSwitch can help develop GenAI applications in several ways...