LangChain 54 深入理解LangChain 表达式语言十七 Chains Route动态路由 LangChain Expression Language (LCEL)

最新推荐文章于 2024-10-18 08:08:06 发布

AI架构师易筋

最新推荐文章于 2024-10-18 08:08:06 发布

阅读量792

点赞数 18

分类专栏： LLM-Large Language Models 文章标签： langchain chatgpt 人工智能 LLM prompt 语言模型

本文链接：https://blog.csdn.net/zgpeace/article/details/135332644

版权

LLM-Large Language Models 专栏收录该内容

185 篇文章 124 订阅

订阅专栏

LangChain系列文章

在这里插入图片描述

1. Dynamically route根据输入动态路由逻辑

本笔记本涵盖了如何在LangChain表达语言中进行路由。

路由允许您创建非确定性链，其中上一步的输出定义了下一步。路由有助于在与LLMs的交互中提供结构和一致性。

有两种方法可以执行路由：

使用RunnableBranch。
编写自定义工厂函数，该函数获取上一步的输入并返回一个可运行的函数。重要的是，这应该返回一个可运行的函数，而不是实际执行。

我们将使用两个步骤的序列来说明这两种方法，其中第一步将输入问题分类为关于LangChain、ChatGPT或其他，然后路由到相应的提示链。

2. 创建三个子链：

from langchain_core.runnables import RunnablePassthrough
from langchain.prompts import ChatPromptTemplate
from langchain.chat_models import ChatOpenAI
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnableParallel
from langchain.prompts import PromptTemplate

from dotenv import load_dotenv
load_dotenv()
# from langchain.globals import set_debug
# set_debug(True)

langchain_chain = (
    PromptTemplate.from_template(
        """You are an expert in langchain. \
Always answer questions starting with "As Harrison Chase told me". \
Respond to the following question:

Question: {question}
Answer:"""
    )
    | ChatOpenAI()
)
chatgpt_chain = (
    PromptTemplate.from_template(
        """You are an expert in ChatGPT. \
Always answer questions starting with "As Sam Altman told me". \
Respond to the following question:

Question: {question}
Answer:"""
    )
    | ChatOpenAI()
)
general_chain = (
    PromptTemplate.from_template(
        """Respond to the following question:

Question: {question}
Answer:"""
    )
    | ChatOpenAI()
)

from langchain_core.runnables import RunnableBranch

branch = RunnableBranch(
    (lambda x: "chatgpt" in x["topic"].lower(), chatgpt_chain),
    (lambda x: "langchain" in x["topic"].lower(), langchain_chain),
    general_chain,
)

full_chain = {"topic": chain, "question": lambda x: x["question"]} | branch

response = full_chain.invoke({"question": "how do I use ChatGPT ?"})
print('response >> ', response)

response = full_chain.invoke({"question": "how do I use LangChain ?"})
print('response >> ', response)

response = full_chain.invoke({"question": "whats 2 + 2"})
print('response >> ', response)

2.1 ChatGPT 问题 response

zgpeace@zgpeaces-MacBook-Pro ~/W/L/langchain-llm-app> python LCEL/router.py                                              [8:50:46] develop?
response >>  content='As Sam Altman told me, to use ChatGPT, you can access it through OpenAI\'s API. You need to make API calls to send a series of messages as input to the model. Start with a system message to set the behavior of the assistant, followed by user messages to have a conversation. Each message should have a \'role\' (either "system", "user", or "assistant") and \'content\' (the text of the message). You can continue the conversation by extending the list of messages and making additional API calls as needed.'

2.2 Langchain 问题 response

zgpeace@zgpeaces-MacBook-Pro ~/W/L/langchain-llm-app> python LCEL/router.py                                              [8:49:42] develop?
response >>  content='As Harrison Chase told me, to use LangChain, you can follow these steps:\n\n1. Visit the LangChain website and create an account.\n2. Once you have created an account, log in to your LangChain account.\n3. On the dashboard, you will find various options and tools to use the LangChain platform.\n4. To start using LangChain, you can either request a translation or offer your translation services.\n5. If you want to request a translation, provide the necessary details such as the source and target languages, the document or text you need translating, and any specific instructions.\n6. LangChain will then connect you with a suitable translator who will work on your request.\n7. If you want to offer your translation services, you can create a profile showcasing your language skills and expertise.\n8. Clients looking for translators will be able to find and hire you based on your profile and qualifications.\n9. Once you have completed a translation or received a translated document, you can communicate with the client through the LangChain platform for any further clarifications or revisions.\n10. LangChain also provides a secure payment system for both clients and translators, ensuring a smooth and reliable transaction process.\n\nRemember to always refer to the LangChain platform for any specific instructions or guidance as they may have updated features or processes.'

2.3 通用问题response

zgpeace@zgpeaces-MacBook-Pro ~/W/L/langchain-llm-app> python LCEL/router.py                                              [8:48:19] develop?
response >>  content='The sum of 2 and 2 is 4.'