五分钟了解 LangChain 的路由链

愤怒的蜗牛

LangChain里的另外1个重要的链:路由链。

五分钟了解 LangChain 的路由链

1. 路由链概念

路由链(RouterChain)是由LLM根据输入的Prompt去选择具体的某个链。路由链中一般会存在多个Prompt,Prompt结合LLM决定下一步选择哪个链。

五分钟了解 LangChain 的路由链

2. 路由链的使用场景

路由链一般涉及到2个核心类,LLMRouterChain和MultiPromptChain,一起看看官网介绍:

五分钟了解 LangChain 的路由链

  • LLMRouterChain:使用LLM路由到可能的选项中。

  • MultiPromptChain:该链可用于在多个提示词之间路由输入,当你有多个提示词并且只想路由到其中一个时,可以用这个链。

一般使用路由链时,有固定的几个步骤:

  • 准备多个链的Prompt提示词,然后各自封装成链。

  • 将可能路由到的链封装到destination_chains里。

  • 构建多提示词和RouterChain ,负责选择下一个要调用的链。

  • 构建默认链。

  • 使用MultiPromptChain选择某个链,然后再去执行此链。

3. 使用路由链的案例

假设我们有一个常见的场景,根据用户的输入内容选择不同的处理路径,如果没有选到合适的链,则使用默认链。比如:根据用户的输入问题,选择不同的链去处理,如果没选到合适的,则走默认链。

具体代码如下:

from langchain_openai import ChatOpenAI

model = ChatOpenAI(
    model_name="gpt-3.5-turbo",
    openai_api_key="sk-xxxx",
    openai_api_base="https://api.302.ai/v1",)from langchain.chains.router import LLMRouterChain, MultiPromptChain
from langchain.chains.router.llm_router import RouterOutputParser
from langchain.chains.router.multi_prompt_prompt import MULTI_PROMPT_ROUTER_TEMPLATEfrom langchain.chains import LLMChain, ConversationChain
from langchain.prompts import PromptTemplate

# 准备2条目的链:一条物理链,一条数学链
# 1. 物理链
physics_template = """
你是一位物理学家,擅长回答物理相关的问题,当你不知道问题的答案时,你就回答不知道。
具体问题如下:{input}"""
physics_prompt = PromptTemplate.from_template(physics_template)physics_chain = LLMChain(llm=model, prompt=physics_prompt)# 2. 数学链
math_template = """
你是一个数学家,擅长回答数学相关的问题,当你不知道问题的答案时,你就回答不知道。
具体问题如下:{input}"""
math_prompt = PromptTemplate.from_template(math_template)math_chain = LLMChain(llm=model, prompt=math_prompt)# 3. 英语链
english_template = """
你是一个非常厉害的英语老师,擅长回答英语相关的问题,当你不知道问题的答案时,你就回答不知道。
具体问题如下:{input}"""
english_prompt = PromptTemplate.from_template(english_template)english_chain = LLMChain(llm=model, prompt=english_prompt)######### 所有可能的目的链
destination_chains = {}destination_chains["physics"] = physics_chain
destination_chains["math"] = math_chain
destination_chains["english"] = english_chain


######### 默认链
default_chain = ConversationChain(llm=model, output_key="text")# 让多路由模板 能找到合适的 提示词模板
destinations_template_str = """physics:擅长回答物理问题math:擅长回答数学问题english:擅长回答英语问题"""
router_template = MULTI_PROMPT_ROUTER_TEMPLATE.format(
    destinations=destinations_template_str)# 通过路由提示词模板,构建路由提示词
router_prompt = PromptTemplate(
    template=router_template,
    input_variables=["input"],
    output_parser=RouterOutputParser(),)######### 路由链
router_chain = LLMRouterChain.from_llm(llm=model, prompt=router_prompt)######### 最终的链
multi_prompt_chain = MultiPromptChain(
    router_chain=router_chain,
    destination_chains=destination_chains,
    default_chain=default_chain,
    verbose=True,)# multi_prompt_chain.invoke({"input": "重力加速度是多少?"})# multi_prompt_chain.invoke("y=x^2+2x+1的导数是多少?")multi_prompt_chain.invoke("将以下英文翻译成中文,只输出中文翻译结果:\n The largest community building the future of LLM apps.")# multi_prompt_chain.invoke("你是怎么理解java的面向对象的思想的?")1.2.3.4.5.6.7.8.9.10.11.12.13.14.15.16.17.18.19.20.21.22.23.24.25.26.27.28.29.30.31.32.33.34.35.36.37.38.39.40.41.42.43.44.45.46.47.48.49.50.51.52.53.54.55.56.57.58.59.60.61.62.63.64.65.66.67.68.69.70.71.72.73.74.75.76.77.78.79.80.81.82.83.84.85.86.87.88.

执行结果跟我们预想的一致,执行结果如下:

五分钟了解 LangChain 的路由链

五分钟了解 LangChain 的路由链

五分钟了解 LangChain 的路由链

4. 总结

这篇博客主要介绍了LangChain中的路由链(RouterChain)的概念,它主要用在不确定性的场景下,根据提示词,选择具体的某个链去执行。还聊了它的使用场景和具体案例,希望对你有帮助!


您需要 登录账户 后才能发表评论

发表评论

快捷回复: 表情:
AddoilApplauseBadlaughBombCoffeeFabulousFacepalmFecesFrownHeyhaInsidiousKeepFightingNoProbPigHeadShockedSinistersmileSlapSocialSweatTolaughWatermelonWittyWowYeahYellowdog
评论列表 (暂无评论,144人围观)

还没有评论,来说两句吧...

目录[+]

取消
微信二维码
微信二维码
支付宝二维码