LangChain的Memory的4种用法（一）_conversationbuffermemory

什么是LangChain

LangChain 是一个基于语言模型开发应用程序的开源框架，用于构建基于大语言模型（LLMs）的端到端语言模型应用。它的核心理念是为各种大语言模型应用提供通用的接口，简化开发过程，并提供一套工具、组件和接口来创建由LLMs和聊天模型支持的应用程序。LangChain提供了多种组件来简化我们开发大语言模型应用的难度，本文来介绍其中Memory的使用方式。

Memory的作用

大多数 LLM 模型都有一个会话接口，当我们使用接口调用大模型能力时，每一次的调用都是新的一次会话。如果我们想和大模型进行多轮的对话，而不必每次重复之前的上下文时，就需要一个Memory来记忆我们之前的对话内容。

Memory就是这样的一个模块，来帮助开发者可以快速的构建自己的应用“记忆”。

用法一：Conversation buffer memory

Conversation buffer memory 是最简单的一种memory，它会把之前的对话信息全部记录下来。

import os
 
from dotenv import load_dotenv, find_dotenv
from langchain.chat_models import ChatOpenAI
from langchain.chains import ConversationChain
from langchain.memory import ConversationBufferMemory
 
llm = ChatOpenAI(temperature=0.0)
 
# 创建一个对话ConversationBufferMemory的缓存
memory = ConversationBufferMemory()
 
#创建一个对话链
conversation = ConversationChain(
    llm=llm, 
    memory = memory,
    verbose=True
)

通过ConversationChain向LLM介绍，提出第一个问题

conversation.predict(input="Hi, my name is Andrew")

LLM回答第一个问题

> Entering new  chain...
Prompt after formatting:
The following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.
 
Current conversation:
 
Human: Hi, my name is Andrew
AI:
 
> Finished chain.
 
"Hello Andrew! It's nice to meet you. How can I assist you today?"

向LLM提出第二个问题

conversation.predict(input="What is 1+1?")

LLM回答第二问题

> Entering new  chain...
Prompt after formatting:
The following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.
 
Current conversation:
Human: Hi, my name is Andrew
AI: Hello Andrew! It's nice to meet you. How can I assist you today?
Human: What is 1+1?
AI:
> Finished chain.
'1+1 is equal to 2.'

向LLM再次询问第一个问题的信息，LLM正确的回答出来了

# 可以记住上一次的对话提到的人类的名字
conversation.predict(input="What is my name?")

> Entering new  chain...
Prompt after formatting:
The following is a friendly conversation between a human and an AI. The AI is talkative and provides lots of specific details from its context. If the AI does not know the answer to a question, it truthfully says it does not know.
 
Current conversation:
Human: Hi, my name is Andrew
AI: Hello Andrew! It's nice to meet you. How can I assist you today?
Human: What is 1+1?
AI: 1+1 is equal to 2.
Human: What is my name?
AI:
> Finished chain.
'Your name is Andrew.'

查看memory的变量

# 加载缓存的变量内容
memory.load_memory_variables({})

{'history': "Human: Hi, my name is Andrew\nAI: Hello Andrew! It's nice to meet you. How can I assist you today?\nHuman: What is 1+1?\nAI: 1+1 is equal to 2.\nHuman: What is my name?\nAI: Your name is Andrew."}

我们也可以手动添加memory的内容

# 手动添加缓存的内容
memory.save_context({"input": "Hi"}, 
                    {"output": "What's up"})

用法二：Conversation buffer window memory

上面的 Conversation buffer memory 会随着我们和LLM的对话次数的增长而不断的增加，这个时候memory就可能变大的很大，可能会超出LLM的tokens限制。Conversation buffer window memory 提供了一个滑动窗口的memory，它可以记录最后K轮对话的内容（一般我们认为越远离我们的对话，可能与我们当前讨论的主题无关）。

from langchain.memory import ConversationBufferWindowMemory
# 创建一个对话缓存的窗口（k=1）
memory = ConversationBufferWindowMemory(k=1) 
 
# 连续两次添加对话内容
memory.save_context({"input": "Hi"},
                    {"output": "What's up"})
memory.save_context({"input": "Not much, just hanging"},
                    {"output": "Cool"})

手动添加了两轮对话内容后，去查看memory的变量，发现只保持了最后一轮的内容（之前设置是k=1）。

# 只能记忆最后一次的对话内容
memory.load_memory_variables({})

{'history': 'Human: Not much, just hanging\nAI: Cool'}

通过ConversationChain测试对话能力

# 创建conversation chain并且设置memory
llm = ChatOpenAI(temperature=0.0)
memory = ConversationBufferWindowMemory(k=1)
conversation = ConversationChain(
    llm=llm, 
    memory = memory,
    verbose=False
)

向LLM提出第一个问题

conversation.predict(input="Hi, my name is Andrew")

"Hello Andrew! It's nice to meet you. How can I assist you today?"

向LLM提出第二个问题

conversation.predict(input="What is 1+1?")

'1+1 is equal to 2.'

向LLM询问第一个问题相关的内容，发现LLM无法回答

# memory只能记住最后一次的对话内容
conversation.predict(input="What is my name?")

"I'm sorry, but I don't have access to personal information."