今天内容涉及如下：
（图片来源网络，侵删）
1.initialize_agent,：执行gent工作，并把工具Tool传入
2.Tool：选取行为函数工具类
（图片来源网络，侵删）
之前我们学习的都是把问题给AI，让AI模型给出答案，那么这种情况下应该怎么处理呢，我需要根据不同的问题选择不同的答案，比如我问AI我想选择一件衣服就去调用挑选衣服的方法，如果是查询订单，那么就专门去调用搜索订单的方法，如果是查询物流就专门去调用物流方面的方法，但是怎么识别出来哪个调用哪个呢？
我们下面代码先模拟出怎么让AI根据我们的话语做选择，
#! pip install openai
! pip install langchain
import openai,os
from langchain.prompts import PromptTemplate
from langchain.llms import OpenAIChat
from langchain.chains import LLMChain
os.environ["OPENAI_API_KEY"] = ""
openai.api_key = os.environ.get("OPENAI_API_KEY")
llm=OpenAIChat(max_tokens=2048,temperature=0.5)
multiple_choice="""
请针对 >>> 和 >{question} Entering new AgentExecutor chain...
I need to find out the shipping policy and delivery time.
Action: FAQ
Action Input: shipping policy and delivery time
> Entering new VectorDBQA chain...
> Finished chain.
Observation: 我们支持全国大部分省份的配送，一般情况下，大部分城市的订单在2-3个工作日内送达，偏远地区可能需要5-7个工作日。具体送货时间可能因订单商品、配送地址和物流公司而异。
Thought: I now know the final answer
Final Answer: 我们支持全国大部分省份的配送，一般情况下，大部分城市的订单在2-3个工作日内送达，偏远地区可能需要5-7个工作日。具体送货时间可能因订单商品、配送地址和物流公司而异。
> Finished chain.
我们支持全国大部分省份的配送，一般情况下，大部分城市的订单在2-3个工作日内送达，偏远地区可能需要5-7个工作日。具体送货时间可能因订单商品、配送地址和物流公司而异。
对于商品的推荐，我们也可以使用类似方式，也把对应的商品信息，存到 VectorStore 里，然后通过先搜索后问答的方式来解决。
关于商品csv文件资源链接：
https://download.csdn.net/download/dfBeautifulLive/88393780?spm=1001.2014.3001.5503
from langchain.text_splitter import CharacterTextSplitter
from langchain.document_loaders import CSVLoader
product_loader = CSVLoader('./data/ecommerce_products.csv')
product_documents = product_loader.load()
product_text_splitter = CharacterTextSplitter(chunk_size=1024, separator="\n")
product_texts = product_text_splitter.split_documents(product_documents)
product_search = FAISS.from_documents(product_texts, OpenAIEmbeddings())
product_chain = VectorDBQA.from_chain_type(llm=llm, vectorstore=product_search, verbose=True)
@tool("FAQ")
def faq(intput: str) -> str:
""""useful for when you need to answer questions about shopping policies, like return policy, shipping policy, etc."""
return faq_chain.run(intput)
@tool("Recommend Product")
def recommend_product(input:str) ->str:
""""useful for when you need to search and recommend products and recommend it to the user"""
return product_chain.run(input)
tools=[
recommend_product,
faq
]
agent = initialize_agent(tools, llm, agent="zero-shot-react-description", verbose=True)
问答：
question = "我想买一件衣服，想要在春天去公园穿，但是不知道哪个款式好看，你能帮我推荐一下吗？" answer = agent.run(question) print(answer)
结果：
> Entering new AgentExecutor chain... I need to find a product that is suitable for the user. Action: Recommend Product Action Input: Clothes for park in spring > Entering new VectorDBQA chain... > Finished chain. Observation: 长款风衣、卫衣连衣裙、长款卫衣。 Thought: I now know the final answer Final Answer: 我建议你可以考虑长款风衣、卫衣连衣裙、长款卫衣，这些款式都很适合在春天去公园穿。 > Finished chain. 我建议你可以考虑长款风衣、卫衣连衣裙、长款卫衣，这些款式都很适合在春天去公园穿。
4.优化 Prompt，让 AI 不要胡思乱想
关于订单号的搜索，我们用向量库不合适，可以用数据库查询就可以，我们模拟一下不同的订单号匹配，然后匹配到返回什么，匹配不到返回什么，业务如下：
import json
import re
ORDER_1="20230101ABC"
ORDER_2="20230101EFG"
ORDER_1_DETAIL={
"order_number":ORDER_1,
"status":"已发货",
"shipping_date": "2023-01-03",
"estimated_delivered_date": "2023-01-05",
}
ORDER_2_DETAIL={
"order_number":ORDER_2,
"status":"未发货",
"shipping_date": None,
"estimated_delivered_date": None,
}
@tool("SearchOrder")
def search_order(input:str)->str:
"""useful for when you need to answer questions about customers orders"""
if input.strip()==ORDER_1:
return json.dumps(ORDER_1_DETAIL)
elif input.strip()==ORDER_2:
return json.dumps(ORDER_2_DETAIL)
else:
return f"对不起，根据{input}没有找到您的订单"
tools=[search_order,recommend_product,faq]
agent=initialize_agent(tools,llm=OpenAI(temperature=0),agent="zero-shot-react-description", verbose=True)
在这里我们匹配到的就直接返回了和我们期望一致的数据，那如果我们查询一下上面没有的订单号，也就是匹配不上，会出现什么？
question = "我有一张订单，订单号是 2022ABCDE，一直没有收到，能麻烦帮我查一下吗？" answer = agent.run(question) print(answer)
结果：
> Entering new AgentExecutor chain... I need to find out the status of the order Action: SearchOrder Action Input: 2022ABCDE Observation: 对不起，根据2022ABCDE没有找到您的订单 Thought: I need to find out more information about the order Action: SearchOrder Action Input: 2022ABCDE Observation: 对不起，根据2022ABCDE没有找到您的订单 Thought: I need to provide more information about the order Action: FAQ Action Input: 订单查询 > Entering new VectorDBQA chain... WARNING:langchain.llms.base:Retrying langchain.llms.openai.completion_with_retry.._completion_with_retry in 4.0 seconds as it raised RateLimitError: Rate limit reached for default-text-davinci-003 in organization org-WdebHKFlmMZipWcgP2HMj6CQ on requests per min. Limit: 3 / min. Please try again in 20s. Contact us through our help center at help.openai.com if you continue to have issues. Please add a payment method to your account to increase your rate limit. Visit https://platform.openai.com/account/billing to add a payment method.. WARNING:langchain.llms.base:Retrying langchain.llms.openai.completion_with_retry.._completion_with_retry in 4.0 seconds as it raised RateLimitError: Rate limit reached for default-text-davinci-003 in organization org-WdebHKFlmMZipWcgP2HMj6CQ on requests per min. Limit: 3 / min. Please try again in 20s. Contact us through our help center at help.openai.com if you continue to have issues. Please add a payment method to your account to increase your rate limit. Visit https://platform.openai.com/account/billing to add a payment method.. WARNING:langchain.llms.base:Retrying langchain.llms.openai.completion_with_retry.._completion_with_retry in 4.0 seconds as it raised RateLimitError: Rate limit reached for default-text-davinci-003 in organization org-WdebHKFlmMZipWcgP2HMj6CQ on requests per min. Limit: 3 / min. Please try again in 20s. Contact us through our help center at help.openai.com if you continue to have issues. Please add a payment method to your account to increase your rate limit. Visit https://platform.openai.com/account/billing to add a payment method.. WARNING:langchain.llms.base:Retrying langchain.llms.openai.completion_with_retry.._completion_with_retry in 8.0 seconds as it raised RateLimitError: Rate limit reached for default-text-davinci-003 in organization org-WdebHKFlmMZipWcgP2HMj6CQ on requests per min. Limit: 3 / min. Please try again in 20s. Contact us through our help center at help.openai.com if you continue to have issues. Please add a payment method to your account to increase your rate limit. Visit https://platform.openai.com/account/billing to add a payment method.. WARNING:langchain.llms.base:Retrying langchain.llms.openai.completion_with_retry.._completion_with_retry in 4.0 seconds as it raised RateLimitError: Rate limit reached for default-text-davinci-003 in organization org-WdebHKFlmMZipWcgP2HMj6CQ on requests per min. Limit: 3 / min. Please try again in 20s. Contact us through our help center at help.openai.com if you continue to have issues. Please add a payment method to your account to increase your rate limit. Visit https://platform.openai.com/account/billing to add a payment method.. > Finished chain. Observation: 登录您的帐户，点击“我的订单”，在此页面上，您可以查看所有订单及其当前状态。 Thought: WARNING:langchain.llms.base:Retrying langchain.llms.openai.completion_with_retry.._completion_with_retry in 4.0 seconds as it raised RateLimitError: Rate limit reached for default-text-davinci-003 in organization org-WdebHKFlmMZipWcgP2HMj6CQ on requests per min. Limit: 3 / min. Please try again in 20s. Contact us through our help center at help.openai.com if you continue to have issues. Please add a payment method to your account to increase your rate limit. Visit https://platform.openai.com/account/billing to add a payment method.. WARNING:langchain.llms.base:Retrying langchain.llms.openai.completion_with_retry.._completion_with_retry in 4.0 seconds as it raised RateLimitError: Rate limit reached for default-text-davinci-003 in organization org-WdebHKFlmMZipWcgP2HMj6CQ on requests per min. Limit: 3 / min. Please try again in 20s. Contact us through our help center at help.openai.com if you continue to have issues. Please add a payment method to your account to increase your rate limit. Visit https://platform.openai.com/account/billing to add a payment method.. WARNING:langchain.llms.base:Retrying langchain.llms.openai.completion_with_retry.._completion_with_retry in 8.0 seconds as it raised RateLimitError: Rate limit reached for default-text-davinci-003 in organization org-WdebHKFlmMZipWcgP2HMj6CQ on requests per min. Limit: 3 / min. Please try again in 20s. Contact us through our help center at help.openai.com if you continue to have issues. Please add a payment method to your account to increase your rate limit. Visit https://platform.openai.com/account/billing to add a payment method.. I now know the final answer Final Answer: 登录您的帐户，点击“我的订单”，在此页面上，您可以查看所有订单及其当前状态。 > Finished chain. 登录您的帐户，点击“我的订单”，在此页面上，您可以查看所有订单及其当前状态。
最终返回的结果不是我们要的结果，想要让它返回对不起，根据2022ABCDE没有找到您的订单，但他最后却重复调用 OpenAI 的思考策略，并最终尝试从 FAQ 里拿一个查询订单的问题来敷衍用户。
解决的方案就是更改下提示语，通过这个提示语，Agent 会知道，这个工具就应该在找不到订单的时候，告诉用户找不到订单或者请它再次确认。
@tool("SearchOrder")
def search_order(input:str)->str:
"""一个帮助用户查询最新订单状态的工具，并且能处理以下情况：
1. 在用户没有输入订单号的时候，会询问用户订单号
2. 在用户输入的订单号查询不到的时候，会让用户二次确认订单号是否正确"""
# 匹配多个数字以及大写字母
pattern=r"\d+[A-Z]+"
match=re.search(pattern,input)
order_number=input
if match:
# 得到整个订单字符串
order_number=match.group(0)
else:
return "请问您的订单号是多少？"
if order_number==ORDER_1:
return json.dumps(ORDER_1_DETAIL)
elif order_number==ORDER_2:
return json.dumps(ORDER_2_DETAIL)
else:
return f"对不起，根据{input}没有找到您的订单"
tools=[search_order,recommend_product,faq]
agent=initialize_agent(tools,llm=OpenAI(temperature=0),agent="zero-shot-react-description", verbose=True)
question = "我有一张订单，订单号是 2022ABCDE，一直没有收到，能麻烦帮我查一下吗？"
answer = agent.run(question)
print(answer)
结果：
> Entering new AgentExecutor chain... 我需要查询订单状态 Action: SearchOrder Action Input: 2022ABCDE Observation: 对不起，根据2022ABCDE没有找到您的订单 Thought: 我需要再次确认订单号是否正确 Action: SearchOrder Action Input: 2022ABCDE Observation: 对不起，根据2022ABCDE没有找到您的订单 Thought: 我现在知道最终答案 Final Answer: 对不起，根据您提供的订单号2022ABCDE没有找到您的订单，请确认订单号是否正确。 > Finished chain. 对不起，根据您提供的订单号2022ABCDE没有找到您的订单，请确认订单号是否正确。
5.订单查询多伦对话
我们再继续优化：
我们应该支持多轮聊天。因为用户不一定是在第一轮提问的时候，就给出了自己的订单号。
我们其实可以直接让 Search Order 这个 Tool，回答用户的问题，没有必要再让 Agent 思考一遍，使用return_direct这个参数就不回在经过Thought，直接把回答传给用户。
from langchain.memory import ConversationBufferMemory
from langchain.chat_models import ChatOpenAI
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain
answer_order_info=PromptTemplate(
template="请把下面的订单信息回复给用户： \n\n {order}?",
input_variables=["order"]
)
answer_order_llm=LLMChain(llm=ChatOpenAI(temperature=0), prompt=answer_order_info)
# return_direct这个参数是不要再经过Thought，直接把回答传给用户
@tool("SearchOrder",return_direct=True)
def search_order(input:str)->str:
"""useful for when you need to answer questions about customers orders"""
pattern=r"\d+[A-Z]+"
match=re.search(pattern,input)
order_number=input
if match:
# 取出整个订单字符集
order_number=match.group(0)
else:
return "请问您的订单号是多少？"
if order_number==ORDER_1:
# 通过 answer_order_llm 这个工具来组织语言文字，将python对象编码成Json字符串
return answer_order_llm.run(json.dumps(ORDER_1_DETAIL))
elif order_number==ORDER_2:
return answer_order_llm.run(json.dumps(ORDER_2_DETAIL))
else:
return f"对不起，根据{input}没有找到您的订单"
tools=[search_order,recommend_product,faq]
# 更换为ChatOpenAI成本更低
chatllm=ChatOpenAI(temperature=0)
# 保留对话记录到内存
memory=ConversationBufferMemory(memory_key="chat_history",return_message=True)
# 更改agent为conversational-react-description支持多语言对话
conversation_agent=initialize_agent(tools,chatllm,
agent="conversational-react-description",
memory=memory,
verbose=True)
问题一：
question1 = "我有一张订单，一直没有收到，能麻烦帮我查一下吗？" answer1 = conversation_agent.run(question1) print(answer1)
结果：
> Entering new AgentExecutor chain... Thought: Do I need to use a tool? Yes Action: SearchOrder Action Input: 我有一张订单，一直没有收到 Observation: 请问您的订单号是多少？ > Finished chain. 请问您的订单号是多少？
问题二：
question2 = "我的订单号是20230101ABC" answer2 = conversation_agent.run(question2) print(answer2)
结果：
> Entering new AgentExecutor chain... Thought: Do I need to use a tool? Yes Action: SearchOrder Action Input: 20230101ABC Observation: 尊敬的用户，以下是您的订单信息： 订单编号：20230101ABC 订单状态：已发货 发货日期：2023年1月3日 预计送达日期：2023年1月5日 如有任何问题，请随时与我们联系。感谢您的支持！ > Finished chain. 尊敬的用户，以下是您的订单信息： 订单编号：20230101ABC 订单状态：已发货 发货日期：2023年1月3日 预计送达日期：2023年1月5日 如有任何问题，请随时与我们联系。感谢您的支持！
本篇文章视频：
让AI做决策，学会langChain的Agent_哔哩哔哩_bilibili
