Function Calling + LangChain 拉通业务系统的技术架构

import openaiimport osimport tiktoken# 加载 .env 文件from dotenv import load_dotenv, find_dotenvfrom langchain.prompts import PromptTemplatefrom langchain.llms import OpenAIfrom langchain.chains import LLMChainfrom langchain.chains import LLMRequestsChain#from langchain.chat_models import AzureChatOpenAIfrom langchain.chat_models import ChatOpenAI #直接访问OpenAI的GPT服务
_ = load_dotenv(find_dotenv())# 从环境变量中获得你的 OpenAI Key和配置URLopenai.api_key = os.getenv('OPENAI_API_KEY')openai.api_base = os.getenv('OPENAI_API_URL')model = os.getenv('OPENAI_API_MODEL')


llm = ChatOpenAI(model_name=model, temperature=0) #直接访问OpenAI的GPT服务#llm = AzureChatOpenAI( model_name=model, temperature=0, max_tokens=200) # 通过Azure的OpenAI服务#根据查询的结果结果返回给大模型，大模型再组装后进行返回def query_baidu(question):      template = """Between >>> and <<< are the raw search result text from web.      Extract the answer to the question '{query}' or say "not found" if the information is not contained.      Use the format      Extracted:<answer or "not found">      >>> {requests_result} <<<      Extracted:"""
      PROMPT = PromptTemplate(          input_variables=["query", "requests_result"],          template=template,      )
      inputs = {          "query": question,          "url": "http://www.baidu.com/s?wd=" + question.replace(" ", "+")      }      requests_chain = LLMRequestsChain(llm_chain = LLMChain(llm=llm, prompt=PROMPT), output_key="query_info", verbose=True)      res = requests_chain.run(inputs)      return res
#python 程序入口if __name__ == "__main__":   print(query_baidu("今天长沙的天气？"))