基于LangChain的新闻相声

事先需要下载

!pip install -U langchain-community
!pip install -U langchain
!pip install -U pyPDF
!pip install tiktoken
!pip install langchain_openai
!pip install unstructured
!pip install gradio

相关代码

from langchain.document_loaders import UnstructuredURLLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
url='https://news.sina.com.cn/zx/gj/2024-12-20/doc-ineaarnv9239635.shtml'
def url3News(url):
  text_splitter=RecursiveCharacterTextSplitter(separators=["正文","统筹"],chunk_size=1000,chunk_overlap=20,length_function=len)
  loader=UnstructuredURLLoader(urls=[url])
  data=loader.load_and_split(text_splitter=text_splitter)
  return data[1:2]
new_data=url3News(url)
print(new_data)

from langchain.chains.summarize import load_summarize_chain
from langchain.chat_models import ChatOpenAI
from langchain.prompts import PromptTemplate
from langchain.schema import HumanMessage
from pydantic import BaseModel,Field
from langchain.output_parsers import PydanticOutputParser

class Line(BaseModel):
  character:str =Field(description="说这句台词的角色的名字")
  content:str = Field(description="台词的具体内容，其中不再包括角色的名字")
class XiangSheng(BaseModel):
  script:list[Line]=Field(description="一段相声的台词剧本")

def news2scripte(new_data):
  prompt_template = """总结这段新闻的内容：
  "{text}"

  总结："""
  chinese_template = PromptTemplate(template=prompt_template, input_variables=["text"])
  llm = ChatOpenAI(base_url="http://chatapi.littlewheat.com/v1", openai_api_key="",max_tokens=1500)

  # Use map_prompt and combine_prompt instead of prompt
  chain = load_summarize_chain(llm, chain_type="map_reduce", map_prompt=chinese_template, combine_prompt=chinese_template)
  summary = chain.run(new_data)


  openaichat=ChatOpenAI(base_url="http://chatapi.littlewheat.com/v1", openai_api_key="",max_tokens=1500)
  template="""\
    我将给你一段新闻的概况，请按照要求把这段新闻改写成郭德纲和于谦的对口相声剧本

    新闻：{新闻}
    要求：{要求}
    {output_instructions}
  """
  parser=PydanticOutputParser(pydantic_object=XiangSheng)
  prompt=PromptTemplate(template=template,input_variables=["新闻","要求"],partial_variables={"output_instructions":parser.get_format_instructions()})

  msg=[HumanMessage(content=prompt.format(新闻=summary,要求="风趣幽默，十分讽刺，剧本对话角色为郭德纲和于谦，以他们的自我介绍为开头"))]

  res=openaichat(msg)
  xiangsheng=parser.parse(res.content)
  return xiangsheng
re=news2scripte(new_data)
print(re)

封装

def url2xiangsheng(url):
  doc=url3News(url)
  xiangsheng=news2scripte(doc)
  return xiangsheng

交互界面

import gradio as gr

with gr.Blocks() as demo:
  url = gr.Textbox()
  chatbot = gr.Chatbot()
  submit_btn = gr.Button("生成相声")
  def generate_conversation(url):
    xiangsheng: XiangSheng = url2xiangsheng(url)
    chat_history = []
    def parse_line(line: Line):
      if line is None:
        return ""
      return f"{line.character}:{line.content}"
    for i in range(0, len(xiangsheng.script), 2):
      line1 = xiangsheng.script[i]
      line2 = xiangsheng.script[i+1] if (i+1) < len(xiangsheng.script) else None
      chat_history.append((parse_line(line1), parse_line(line2)))
    return chat_history
  submit_btn.click(fn=generate_conversation, inputs=url, outputs=chatbot)

demo.launch()