🔗 My_LangChain

📋 프로젝트 개요

📝 개선 예정 사항

🚀 프로젝트 구성

1. ChatbotWithPDF

사용법

def main():
 chat = ChatbotWithPDF(pdf = "https://snuac.snu.ac.kr/2015_snuac/wp-content/uploads/2015/07/asiabrief_3-26.pdf")
 print('첫 번째 질문')
 chat('저출산을 극복한 나라들은 어디가 있어?')
 print('두 번째 질문')
 chat('최신 관련 자료들의 문헌 제목을 알려줘')

ChatbotWithPDF 클래스

class ChatbotWithPDF:
 def __init__(self, pdf: str):
 self.llm = ChatOpenAI();
 
 loader = PyPDFLoader(pdf)
 pages = loader.load()
 text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=150)
 splits = text_splitter.split_text(pages)
 self.vectorestore = Chroma.from_texts(texts=splits)
 
 def __call__(self, query: str):
 doc = self.vectorestore.similarity_search(query)
 self.qa = LLMChain.from_llm(llm = self.llm, retriver=doc)
 print(self.qa(query))

필요 모듈

from my_langchain.chains.llm_chain import LLMChain
from my_langchain.text_splitter import RecursiveCharacterTextSplitter
from my_langchain.document_loaders import PyPDFLoader
from my_langchain.vectorstores import Chroma
from my_langchain.chat_models import ChatOpenAI

2. ChatbotWithWeb

사용법

def main():
 chat2 = ChatbotWithWeb(url = 'http://18children.president.pa.go.kr/mobile/our_space/fairy_tales.php?srh%5Bcategory%5D=07&srh%5Bview_mode%5D=detail&srh%5Bseq%5D=1204')
 chat2('삼형제의 재주는 각각 무엇인가요?')
 chat2('누구의 음식을 훔쳤나요?') 

ChatbotWithWeb 클래스

class ChatbotWithWeb:
 def __init__(self, url: str):
 self.llm = ChatOpenAI();
 
 loader = WebBaseLoader(url)
 pages = loader.load()
 text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=200)
 splits = text_splitter.split_text(pages)
 self.vectorestore = Chroma.from_texts(texts=splits)
 
 def __call__(self, query: str):
 doc = self.vectorestore.similarity_search(query)
 self.qa = LLMChain.from_llm(llm = self.llm, retriver=doc)
 print(self.qa(query))

필요 모듈

from langchain_text_splitters import RecursiveCharacterTextSplitter
from my_langchain.chains.llm_chain import LLMChain
from my_langchain.vectorstores import Chroma
from my_langchain.chat_models import ChatOpenAI
from my_langchain.document_loaders import WebBaseLoader

3. ConversationBot

사용법

def main():
 #보기에 심심해 보여서 chat gpt를 통해 gui로 변경
 import tkinter as tk
 bot = ConversationBot()
 def send_message(event=None):
 user_message = entry_field.get()
 if user_message.strip() == '':
 return
 chat_log.config(state=tk.NORMAL)
 chat_log.insert(tk.END, "You: " + user_message + "\n\n")
 chat_log.config(state=tk.DISABLED)
 entry_field.delete(0, tk.END)
 receive_message(f"{bot.conversation(user_message)}.")
 def receive_message(response):
 chat_log.config(state=tk.NORMAL)
 chat_log.insert(tk.END, "Bot: " + response + "\n\n")
 chat_log.config(state=tk.DISABLED)
 # GUI 창 생성
 root = tk.Tk()
 root.title("Chatbot")
 # 대화 내용 표시할 곳
 chat_log = tk.Text(root, bd=0, height=8, width=50, font="Arial")
 chat_log.config(state=tk.DISABLED)
 scrollbar = tk.Scrollbar(root, command=chat_log.yview, cursor="heart")
 chat_log['yscrollcommand'] = scrollbar.set
 # 입력 필드
 entry_field = tk.Entry(root, bg="white", width=50)
 entry_field.bind("<Return>", send_message)
 # 전송 버튼
 send_button = tk.Button(root, text="Send", command=send_message)
 # 배치
 scrollbar.pack(side=tk.RIGHT, fill=tk.Y)
 chat_log.pack(side=tk.TOP, fill=tk.BOTH, expand=True)
 entry_field.pack(side=tk.LEFT, fill=tk.BOTH, expand=True)
 send_button.pack(side=tk.RIGHT)
 # GUI 시작
 root.mainloop()

ConversationBot 클래스

class ConversationBot:
 def __init__(self):
 llm = ChatOpenAI()
 memory = ConversationBufferMemory()
 self.chain = ConversationChain.from_llm(llm=llm, memory=memory, retriver=None)
 
 def conversation(self, txt):
 return self.chain(txt)

필요 모듈

from my_langchain.chat_models import ChatOpenAI
from my_langchain.memory import ConversationBufferMemory
from my_langchain.chains import ConversationChain

4. MyGoogleAgent

사용법

def main():
 # 1. Get user's question in English
 user_query = input("Enter your question in English: ")
 # 2. Search Google with the user's question
 google_query = make_query(user_query)
 urls = search_google(google_query)
 # 3. Provide answer using RAG approach with the returned URL
 play_rag(user_query, urls)

play_rag함수와 make_query함수

def play_rag(query, urls):
 llm = ChatOpenAI()
 for url in urls:
 loader = WebBaseLoader(url)
 text = loader.load()
 if text == ' ':
 continue
 splitter = RecursiveCharacterTextSplitter(chunk_size=700, chunk_overlap=0)
 text_chunks = splitter.split_text(text)
 
 db = Chroma.from_texts(text_chunks)
 retriever = db.similarity_search(query)
 if retriever == [' ']:
 continue
 
 chain = LLMChain(llm, retriever=retriever)
 response = chain(query)
 print(response)
 return response
 
 print('죄송합니다. 정보를 찾을 수 없습니다.')
def make_query(query):
 llm = ChatOpenAI()
 return llm.predict(f'to aquire the resopnse about "{query}", make the a proper question, and just only say that in English.')

필요 모듈

from my_langchain.agents.tool import search_google
from my_langchain.document_loaders import WebBaseLoader
from my_langchain.text_splitter import RecursiveCharacterTextSplitter
from my_langchain.chat_models import ChatOpenAI
from my_langchain.vectorstores import Chroma
from my_langchain.chains import LLMChain