Rag 기반 챗봇 만들기

1. PDF 기반 학습 챗봇

!pip install langchain_community
!pip install pypdf
!pip install tiktoken
!pip install faiss-cpu

 

from langchain_community.document_loaders import PyPDFLoader

api_key="sk-proj-본인의 API키"
pdf_filepath = '/content/차세대 한국형 스마트팜 개발.pdf'
loader = PyPDFLoader(pdf_filepath)
data  = loader.load()

 

from langchain.text_splitter import CharacterTextSplitter

text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
data = text_splitter.split_documents(data)

 

from langchain.embeddings.openai import OpenAIEmbeddings

embeddings = OpenAIEmbeddings(model="text-embedding-ada-002", api_key=api_key)

 

from langchain.vectorstores import FAISS

vectorstore = FAISS.from_documents(data, embeddings)

 

from langchain.chat_models import ChatOpenAI

llm = ChatOpenAI(model="gpt-3.5-turbo", api_key=api_key)

 

from langchain.memory import ConversationBufferMemory

memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)

 

from langchain.chains import ConversationalRetrievalChain
conversation_chain = ConversationalRetrievalChain.from_llm(
    llm=llm,
    chain_type="stuff",
    retriever=vectorstore.as_retriever(),
    memory=memory
)

 

query = "한국형 스마트팜이란?"
result = conversation_chain({"question": query})
answer = result["answer"]
answer

 

query = "스마트팜 농가의 통합관리 및 유지보수가 어려운 이유는?"
result = conversation_chain({"question": query})
answer = result["answer"]
answer

 

 

2. Gradio를 이용한 간단한 챗봇

import gradio as gr
from langchain.chat_models import ChatOpenAI
from langchain.prompts import PromptTemplate
import os

# OpenAI API 키 설정 (환경 변수 또는 직접 입력)
os.environ["OPENAI_API_KEY"] = api_key  # 여기에 OpenAI API 키를 입력하세요

# 프롬프트 템플릿 설정
prompt = PromptTemplate(
    input_variables=["chat_history", "question"],
    template="""You are an AI assistant. You are
    currently having a conversation with a human. Answer the questions.
    
    chat_history: {chat_history},
    Human: {question}
    AI:"""
)

# LLM 모델 설정
llm = ChatOpenAI(
    temperature=0,  # 창의성 설정
    model_name='gpt-4o-mini-2024-07-18'  # 모델명
)

# 최신 메모리 사용 방식 적용 (ConversationBufferWindowMemory 제거)
def chat_with_ai(user_input, history):
    """사용자 입력을 받아 AI 응답을 반환하는 함수."""
    chat_history = "\n".join([f"Human: {h['content']}" if h['role'] == 'user' else f"AI: {h['content']}" for h in history])
    response = (prompt | llm).invoke({"chat_history": chat_history, "question": user_input}).content
    return {"role": "assistant", "content": response}

# Gradio 인터페이스 설정
demo = gr.ChatInterface(chat_with_ai, type="messages")

demo.launch(share=True, debug=True)