LangChain 아키텍처: 모델, 프롬프트, 체인

LangChain을 사용하는 이유

세 가지 핵심 개념

모델

from langchain_openai import ChatOpenAI
from langchain_anthropic import ChatAnthropic

openai_model = ChatOpenAI(model='gpt-4o-mini', temperature=0.2)
claude_model = ChatAnthropic(model='claude-sonnet-4-5')

resp = openai_model.invoke('Hello!')
print(resp.content)

프롬프트 템플릿

from langchain.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages([
    ('system', 'You are a {role} assistant.'),
    ('user', '{question}')
])

messages = prompt.invoke({'role': 'SQL', 'question': 'How do I delete rows?'})
print(messages.to_messages())

체인: 파이프 연산자

from langchain.schema.output_parser import StrOutputParser

chain = prompt | openai_model | StrOutputParser()
result = chain.invoke({'role': 'SQL', 'question': 'How do I delete rows?'})
print(result)

출력 파서

from langchain.output_parsers import PydanticOutputParser
from pydantic import BaseModel

class Plan(BaseModel):
    steps: list[str]

parser = PydanticOutputParser(pydantic_object=Plan)
chain = prompt | model | parser

실행 가능 객체와 실행 가능 프로토콜

Streaming

for chunk in chain.stream({'role': 'helpful', 'question': 'Tell a story.'}):
    print(chunk, end='', flush=True)

Batch

results = chain.batch([
    {'role': 'SQL', 'question': 'Q1'},
    {'role': 'SQL', 'question': 'Q2'}
])
# Runs in parallel.

실행 중 구성

result = chain.invoke(
    {'role': 'helpful', 'question': '...'},
    config={'configurable': {'temperature': 0.0}}
)

LangChain Hub

from langchain import hub
rag_prompt = hub.pull('rlm/rag-prompt')

LangChain을 NOT 사용할 때

Chain 연산자

복습