LlamaIndex의 RAG pipeline에 대한 스터디

# 필요한 라이브러리와 모듈 임포트
from dotenv import load_dotenv
from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
from llama_index.llms.openai import OpenAI
from llama_index.embeddings.openai import OpenAIEmbedding
from llama_index.core import Settings
import chromadb

# LlamaIndex Settings
Settings.llm = OpenAI(model="gpt-4o-mini", temperature=0)
Settings.embed_model = OpenAIEmbedding(model="text-embedding-3-small")

# .env 파일에서 환경 변수(OPENAI_API_KEY) 로드
load_dotenv()

# Chromadb 관련 모듈 임포트
from llama_index.vector_stores.chroma import ChromaVectorStore
from llama_index.core import StorageContext

# Chromadb 클라이언트 초기화
db = chromadb.PersistentClient(path="./chroma_db")

# "quickstart" 컬렉션 생성 또는 가져오기
chroma_collection = db.get_or_create_collection("quickstart")

vector_store = ChromaVectorStore(chroma_collection=chroma_collection)
storage_context = StorageContext.from_defaults(vector_store=vector_store)

if chroma_collection.count() == 0:
    print("Collection is empty. Indexing documents...")
    # load documents
    documents = SimpleDirectoryReader("./data").load_data()
    
    # create your index
    index = VectorStoreIndex.from_documents(
        documents, storage_context=storage_context
    )
    print("Indexing completed.")
else:
    print(f"Collection already contains {chroma_collection.count()} items. Loading existing index...")
    # load your index from stored vectors
    index = VectorStoreIndex.from_vector_store(
        vector_store, storage_context=storage_context
    )
    print("Existing index loaded.")

Collection already contains 22 items. Loading existing index...
Existing index loaded.

from llama_index.core import get_response_synthesizer
from llama_index.core.retrievers import VectorIndexRetriever
from llama_index.core.query_engine import RetrieverQueryEngine
from llama_index.core.postprocessor import SimilarityPostprocessor, KeywordNodePostprocessor

retriever = VectorIndexRetriever(index=index, similarity_top_k=2)

from llama_index.core import PromptTemplate
text_qa_template_str = (
    "We have provided context information below. \n"
    "---------------------\n"
    "{context_str}"
    "\n---------------------\n"
    "Given this information, please answer the question: {query_str}\n Write in Korean."
)

text_qa_template = PromptTemplate(text_qa_template_str)

response_synthesizer = get_response_synthesizer(text_qa_template=text_qa_template)

query_engine = RetrieverQueryEngine(
    retriever=retriever,
    response_synthesizer=response_synthesizer,
    # node_postprocessors=[SimilarityPostprocessor(similarity_cutoff=0.2), KeywordNodePostprocessor(required_keywords=["Combinator"], exclude_keywords=["Italy"])],
    node_postprocessors=[SimilarityPostprocessor(similarity_cutoff=0.2), KeywordNodePostprocessor(exclude_keywords=["Italy"])]
)

query = "What did the author do growing up?"

response = query_engine.query(query)

print(response.response)

저자는 성장하면서 주로 글쓰기와 프로그래밍에 집중했습니다. 대학에 가기 전, 그는 짧은 이야기를 쓰며 글쓰기 연습을 했지만, 그 이야기들은 플롯이 거의 없고 강한 감정을 가진 캐릭터들로만 구성되어 있어 좋지 않았다고 회상합니다. 프로그래밍에 대해서는 9학년 때 IBM 1401 컴퓨터를 사용하여 처음 시도했으며, 당시에는 펀치 카드로 프로그램을 작성해야 했습니다. 이후 그는 TRS-80이라는 마이크로컴퓨터를 구입하여 간단한 게임과 모델 로켓 비행 예측 프로그램, 그리고 아버지를 위해 사용한 워드 프로세서를 만들었습니다. 이러한 경험들은 그가 나중에 Y Combinator를 시작하고, 새로운 프로그래밍 언어인 Arc를 개발하는 데 큰 도움이 되었습니다.

for node in response.source_nodes:
    print(node)

Node ID: 9c1fd7de-910e-4212-a759-8c7857cec05d
Text: What I Worked On  February 2021  Before college the two main
things I worked on, outside of school, were writing and programming. I
didn't write essays. I wrote what beginning writers were supposed to
write then, and probably still are: short stories. My stories were
awful. They had hardly any plot, just characters with strong feelings,
which I ...
Score:  0.297

Node ID: 8ef64de0-d29b-48a2-a167-29722516b073
Text: Much to my surprise, the time I spent working on this stuff was
not wasted after all. After we started Y Combinator, I would often
encounter startups working on parts of this new architecture, and it
was very useful to have spent so much time thinking about it and even
trying to write some of it.  The subset I would build as an open
source proje...
Score:  0.291

response.metadata

{'9c1fd7de-910e-4212-a759-8c7857cec05d': {'file_path': 'c:\\llamaindex\\01-Learn\\data\\paul_graham_essay.txt',
  'file_name': 'paul_graham_essay.txt',
  'file_type': 'text/plain',
  'file_size': 75042,
  'creation_date': '2024-08-26',
  'last_modified_date': '2024-08-21'},
 '8ef64de0-d29b-48a2-a167-29722516b073': {'file_path': 'c:\\llamaindex\\01-Learn\\data\\paul_graham_essay.txt',
  'file_name': 'paul_graham_essay.txt',
  'file_type': 'text/plain',
  'file_size': 75042,
  'creation_date': '2024-08-26',
  'last_modified_date': '2024-08-21'}}