# 1. プロジェクトフォルダ作成
mkdir rag-api
cd rag-api
# 2. 仮想環境の作成と有効化
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
# 3. 必要なライブラリをインストール
pip install fastapi uvicorn[standard] langchain langchain-openai pypdf faiss-cpu sentence-transformers python-dotenv

OPENAI_API_KEY="sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"

# rag_pipeline.py
import os
from dotenv import load_dotenv
from langchain_community.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import FAISS
from langchain_openai import OpenAIEmbeddings, ChatOpenAI
from langchain.chains import RetrievalQA
load_dotenv()
FAISS_INDEX_PATH = "faiss_index"
def create_vector_store(pdf_path: str):
    """PDFを読み込み、ベクトルストアを作成・保存する"""
print("Loading PDF...")
loader = PyPDFLoader(pdf_path)
documents = loader.load()
print("Splitting text into chunks...")
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
chunks = text_splitter.split_documents(documents)
print("Creating embeddings and vector store...")
embeddings = OpenAIEmbeddings()
vector_store = FAISS.from_documents(chunks, embeddings)
vector_store.save_local(FAISS_INDEX_PATH)
print(f"Vector store created and saved at {FAISS_INDEX_PATH}")
def create_qa_chain():
    """ベクトルストアを読み込み、QAチェーンを作成する"""
embeddings = OpenAIEmbeddings()
if not os.path.exists(FAISS_INDEX_PATH):
raise FileNotFoundError("Vector store not found.")
vector_store = FAISS.load_local(
FAISS_INDEX_PATH, 
embeddings, 
allow_dangerous_deserialization=True # FAISSのローカル読み込みに必要
)
llm = ChatOpenAI(model_name="gpt-4o")
# LangChain Expression Language (LCEL) を使った書き方
qa_chain = RetrievalQA.from_chain_type(
llm=llm,
chain_type="stuff",
retriever=vector_store.as_retriever()
)
return qa_chain

# main.py
import os
import shutil
from fastapi import FastAPI, UploadFile, File, HTTPException, Form
from fastapi.responses import JSONResponse
from rag_pipeline import create_vector_store, create_qa_chain
app = FastAPI(title="PDF RAG API")
UPLOAD_DIR = "uploaded_pdfs"
os.makedirs(UPLOAD_DIR, exist_ok=True)
# グローバル変数としてQAチェーンを保持
qa_chain = None
@app.post("/upload-pdf/")
async def upload_pdf(file: UploadFile = File(...)):
    """PDFをアップロードし、ベクトルストアを作成するエンドポイント"""
global qa_chain
if not file.filename.endswith(".pdf"):
raise HTTPException(status_code=400, detail="Only PDF files are allowed.")
file_path = os.path.join(UPLOAD_DIR, file.filename)
with open(file_path, "wb") as buffer:
shutil.copyfileobj(file.file, buffer)
try:
create_vector_store(file_path)
qa_chain = create_qa_chain() # チェーンを更新
return JSONResponse(status_code=200, content={"message": f"'{file.filename}' processed successfully."})
except Exception as e:
raise HTTPException(status_code=500, detail=f"Failed to process PDF: {str(e)}")
@app.post("/ask/")
async def ask_question(question: str = Form(...)):
    """アップロードされたPDFについて質問するエンドポイント"""
if qa_chain is None:
raise HTTPException(status_code=400, detail="Please upload a PDF first.")
try:
response = qa_chain.invoke({"query": question})
return JSONResponse(status_code=200, content={"answer": response.get("result")})
except Exception as e:
raise HTTPException(status_code=500, detail=f"Error: {str(e)}")

uvicorn main:app --reload