vdb_script/faiss_vdb_script.py

import os
from langchain.document_loaders import PyPDFLoader, DirectoryLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.embeddings.openai import OpenAIEmbeddings
from langchain_community.vectorstores import FAISS
from langchain.chains import RetrievalQA
from langchain.llms import OpenAI
from dotenv import load_dotenv

# Load environment variable for OpenAI key
load_dotenv()

OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
if not OPENAI_API_KEY:
    raise ValueError("Missing OPENAI_API_KEY in environment variables.")

# Extract Data from the PDFs
def load_pdf_file(data_path):
    loader = DirectoryLoader(data_path, glob="*.pdf", loader_cls=PyPDFLoader)
    documents = loader.load()
    return documents

# Split the data into chunks
def text_split(docs):
    splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=20)
    return splitter.split_documents(docs)

# Set up LLM and Embedding
llm = OpenAI(model_name="gpt-4o-mini", temperature=0.5, openai_api_key=OPENAI_API_KEY)
embeddings = OpenAIEmbeddings(openai_api_key=OPENAI_API_KEY)

# Load PDF, chunk it, embed it, and store in FAISS
pdf_docs = load_pdf_file("/kaggle/input/rag-test")  # Update this to your PDF folder
chunks = text_split(pdf_docs)

vectorstore = FAISS.from_documents(chunks, embeddings)
vectorstore.save_local("faiss_index_sysml")

# Load FAISS and create retriever QA chain
# new_vectorstore = FAISS.load_local("faiss_index_sysml", embeddings, allow_dangerous_deserialization=True)
# qa = RetrievalQA.from_chain_type(
#     llm=llm,
#     chain_type="stuff",
#     retriever=new_vectorstore.as_retriever()
# )

# # Run a sample query
# query = "What is SysML used for?"
# print("User:", query)
# print("Bot:", qa.run(query))