utils.py

import os
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.document_loaders import PyPDFLoader
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import Pinecone

INDEX_NAME = os.getenv("INDEX_NAME", "my-vector-db")
LC_Pinecone = Pinecone

def load_pdf_and_create_vectorstore(pdf_path: str, user_id: str):
    loader = PyPDFLoader(pdf_path)
    documents = loader.load()
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
    texts = text_splitter.split_documents(documents)
    embeddings = HuggingFaceEmbeddings(model_name='sentence-transformers/all-MiniLM-L6-v2')
    namespace = user_id or "default_namespace"
    vectorstore = LC_Pinecone.from_documents(
        texts,
        embeddings,
        index_name=INDEX_NAME,
        namespace=namespace
    )
    return vectorstore