ingest.py

import os
from langchain_community.embeddings import SentenceTransformerEmbeddings
from langchain_community.document_loaders import UnstructuredFileLoader, DirectoryLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import Qdrant

embeddings = SentenceTransformerEmbeddings(model_name="NeuML/pubmedbert-base-embeddings")

loader = DirectoryLoader('data/', glob="**/*.pdf", show_progress=True, loader_cls=UnstructuredFileLoader)

documents = loader.load()

text_splitter = RecursiveCharacterTextSplitter(chunk_size=700, chunk_overlap=70)

texts = text_splitter.split_documents(documents)



url = "http://localhost:6333"
qdrant = Qdrant.from_documents(
    texts,
    embeddings,
    url=url,
    prefer_grpc=False,
    collection_name="vector_db"
)

print("Vector DB Successfully Created!")