db.py

from langchain_community.document_loaders.csv_loader import CSVLoader
from langchain_text_splitters import RecursiveCharacterTextSplitter
import torch
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_chroma import Chroma


file_path="./paul_graham_essays.csv"
db_persist_directory = './docs/chroma/'


def load_data():
    loader = CSVLoader(
        file_path=file_path,
        csv_args={
            "delimiter": ",",
            "fieldnames": ['id', 'title', 'date', 'text'],
        },
        source_column='title',
        metadata_columns=['date'],
        content_columns=['text'],
    )

    data = loader.load()
    return data[1:]


def split_data(data, chunk_size, chunk_overlap):
    splitter = RecursiveCharacterTextSplitter(
        chunk_size=chunk_size,
        chunk_overlap=chunk_overlap,
        separators=['\n\n', '\n', '(?<=\. )', ' ', '']
    )

    return splitter.split_documents(data)


def generate_embeddings():
    device = "cuda" if torch.cuda.is_available() else "cpu"
    model_kwargs = {'device': device} 

    encode_kwargs = {'normalize_embeddings': False}

    return HuggingFaceEmbeddings(
        model_name=model_path,
        model_kwargs=model_kwargs,
        encode_kwargs=encode_kwargs,
    )


def get_db(
    chunk_size=1000,
    chunk_overlap=200,
    model_path = 'intfloat/multilingual-e5-large-instruct',
):
    data = load_data()
    
    splits = split_data(data, chunk_size, chunk_overlap)

    embedding = generate_embeddings()
    
    return Chroma.from_documents(
        documents=splits,
        embedding=embedding,
        persist_directory=persist_directory,
    )