index.py

from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.document_loaders import WebBaseLoader
from langchain_community.vectorstores import Chroma
from langchain_openai import OpenAIEmbeddings

def build_index(urls):
    embd = OpenAIEmbeddings()
    docs = [WebBaseLoader(url).load() for url in urls]
    docs_list = [item for sublist in docs for item in sublist]

    text_splitter = RecursiveCharacterTextSplitter.from_tiktoken_encoder(
        chunk_size=500, chunk_overlap=0
    )
    doc_splits = text_splitter.split_documents(docs_list)

    vectorstore = Chroma.from_documents(
        documents=doc_splits,
        collection_name="rag-chroma",
        embedding=embd,
    )
    return vectorstore.as_retriever()