Pixeltable vs LangChain

import pixeltable as pxt
from pixeltable.functions.document import document_splitter
from pixeltable.functions.huggingface import clip, sentence_transformer

# Multimodal knowledge base
kb = pxt.create_table('app.knowledge', {
    'document': pxt.Document,
    'image': pxt.Image,
    'title': pxt.String,
})

# Document RAG: chunk + embed (incremental)
chunks = pxt.create_view(
    'app.chunks', kb,
    iterator=document_splitter(
        document=kb.document,
        separators='sentence', limit=512, overlap=50
    )
)
chunks.add_embedding_index(
    'text',
    string_embed=sentence_transformer.using(
        model_id='sentence-transformers/all-MiniLM-L6-v2'
    )
)

# Image search: CLIP embedding index
kb.add_embedding_index(
    'image',
    embedding=clip.using(model_id='openai/clip-vit-base-patch32')
)

@pxt.query
def search_docs(question: str, n: int = 5):
    return chunks.select(chunks.text, chunks.title).order_by(
        chunks.text.similarity(string=question), asc=False
    ).limit(n)

@pxt.query
def search_images(query_text: str, n: int = 5):
    sim = kb.image.similarity(string=query_text)
    return kb.order_by(sim, asc=False).limit(n).select(
        kb.title, kb.image, sim
    )

# Insert assets anytime — chunking and indexes stay in sync

from langchain_community.document_loaders import DirectoryLoader
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_openai import OpenAIEmbeddings, ChatOpenAI
from langchain_community.vectorstores import Chroma
from langchain.chains import RetrievalQA

loader = DirectoryLoader("./documents", glob="**/*.pdf")
docs = loader.load()

splitter = RecursiveCharacterTextSplitter(
    chunk_size=512, chunk_overlap=50
)
chunks = splitter.split_documents(docs)

vectorstore = Chroma.from_documents(
    chunks, OpenAIEmbeddings(model="text-embedding-3-small")
)

qa = RetrievalQA.from_chain_type(
    llm=ChatOpenAI(model="gpt-4o-mini"),
    retriever=vectorstore.as_retriever(search_kwargs={"k": 5}),
)
answer = qa.invoke({"query": "What is our refund policy?"})

# Images/video: separate loaders, embedding models, and vector stores