script/extraction.py

from langchain_community.document_loaders import PyPDFLoader
from langchain_core.documents import Document
from script.errors import IngestionError

MIN_TEXT_LENGTH = 500  # heuristic, tune later


def extract_text_native(pdf_path: str) -> list[Document]:
    loader = PyPDFLoader(pdf_path)
    documents = loader.load()

    if not documents:
        raise IngestionError(
            code="PDF_PARSE_FAILED",
            message="No pages extracted from PDF"
        )

    full_text = "\n".join(d.page_content for d in documents).strip()

    if len(full_text) < MIN_TEXT_LENGTH:
        raise IngestionError(
            code="INSUFFICIENT_TEXT",
            message="PDF likely scanned or image-based"
        )

    return documents