Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.
Python ai4sciencedocument-analysisdocxextract-datalayout-analysisocrparserpdfpdf-converterpdf-extractor-llmpdf-extractor-pretrainpdf-extractor-ragpdf-parserpptxpythonxlsx
import pytest
from mineru.doclib.locators import block_char_ref, block_ref, locator_for_block, page_ref, parse_content_cursor
def test_doclib_locator_helpers_use_public_one_based_numbers() -> None:
assert locator_for_block(1, 3) == "page:1/block:3"
assert page_ref("ab12cd3", "standard", 1) == "doc:ab12cd3/tier:standard/page:1"
assert block_ref("ab12cd3", "standard", 1, 3) == "doc:ab12cd3/tier:standard/page:1/block:3"
assert block_char_ref("ab12cd3", "standard", 1, 3, 0) == "doc:ab12cd3/tier:standard/page:1/block:3/char:0"
def test_parse_content_cursor_supports_page_block_and_char_refs() -> None:
page = parse_content_cursor("doc:ab12cd3/tier:standard/page:5")
block = parse_content_cursor("doc:ab12cd3/tier:standard/page:5/block:18")
char = parse_content_cursor("doc:ab12cd3/tier:standard/page:5/block:18/char:32784")
assert page.page_no == 5
assert page.block_no is None
assert block.block_no == 18
assert block.char_offset is None
assert char.char_offset == 32784
def test_content_cursor_rejects_zero_based_public_numbers() -> None:
with pytest.raises(ValueError):
locator_for_block(0, 1)
with pytest.raises(ValueError):
parse_content_cursor("doc:ab12cd3/tier:standard/page:0")
with pytest.raises(ValueError):
parse_content_cursor("doc:ab12cd3/tier:standard/page:1/block:0")