langchain-oxidize-pdf

LangChain document loader backed by oxidize-pdf, a fast Rust-powered PDF engine with first-class RAG chunking.

0.1.0 (2026-04-24) — Requires oxidize-pdf>=0.4.3 (oxidize-pdf-core 2.5.5). First release. The sibling llama-index-readers-oxidize-pdf 0.1.0 shipped with shape-only tests that missed a quadratic accumulation bug in the underlying chunker; this loader ships from day one with the semantic regression suite (test_loader_disjoint.py) that guarantees the disjointness contract end-to-end.

Install

pip install langchain-oxidize-pdf

Usage

LangChain convention binds the file path to the loader instance and uses lazy_load() as the primary entry point; load() is inherited from BaseLoader as a convenience that materializes the iterator.

RAG chunks (default)

from langchain_oxidize_pdf import OxidizePdfLoader

loader = OxidizePdfLoader("paper.pdf")  # mode="rag" by default
documents = loader.load()

for doc in documents:
    print(doc.metadata["chunk_index"], doc.metadata["heading_context"])
    print(doc.page_content[:200])

Each Document carries:

One document per page

loader = OxidizePdfLoader("paper.pdf", mode="pages")
for doc in loader.lazy_load():
    print(doc.metadata["page_number"], len(doc.page_content))

Whole PDF as markdown

loader = OxidizePdfLoader("paper.pdf", mode="markdown")
[doc] = loader.load()
print(doc.page_content)

Adding caller metadata

loader = OxidizePdfLoader(
    "paper.pdf",
    extra_info={"source": "arxiv:2501.12345", "collection": "benchmarks"},
)

Keys in extra_info override base metadata (file_path, file_name, total_pages, pdf_version) if they collide — explicit caller intent.

Why oxidize-pdf

• Rust parser: fast on large PDFs, low memory footprint.
• Native RAG primitives: element-disjoint semantic chunking, element partitioning, heading-aware context — no post-processing needed. The disjointness contract (no chunk's text is a substring of another's; each source element appears in exactly one chunk) is enforced by regression tests in both this loader and the underlying bridge.
• CJK friendly: compact output for multibyte documents (see oxidize-pdf 2.5.4 subsetter fixes).
• Pure Python install: ships as a wheel for Linux/macOS/Windows via the oxidize-pdf package; no system dependencies.
• Real lazy loading: lazy_load() returns a generator, so large PDFs don't force every Document into memory upfront.

Source

Part of oxidize-pdf-integrations, the ecosystem of integrations around oxidize-pdf. The Rust core and Python bridge live in oxidize-python.

License

MIT