langchain-docling

What it is and what it does

langchain-docling bridges Docling, a document conversion and understanding library, with LangChain's document loading and processing ecosystem. It provides a `DoclingLoader` that can ingest documents from URLs or local files and convert them into LangChain-compatible document objects suitable for downstream language model tasks.

You can run document conversion locally (with the `local` extra, which includes AI runtimes like PyTorch) or remotely via Docling Serve, Managed Docling, or IBM watsonx. The loader supports multiple export modes (doc chunks or Markdown), custom chunking strategies, metadata extraction, and backend-specific conversion options. It's designed for developers building RAG pipelines, document analysis workflows, or AI-powered search systems that need reliable document ingestion.

Use it for:

• Build a RAG pipeline that ingests PDFs and technical documents into LangChain for semantic search and question-answering.
• Convert scanned documents or images to structured text using OCR, then feed them into language models for analysis.
• Process multiple document formats (PDF, DOCX, images) in bulk and export as Markdown for downstream NLP tasks.
• Integrate document conversion into a LangChain agent that needs to extract and reason over document content.
• Use a remote Docling service to offload heavy document processing while keeping your application lightweight.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Integrates Docling document conversion with LangChain, enabling you to load and process documents (PDFs, images, etc.) into LangChain-compatible formats for use in language model pipelines.

Yes. Active maintenance, no vulnerabilities, low install friction, and permissive MIT license make this a safe choice. Install it if you're building a LangChain application that needs to ingest and process documents—either locally (with the `local` extra) or via a remote Docling endpoint. The base install is lightweight; add the `local` extra only if you need local conversion.

Install

langchain-docling on PyPI

pip install langchain-docling

uv add langchain-docling

poetry add langchain-docling

Installing langchain-docling

Before you install

Low friction install with a pure Python wheel. Active maintenance as of 2026-08-14 with no known vulnerabilities. Requires Python 3.10–3.14; the base install pulls only langchain-core and docling-slim, but local document conversion requires the `local` extra with additional dependencies.

License in practice

MIT license permits commercial and private use with minimal restrictions; you may use, modify, and distribute this package freely provided you include the license notice.

Quickstart

pip install langchain-docling

from langchain_docling import DoclingLoader

loader = DoclingLoader(file_path=["https://arxiv.org/pdf/2408.09869"])
docs = loader.load()

Requires Python 3.10 or later. Local document conversion requires the `local` extra (`pip install "langchain-docling[local]"`); without it, you must use a remote Docling Serve or Managed Docling endpoint.

Verify before relying

• Whether the base install (without `local` extra) can convert documents locally or requires a remote service endpoint.
• Performance characteristics and throughput limits when processing large document batches.
• Supported document formats beyond PDF (e.g., DOCX, images, HTML).

Package facts

Evidence: langchain_docling-3.0.0-py3-none-any.whl

Tags

More Artificial Intelligence packages