community: add Docling document loader

[vagenas]

Description

This adds a document loader for Docling doc parsing package from IBM that parses PDF, DOCX, PPTX, HTML, and other formats into a rich unified representation including document layout, tables etc., making them ready for generative AI workflows like RAG.

Some references:

• https://research.ibm.com/blog/docling-generative-AI
• https://www.redhat.com/en/blog/docling-missing-document-processing-companion-generative-ai
• Docling Technical Report

The introduced DoclingLoader enables users to:

• use various document types in their LLM applications with ease and speed, and
• leverage Docling's rich representation for advanced, document-native grounding.

Issue

No issue, but discussion initiated couple weeks ago: #27641

Dependencies

docling

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
langchain	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Nov 13, 2024 3:15pm

[vagenas]

Most CI findings fixed — just two remaining points, for which some feedback from the maintainers would be appreciated:

• Check make extended_tests #3.9: How should this best be handled given that Docling does not support Python 3.9?
• Check extended_tests #3.11: This fails on a transitive dependency conflict between docling and another another package from extended_testing_deps.txt. Do you have some suggestion as to how to address this (besides attempting relaxing the constraints for that transitive dependency)?

[vagenas]

Hey @ccurme, can you have a look at this Docling PR and my two questions above?

FYI I have updated the description with some further references for more context 😉

Happy to help in case of any questions!

[efriis]

Hey! This adds a community integration, which is no longer recommended. Would you be interested in publishing your own integration package, and contributing docs via this guide? https://python.langchain.com/docs/contributing/how_to/integrations/

[vagenas]

Hey! This adds a community integration, which is no longer recommended. Would you be interested in publishing your own integration package, and contributing docs via this guide? https://python.langchain.com/docs/contributing/how_to/integrations/

Hi @efriis, I see, then we can do it like that, yes.
Will come back with a docs PR once our package is ready! 🔜