Skip to content

Lumberjack documentation

中文

Lumberjack turns Markdown, HTML, and DOCX into retrieval-ready chunks while preserving document structure and heading context. Start with the installation guide, then split your first document in the quickstart.

What makes a chunk useful for retrieval?

Lumberjack separates a chunk's body from its heading path, tracks source provenance, and respects document blocks while applying a configurable token budget. One format-neutral DocTree supports all built-in parser formats and all splitter strategies.

I want to… Start here
Choose a strategy or counting mode Splitting and counting
Understand output fields and source locations Chunks and provenance
Configure code/table behavior Configuration
Extend the pipeline Custom components
Call the CLI or Web API Reference

The project is still building its benchmark suite. This site does not make unverified performance or quality claims.