pdf2wiki converts native-text technical-book PDFs into clean, chapter-split, LLM-ready Markdown using a dual-pass MinerU pipeline plus fidelity-preserving post-processing.
The documentation is organized by what you want to do. Pick the entry point that matches your need — a high-level sense of the layout will tell you where to look.
Tutorials — learn by doing
Start here if pdf2wiki is new to you.
- Convert your first book — a full guided run, PDF to chapters.
How-to guides — get a specific task done
For when you already know the tool and need to accomplish something.
- Install pdf2wiki
- Convert a book
- Post-process and split into chapters
- Run a batch
- QA a conversion
- Configure a remote GPU host (experimental)
- Offload the hybrid pass to a server — no local GPU
- Convert in the cloud — no GPU, no MinerU (sends the PDF to a third-party cloud)
- Troubleshooting
Reference — find the facts
Precise, complete descriptions of the machinery.
- CLI reference — every command, flag, and default
- Configuration — the config file schema
- Pipeline stages — stages and their on-disk artifacts
- Phase 5 steps — the eight post-processors
- Output layout — what gets written, and the fidelity markers
Explanation — understand why
Background and design rationale.
- Architecture — arc42-style overview with C4 diagrams.
A standalone Claude Code plugin. It can consult a knowledge vault of the shape pdf2wiki builds, and review your code against one. This is the query side of the "LLM Wiki" pattern. It works with any such vault and does not require pdf2wiki. See the llm-wiki docs.
This doc set follows the Diátaxis framework: tutorials, how-to guides, reference, and explanation each serve a distinct need and are kept separate on purpose.