minerupress

分类: 工具与效率 | 上传者: aronnaxlin | 下载: 0 | 版本: v1.0（最新）

当 AI 代理需要将上传的 MinerU 结果、单独安装的本地 MinerU CLI 或 MinerU 官方 API 结果转换为可发布的 MkDocs Material 图书网站（具有章节分割、图像过滤、CJK 间距、严格验证和可选的 Cloudflare Pages 部署）时使用。触发涉及 MineruPress、minerupress、MinerU content_list.json、book.yml、“minerupress 导出”、“minerupress 获取”、“minerupress 标题”、“minerupress 指纹”、mineru-export、mineru-fetch、生成的文档/章节或 PDF-to-MkDocs 图书出版自动化的任务。

更新日志: Source: GitHub https://github.com/aronnaxlin/minerupress

目录结构

当前层级: tree/main/skills/minerupress/

📁 agents/
- 📄 openai.yaml 223 B
📁 references/
- 📄 book-yml.md 2.4 KB
📄 SKILL.md 7.3 KB

SKILL.md

---
name: minerupress
description: Use when an AI agent needs to turn uploaded MinerU results, a separately installed local MinerU CLI, or MinerU official API results into a publishable MkDocs Material book site with chapter splitting, image filtering, CJK spacing, strict validation, and optional Cloudflare Pages deployment. Trigger for tasks involving MineruPress, minerupress, MinerU content_list.json, book.yml, `minerupress export`, `minerupress fetch`, `minerupress headings`, `minerupress fingerprint`, mineru-export, mineru-fetch, generated docs/chapters, or PDF-to-MkDocs book publishing automation.
---

# MineruPress

Use this skill to configure, run, verify, and troubleshoot the MineruPress publishing pipeline for MinerU -> MkDocs Material book sites.

## Install The Toolchain

If the current machine does not already have the project, install it first:

```bash
git clone https://github.com/aronnaxlin/minerupress.git
cd minerupress
pip install -e ".[all]"
```

For a new book workspace, run `minerupress init <directory>` outside the
toolchain repo or use the bundled template as the starting point for `book.yml`,
`mkdocs.yml`, `.env.example`, and `Makefile`. Keep real PDFs, MinerU output,
`.env`, generated `docs/`, and generated `site/` out of the toolchain
repository unless the user explicitly asks to commit a book project.

## Preferred Workflow

Default to a two-stage, isolated workflow when the user starts from a raw PDF:

1. Create an isolated book workspace rather than reusing the toolchain root.
2. Copy the source PDF into that workspace under `resources/pdfs/`.
3. Set `source: official_api` and run a first pass with `minerupress fetch` to obtain MinerU output.
4. Inspect `*_content_list.json` to confirm the real chapter boundary shape.
5. Refine `book.yml` chapter boundaries and rerun `minerupress export`.
6. Verify with `mkdocs build --strict` and optionally fingerprints.

This is the safest default because:

- automatic PDF splitting writes chunk files next to the source PDF
- cloud-drive or protected paths may fail with `PermissionError`
- table-of-contents pages often need one refinement pass before chapter boundaries are stable
- generated `docs/`, `site/`, `reports/`, and `resources/mineru/` stay isolated per book

## Workflow

1. Inspect the book workspace:
   - Read `book.yml`, `mkdocs.yml`, and `AGENTS.md` if present.
   - Check `resources/mineru/` for one or more MinerU output directories containing `*_content_list.json`.
   - Treat `docs/chapters/`, `docs/images/`, `site/`, `reports/`, and `.env` as generated or local-only unless the user explicitly asks otherwise.

2. If the user starts from a raw PDF instead of existing MinerU output:
   - Create or use an isolated workspace for that specific book.
   - Copy the PDF into `resources/pdfs/` inside the workspace instead of referencing a cloud-drive or protected external path directly.
   - Choose a source mode:
     - `official_api` is the preferred default for new workspaces.
     - `local_toolchain` is valid only when the user already installed MinerU separately.
     - `uploaded_result` is only for pre-existing parsed output under `resources/mineru/`.
   - Write a minimal `api:` block and a temporary placeholder chapter when using `official_api`.
   - Run `minerupress fetch book.yml` first to fetch MinerU output and get an initial export.
   - Run `minerupress headings resources/mineru --volume-uid <uid> --format yaml --body-only` to generate a chapter YAML draft.

3. Configure `book.yml`:
   - Set `source` explicitly for new configs: `official_api`, `local_toolchain`, or `uploaded_result`.
   - Use a top-level `volume_uid` for a logical book/PDF. Split outputs such as `javaweb_p1` and `javaweb_p2` can share `volume_uid: javaweb`.
   - Prefer chapter `title` plus `slug`; omit `start_pattern` unless the generated boundary matcher is ambiguous.
   - Add `aliases` or `start_patterns` only when MinerU headings differ from the canonical title.
   - Keep `allow_missing_boundaries: false` for production runs.
   - After the first fetch, inspect `*_content_list.json` and replace placeholder chapters with the real chapter list.
   - If the table of contents page causes false matches, prefer a display-only `title` plus an exact `start_pattern` such as `^第\\s*1\\s*章$`.
   - Do not make MineruPress install MinerU as an internal dependency. If the user wants `local_toolchain`, point them to MinerU's official install docs and treat the CLI as an external prerequisite.

4. Run export:

```bash
minerupress export book.yml
```

For cloud API upload and export:

```bash
minerupress fetch book.yml
```

Legacy wrappers `minerupress-export`, `minerupress-fetch`, `minerupress-headings`, `mineru-export`, and `mineru-fetch` are also available for existing projects.

5. Verify:

```bash
mkdocs build --strict
```

If Markdown fingerprints are part of the project workflow:

```bash
minerupress fingerprint --docs-dir docs --out reports/fingerprints.json
```

## Chapter Boundary Guidance

The exporter can infer boundary patterns from chapter titles. Use titles like:

- `第10章 JavaScript`
- `附录A 部分习题的解答`
- `Chapter 3 Arrays`
- `项目二 尚硅谷书城`
- `10.1 JavaScript 简介`

Use `aliases` for alternate visible headings:

```yaml
chapters:
  - slug: appendix-a
    title: 附录A 部分习题的解答
    aliases:
      - Appendix A
```

Use `start_patterns` for precise regex alternatives:

```yaml
chapters:
  - slug: unit-01
    title: 第一单元 Web 基础
    start_patterns:
      - "^Unit\\s*1\\b"
      - "^第\\s*一\\s*单元"
```

Only use `--allow-missing-boundaries` while diagnosing bad MinerU output. A successful production run should find every configured chapter boundary.

If TOC lines such as `第1章 引论 …… 3` are matched before the real body heading,
switch to exact body-heading regexes:

```yaml
chapters:
  - slug: ch01-introduction
    title: 引论
    start_pattern: "^第\\s*1\\s*章$"
```

This avoids the generated title matcher from locking onto the table of contents.

Use the headings helper before hand-writing many boundaries:

```bash
minerupress headings resources/mineru --volume-uid javaweb --format report
minerupress headings resources/mineru --volume-uid javaweb --format yaml --body-only
```

The report marks TOC-looking candidates as `toc?` and body-looking candidates as `body`.

## Generated Outputs

Each export rebuilds:

- `docs/chapters/`
- `docs/images/`

Do not hand-edit generated chapter Markdown as a long-term fix. Instead, update `book.yml`, source MinerU output, or add a plugin.

## Code And HTML Prose

MinerU code items may store their body in `code_body` rather than `text`.
The exporter preserves already-fenced code blocks from MinerU and falls back
to wrapping unfenced code. If a chapter appears to drop examples after phrases
like "示例代码如下", inspect the source `*_content_list.json` for `type: code`
items before editing generated Markdown.

Literal HTML/XML tags in normal prose and captions are escaped during export
so textbooks can discuss tags such as `<span>` without MkDocs rendering them
as real HTML. Raw `table_body` HTML is preserved as table markup.

## References

Read `references/book-yml.md` when you need a compact configuration reference or examples.
For the end-to-end isolated-book workflow, also read the project docs page
`docs/guide/workflow-run-a-book.md` in this repository.

登录后下载/点赞/收藏 ❤ 14 | ★ 0

minerupress

目录结构

SKILL.md

举报内容

提示