knowledge2skills

分类: 文档与知识管理 | 上传者: yuancafe | 下载: 0 | 版本: v1.0（最新）

Transform any source material (PDF, MD, TXT, DOCX, EPUB, MOBI, AZW3) into structured, executable AI skill packages. Specializes in Knowledge Production using LightRAG for graph-enhanced precision. Supports multi-file batching and bundled graph databases. "knowledge2skills", "docs2skills", "ebooks2skills", "把这些文档变成技能", "知识生产流水线".

更新日志: Source: GitHub https://github.com/yuancafe/knowledge2skills

目录结构

当前层级: 根目录

📁 assets/
- 📄 example_asset.txt 865 B
📁 references/
- 📄 api_reference.md 2.3 KB
- 📄 extraction_guide.md 5.1 KB
📁 scripts/
- 📁 __pycache__/
  - 📄 extract_content.cpython-311.pyc 20.0 KB
  - 📄 extract_content.cpython-314.pyc 20.1 KB
  - 📄 generate_skill.cpython-311.pyc 21.9 KB
  - 📄 generate_visualization.cpython-311.pyc 19.8 KB
  - 📄 generate_visualization.cpython-314.pyc 18.7 KB
  - 📄 install_skill.cpython-311.pyc 5.3 KB
  - 📄 lightrag_graph.cpython-314.pyc 6.2 KB
- 📄 domain_detector.py 8.1 KB
- 📄 entity_deduplicator.py 8.2 KB
- 📄 entity_type_learner.py 13.1 KB
- 📄 evomap_register.py 1.7 KB
- 📄 example.py 591 B
- 📄 extract_content.py 14.2 KB
- 📄 generate_skill.py 18.2 KB
- 📄 generate_visualization.py 17.2 KB
- 📄 install_skill.py 3.3 KB
- 📄 knowledge2skills_pipeline.py 10.5 KB
- 📄 lightrag_graph.py 4.4 KB
- 📄 query_graph.py 2.2 KB
- 📄 semantic_engineering.py 3.5 KB
📄 .DS_Store 6.0 KB
📄 CHANGELOG.md 2.6 KB
📄 entity_type_templates.json 1.2 KB
📄 README.md 5.0 KB
📄 README_CN.md 4.9 KB
📄 SKILL.md 3.6 KB

SKILL.md

---
name: knowledge2skills
description: Transform any source material (PDF, MD, TXT, DOCX, EPUB, MOBI, AZW3) into structured, executable AI skill packages. Specializes in Knowledge Production using LightRAG for graph-enhanced precision. Supports multi-file batching and bundled graph databases. "knowledge2skills", "docs2skills", "ebooks2skills", "把这些文档变成技能", "知识生产流水线".
---

# knowledge2skills — Advanced Knowledge Production Pipeline


## Overview

`knowledge2skills` is an advanced evolution of knowledge production

, designed to convert unstructured information from multiple sources and formats into structured, executable AI skill packages. It prioritizes **Graph-Enhanced Precision**, **Multi-source Synthesis**, and **Self-contained Deployment**.

## Core Capabilities

1.  **Multi-format Extraction**: Seamlessly processes `.pdf`, `.md`, `.txt`, `.docx`, `.epub`, `.mobi`, `.azw3`, and `.json`.
2.  **Cross-lingual Synthesis**: Built-in support for mixed-language inputs (Chinese, English, Italian, etc.). It creates unified knowledge structures where concepts across different languages are mapped and linked.
3.  **Batch Synthesis**: Combines multiple files (e.g., a series of history books) into a single unified skill with source-aware sectioning.
3.  **Local-First Graph Construction**: Automatically detects local model backends (e.g., **Ollama**). If found, the pipeline offers to build the knowledge graph locally, eliminating API dependencies and costs.
4.  **Domain-Adaptive Graph RAG**: Integrates advanced logic from `rag-builder` to automatically detect document domains and apply optimized entity types (e.g., Medical, Legal, Business). It learns and updates these types over time.
5.  **Smart Entity Deduplication**: Automatically merges similar entities (e.g., "Dante" + "Dante Alighieri") during graph construction to ensure a clean and accurate knowledge base.
6.  **Interactive Graph Visualization**: Generates a self-contained, web-based interactive graph explorer bundled within the skill (`references/visualization/index.html`).
7.  **Auto-Configuration & Setup**: Scans the environment for LightRAG and necessary keys. If missing, it provides interactive installation and configuration prompts.
8.  **Bundled Graph DB**: If built, the LightRAG database is bundled into the skill package (`references/graph_db/`) for high-fidelity secondary retrieval.

## Workflow

1.  **Extract**: `scripts/extract_content.py <file1> <file2> ...` — Multi-file parallel extraction.
2.  **Graph (Optional)**: `scripts/lightrag_graph.py` — Builds a persistent knowledge graph.
3.  **Produce**: Generates a skill directory containing `SKILL.md`, `references/`, and `scripts/`.
4.  **Install**: Registers the skill in `~/.agents/skills/`.

### One-Command Full Pipeline
```bash
python3 scripts/knowledge2skills_pipeline.py book1.pdf book2.md --name history-expert --graph --install
```

## Sub-agent Guidance

*   **API Check**: Before running the graph enrichment, check for `OPENAI_API_KEY`. If missing, prompt the user.
*   **Graph Retrieval**: If a skill contains a `references/graph_db/` folder, prioritize using `LightRAG` queries over simple text search for complex "how" and "why" questions.
*   **Source Attribution**: Maintain the `[Source: filename]` prefixes in extracted sections to ensure traceability.

## Dependencies

- `pdfplumber`, `pypdf`: PDF parsing.
- `python-docx`: DOCX support.
- `lightrag-hku`: Knowledge graph building.

## Documentation

Detailed usage is available in **[api_reference.md](references/api_reference.md)** and **[extraction_guide.md](references/extraction_guide.md)**.

登录后下载/点赞/收藏 ❤ 3 | ★ 0

knowledge2skills

目录结构

SKILL.md

举报内容

提示