Skip to content
Projects

YangYuHou/StudyAnalysis-Skills

View on GitHub

YangYuHou/StudyAnalysis-Skills has 664 GitHub stars and 51 forks. GitHot measured 0 stars in 24 hours. Last checked 2026-11-15 23:22 UTC. Last push 2026-03-16.

How these counts are measured

Measured star growth

in 24h
0
in 7 days
Update delayed
in 15 days
Update delayed
in 1 month
Update delayed

Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.

About

深度解析链接、文档或代码,生成“全能导师级”的教学笔记

Developer Tools

README_EN.md

Original

📚 Knowledge Absorber

License: MIT Python: 3.8+ Version: 4.1.0

Universal AI Skill Module | Compatible with Trae, Claude, Gemini, VS Code Copilot, etc.

🇺🇸 English | 🇨🇳 简体中文


Knowledge Absorber is an independent "External Brain" module. It empowers AI agents with the ability to deeply read long documents, analyze complex content, and generate structured "Knowledge Crystals" (Markdown + HTML).


🚀 What's New

v4.1.0 Latest Update

  • 🎓 Mentor Mode: Brand new interactive learning experience

    • Intelligent Dialogue Assistant: Click the "Open Mentor Mode" button in the generated HTML page to chat with an AI mentor in real-time
    • Context-Aware: The mentor fully understands the current knowledge card content and can answer questions and deepen understanding specifically
    • Local Relay Service: Start the local relay service via python scripts/mentor_relay.py (default port 8760)
    • Multi-Model Support: Compatible with OpenAI, Claude, Gemini, and any OpenAI-compatible API
    • Sidebar Design: Elegant sidebar interface that doesn't obstruct the main content, can be opened and closed at any time
    • Conversation History: Maintains complete conversation context, supports continuous in-depth questioning
  • 💡 Deep Linking Analysis: Added multi-source input conflict comparison feature, automatically compares conflicts from multiple sources, strengthening AI's fact-checking capabilities

v4.0.0 Previous Updates

  • ⚡ High-Speed Concurrent Engine: Introduced ThreadPoolExecutor and threading.Lock for simultaneous scraping of multiple links/files, improving processing efficiency by 300%+.
  • 🎨 Visual Progress System: Integrated Rich library for high-contrast color tracking, providing a clear task board and real-time progress feedback in the terminal.
  • ⚓ Truth Anchoring Protocol: Added 【💡 Deep Linking】 analysis to automatically identify factual conflicts or complementarities across multiple sources.
  • 🔍 Deep Protocol Alignment: Mandatory integration of the "Seven Holographic Lenses" protocol, ensuring outputs include "Mind Maps" and "Pitfall Guides".
  • 🛡️ Enhanced Robustness: Optimized SSL verification for offline testing and added auto-fix for character encoding issues (Mojibake) on major platforms.
  • ⚛️ Deep Fission Module: Added atomic-level contradiction analysis and version archeology module to reveal counter-intuitive conclusions (Style: .fission-section).
  • 🔍 Strict Search Filter: Upgraded HTML interaction; search box now strictly hides non-matching content blocks instead of just highlighting, providing a focused reading experience.
  • 🛡️ Mermaid Safety Protocol: Built-in syntax auto-correction mechanism that forcibly escapes special characters to prevent diagram rendering crashes.

📂 Portability Guide

This module is designed for "Folder-Level Plug & Play". AI assistants typically scan specific configuration folders in the project root. To let other AIs (like Claude or Gemini) recognize this skill, you simply need to rename the parent directory.

📂 Directory Structure Adaptation

Assuming your skills folder is located at the project root:

  1. For Trae (Default):

    Project_Root/
    └── .trae/              <-- Keep original name
        └── skills/
            └── knowledge-absorber/
    
  2. For Claude Projects:

    • Rename .trae to .claude.
    Project_Root/
    └── .claude/            <-- Rename to .claude
        └── skills/
            └── knowledge-absorber/
    
  3. For Gemini Advanced / AI Studio:

    • Rename .trae to .gemini.
    Project_Root/
    └── .gemini/            <-- Rename to .gemini
        └── skills/
            └── knowledge-absorber/
    
  4. For VS Code (Copilot/Cline):

    • Rename .trae to .vscode.
    Project_Root/
    └── .vscode/            <-- Rename to .vscode
        └── skills/
            └── knowledge-absorber/
    

💡 Core Principle: AI agents usually have permission to read hidden folders (starting with .). As long as the path is correct and you explicitly instruct the AI to "use this skill", it will work.


🛠️ Installation & Usage

Step 1: Environment Preparation

Ensure Python 3.8+ is installed on your machine. Run the following command in the knowledge-absorber directory:

pip install -r requirements.txt

Step 2: When to Activate

Do not use this skill for simple Google searches. Summon it for "High Cognitive Load" scenarios:

  1. Heavy Lifting: When facing hundreds of pages of PDFs, technical framework docs (e.g., BMad, React Source), or ancient texts.
  2. Knowledge Crystals Needed: When you need more than a simple summary—you need a beautifully formatted, archivable HTML card.
  3. Cross-Source Analysis: When you need to scrape and compare multiple links (e.g., Zhihu + Blog + Official Docs) for deep Truth Anchoring.

Step 3: Workflow

You don't need to manually run complex command-line instructions. Just use Natural Language in the chat, and the AI will proxy the execution:

  • User Instruction Examples:

    "Help me deeply analyze this link: https://docs.bmad-method.org/" "Read manual.pdf and apply the [Mechanistic Lens] to generate knowledge cards."

  • AI Execution Logic:

    1. Ingestion: Automatically calls scripts/content_ingester.py to scrape and clean content concurrently.
    2. Lens Analysis: Applies deep reasoning based on protocols defined in SKILL.md (e.g., Mechanistic Lens, Evolution Lens).
    3. Delivery: Automatically generates .md (Deep Notes) and .html (Visual Cards).

📦 Outputs

This skill automatically generates files in two formats (located in the data/ directory):

  1. Markdown Deep Notes (.md):
    • Includes metadata, concept icebreaking, deep deconstruction, mind maps, and pitfall guides.
    • Supports "Mixed Script Protocol" (Traditional/Simplified) or "Tech Stack Templates".
  2. HTML Visual Cards (.html):
    • Beautiful formatting, perfect for sharing or archiving in a knowledge base.
    • Supports Dark/Light mode adaptation, with perfected code highlighting and Mermaid diagram display.

🎓 Mentor Mode Explained

What is Mentor Mode?

Mentor Mode is Knowledge Absorber's interactive learning feature that allows you to chat with an AI mentor in real-time while browsing generated knowledge cards, deepening understanding and answering questions.

Core Features

  1. 📚 Context-Aware

    • The mentor fully understands all content in the current knowledge card
    • Can provide in-depth explanations for specific sections and concepts
    • Automatically correlates related knowledge points
  2. 💬 Intelligent Dialogue

    • Supports continuous multi-turn conversations
    • Maintains complete conversation history
    • Allows follow-up questions, deep dives, and examples
  3. 🎨 Elegant Interface

    • Sidebar design that doesn't obstruct the main content
    • Supports dark/light themes
    • Responsive layout, adapts to various screen sizes
  4. 🔌 Flexible Configuration

    • Supports any OpenAI-compatible API
    • Configurable model, temperature, max tokens, and other parameters
    • Runs locally, data security guaranteed

Usage Steps

1. Start the Relay Service

Run in the knowledge-absorber directory:

python scripts/mentor_relay.py

You will see:

Mentor relay listening on http://127.0.0.1:8760

💡 Tip: The relay service is a local relay server used to pass messages between the browser and AI models. It does not store your conversation content.

2. Open the Knowledge Card

Open the generated knowledge_card.interactive.html file in your browser.

3. Configure AI Connection

Click the "Open Mentor Mode" button in the upper right corner of the page, then in the sidebar:

  1. Select API type (OpenAI / Claude / Gemini / Custom)
  2. Enter API Key
  3. Select model (e.g., gpt-4, claude-3-5-sonnet, etc.)
  4. Click "Test Connection" to confirm the configuration is correct

4. Start Chatting

After successful configuration, you can ask questions in the input box! For example:

  • "Please explain the core mechanism section in detail"
  • "What's the difference between this concept and XXX?"
  • "Can you give a practical application example?"
  • "I encountered XXX problem in practice, what should I do?"

Typical Use Cases

  1. Deepen Understanding

    • Don't quite understand a concept, ask the mentor to explain in simpler terms
    • Need more examples to deepen impression
  2. Extended Learning

    • Ask about related knowledge points
    • Understand deeper principles
  3. Practical Guidance

    • Ask how to apply in actual projects
    • Seek advice on specific problems encountered
  4. Comparative Analysis

    • Compare pros and cons of different technical solutions
    • Understand historical evolution and future trends

Configuration Instructions

Supported API Types

  • OpenAI: GPT-4, GPT-3.5, etc.
  • Claude: Claude 3.5 Sonnet, Claude 3 Opus, etc.
  • Gemini: Gemini Pro, Gemini Ultra, etc.
  • Custom: Any service compatible with OpenAI API format

Advanced Configuration

In the sidebar's "Advanced Settings", you can adjust:

  • Temperature: Controls creativity of responses (0-2, default 0.7)
  • Max Tokens: Maximum length of single response (default 2000)
  • System Prompt: Customize the mentor's role and style

Important Notes

  1. Relay Service Must Be Running

    • If the relay service is not started, Mentor Mode will not work
    • Ensure port 8760 is not occupied
  2. API Key Security

    • API Key is only stored in browser local storage (localStorage)
    • Not uploaded to any server
    • Recommend using API Keys with quota limits
  3. Network Connection

    • Need to be able to access the corresponding AI service API
    • If using domestic networks, may need to configure a proxy
  4. Reading Mode Only

    • If just reading knowledge cards, no need to start relay service
    • Search, theme switching, and other features do not depend on relay

Troubleshooting

Issue: Clicking "Open Mentor Mode" has no response

  • Check if relay service is running
  • Check browser console for error messages

Issue: Test connection failed

  • Check if API Key is correct
  • Check if network connection is normal
  • Check if API endpoint URL is correct

Issue: Dialogue response is very slow

  • May be high model load, wait a moment
  • Can try switching to other models
  • Check network latency

🤖 Skill Protocol

The core logic is defined in the SKILL.md file. If you want to modify the AI's way of thinking (e.g., changing the depth of analysis or output style), please edit SKILL.md directly.


Maintainer: Little Code Sauce Version: v4.1.0 (Mentor Mode Edition)