This project, voidful/hung-yi-lee-skill, creates an AI assistant that answers AI-related questions in the distinct teaching style of Professor Hung-yi Lee. It leverages a comprehensive knowledge graph derived from his lectures and provides explanations structured with his pedagogical approach, aiming to offer deeper understanding than simple retrieval-augmented generation (RAG).

How It Works

The core of the project is a 916-node knowledge graph built from 478 YouTube videos and associated transcripts, capturing Professor Lee's knowledge and thinking framework. Unlike static RAG, this system uses a structured knowledge graph with extracted real relationships between concepts, enabling better cross-topic discovery and traceability. Explanations follow a distilled teaching DNA: intuition first, then the black box, mechanism, potential pitfalls, and a recap, mimicking the professor's pedagogical style.

Quick Start & Requirements

• Installation: Clone the repository (git clone https://github.com/voidful/hung-yi-lee-skill.git), navigate into the directory (cd hung-yi-lee-skill), and install dependencies (pip install -r requirements.txt). The skill is then ready to be integrated with an AI coding assistant by making SKILL.md accessible.
• Prerequisites: Python environment.
• Dependencies: Listed in requirements.txt.
• Resources: Requires disk space for cloned repo and cached transcripts.
• Links: SKILL.md serves as the entry point.

Highlighted Details

• Knowledge Graph: Features 916 nodes, 3,664 edges, and 10 concept communities, with interactive visualization available. Key "God Nodes" include ML Fundamentals, Language Model, and Transformer.
• Distilled Teaching Structure: Adheres to a specific pedagogical DNA: Intuition First, Black Box, Unboxing Mechanism, Trap Warnings, and Recap.
• Enhanced RAG: Offers advantages over static RAG through its structured knowledge graph, extracted concept relations, automatic cross-topic discovery, persistent knowledge, and traceable edges (EXTRACTED/INFERRED).
• Data Provenance: All answers can be traced back to specific lecture segments, with clear distinctions between extracted and inferred relationships.

Maintenance & Community

The README does not explicitly detail maintenance contributors, community channels (like Discord/Slack), or a roadmap. It focuses on the technical implementation and data sources.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The MIT license is permissive and generally compatible with commercial use and closed-source linking, allowing broad adoption.

Limitations & Caveats

The project's effectiveness is tied to the completeness and accuracy of the distilled knowledge from Professor Lee's lectures. While it aims for a comprehensive teaching framework, it is inherently limited to the scope of the source material. The README does not detail performance benchmarks or specific integration requirements beyond making SKILL.md readable by an AI assistant.