YouTube Knowledge Graph for Obsidian

A plugin for Obsidian that creates a semantically connected knowledge graph from YouTube video content. It automatically extracts video transcripts, generates summaries through Claude, and builds connections between related discussions using semantic similarity and conceptual linking.

Features

• Easily add YouTube videos through ribbon icon or command palette
• Automatically transcribes videos using local Whisper server
• Generates concise summaries using Claude
• Creates semantic connections between related discussions using embeddings
• Extracts and links key concepts across videos
• Stores organized transcripts for reference
• Builds a rich knowledge network based on content similarity and shared concepts

Prerequisites

1. Python environment setup for Whisper server:

# Create and setup virtual environment
mkdir whisper_server
cd whisper_server
python3 -m venv whisper_env
source whisper_env/bin/activate

# Install dependencies
pip install fastapi uvicorn faster-whisper python-multipart yt-dlp ffmpeg-python

2. FFmpeg installation:

sudo apt install ffmpeg

3. Anthropic API key for Claude

Installation

1. Copy the main.js and manifest.json files to your Obsidian plugins folder: .obsidian/plugins/obsidian-youtube-knowledge-graph/

2. Enable the plugin in Obsidian Settings → Community Plugins

3. Configure your Anthropic API key in the plugin settings

4. Start the Whisper server:

cd whisper_server
source whisper_env/bin/activate
python whisper_server.py

Usage

1. Click the YouTube icon in the ribbon (left sidebar) or use command palette (Cmd/Ctrl + P) to open "Add YouTube Video"
2. Paste a YouTube video URL in the modal
3. The plugin will automatically:
   • Create a new note with the video title
   • Extract and transcribe the video using Whisper
   • Generate a summary using Claude
   • Extract key concepts from the content
   • Find and link to conceptually related discussions
   • Store transcript in the Transcripts folder
4. View connections:
   • Open Obsidian's graph view to see semantic and conceptual connections
   • Check "Conceptually Related Discussions" section in notes
   • Click through to related content based on shared concepts and similarity scores
   • Browse concepts to explore related discussions

Note Structure

Each processed note will have:

• Video title and link
• Link to full transcript
• Generated summary
• Semantic connections to related discussions with:
  • Shared concepts
  • Similarity scores
  • Relevant quotes showing the connection
  • Bi-directional links for graph visualization

Development

To build the plugin:

# Install dependencies
npm install

# Build
npm run build

Known Issues

• Requires local Whisper server to be running
• Large videos may take longer to process
• Requires stable internet connection for YouTube download and Claude API
• YouTube links must be standard format (e.g., https://www.youtube.com/watch?v=...)

Future Improvements

• Add support for multiple languages
• Implement batch processing for multiple videos
• Add configuration for Whisper model selection
• Add progress indicators for long transcriptions
• Support for different YouTube URL formats
• Improve similarity threshold tuning
• Add concept relationship visualization

Support

For issues or feature requests, please use the GitHub issues page.

License

MIT License