Analyze and visualize your obsidian vault and images based on semantic similarity

How to use

• pip install -r requirements.txt
• drop your obsidian notes inside obsidian/
• drop your images inside images/
• run python main.py
• run python plot.py

Requirements

• 8GB+ VRAM
• CUDA capable GPU
• Python
• ~12 GB storage

Files

• main.py
  • generates notes.pkl which contains all notes from obsidian/ and their embeddings using Qwen3-VL-Embedding-2B
  • generates images.pkl which contains all images from images/ and their embeddings using Qwen3-VL-Embedding-2B
• plot.py creates interactive 2D/3D semantic similarity point maps for notes, images, and the combined notes + images space, also displays word count and other info
• query.py searches top most related notes based on your query
• gap.py detects similar notes that are not linked [[My Note]] <--> [[Related Note]]
• config.py what embedding model to use etc