Semantic Notes for Obsidian

AI-powered note management for Obsidian: semantic search, AI Q&A with OpenRouter integration, and automatic note linking through embeddings.

Source code repository: https://github.com/EclipseAditya/semantic-notes-obsidian

Features

• Semantic Search: Find notes based on meaning, not just keywords
• AI Q&A: Ask questions about your notes and get AI-generated answers
• Automatic Embeddings: Notes are automatically indexed with embeddings
• Exact Match Detection: Combines semantic search with exact text matching for better results
• OpenRouter Integration: Leverage powerful language models through OpenRouter

Installation

1. Download the latest release from the GitHub repository
2. Extract the zip file into your Obsidian plugins folder: {vault}/.obsidian/plugins/
3. Enable the plugin in Obsidian's settings under "Community Plugins"

Setup

1. In the plugin settings, configure your OpenRouter API key (required for AI Q&A)
2. Select your preferred embedding model (UAE-Large-V1 recommended for best quality)
3. Choose your preferred language model for Q&A
4. Adjust other settings to your liking

The plugin will automatically start indexing your notes in the background. You can monitor the progress in the Obsidian notices.

Usage

Semantic Search

1. Use the command palette to open "Semantic Search" (Cmd/Ctrl+P and search for "Semantic Notes: Open Semantic Search")
2. Enter a natural language query (e.g., "What are the key benefits of meditation?")
3. View the search results ranked by relevance
4. Click on a result to open the corresponding note

AI Q&A

1. Use the command palette to open "Ask AI a Question" (Cmd/Ctrl+P and search for "Semantic Notes: Ask AI a Question")
2. Enter a question about your notes (e.g., "What have I written about project management?")
3. The AI will search your notes for relevant information and provide an answer
4. View the source notes that were used to generate the answer

How it Works

The plugin uses a local embedding model to convert your notes into vector representations that capture their semantic meaning. When you search or ask a question, your query is also converted to a vector and compared with your notes using cosine similarity.

For AI Q&A, the plugin:

1. Finds the most relevant notes using semantic search
2. Formats them as context for the language model
3. Sends the context and your question to OpenRouter
4. Returns the AI-generated answer along with the source notes

Images

1.png: Configuration options settings

2.png: Command options through command palette

3_example note.png: eg note created from rough note

3_example-refactor.png: refactored vault graph view

4.png: semantic view of similar notes

5.png: Smart semantic search

6.png: Example usage of Brain working(recreation of all vault notes with better quality and semantics)

7.png: Topic extraction feature usage

8.png: Ask ai about your notes interface(rag model)

Privacy

Your notes never leave your computer. The embedding process happens locally using Transformers.js. For AI Q&A, only the relevant snippets from your notes (needed to answer your question) are sent to OpenRouter.

Configuration

• Embedding Model: Choose between several models with different speed/quality tradeoffs
• LLM Model: Select which language model to use for Q&A
• Context Length: Number of notes to include as context for AI answers
• Re-ranking: Enable advanced re-ranking for better search results
• Persistent Storage: Save embeddings between sessions

License

MIT