Smart Vault Organizer

An Obsidian plugin powered by Rust/WebAssembly that uses local LLMs to intelligently organize your notes with semantic embeddings, AI-powered link suggestions, contextual chat, and smart formatting tools.

Development included the use of LLMs.

Features

Link Suggestions

• Semantic Embeddings - Generate embeddings for your vault using local LLMs (Ollama)
• LLM Reranking - AI ranks suggestions by relevance with explanations
• Sidebar Panel - View and manage suggestions for the current note
• Ignore System - Hide unwanted suggestions permanently

Smart Chat

• Context-Aware Q&A - Ask questions about your current note
• RAG Integration - Automatically retrieves relevant notes from your vault
• Agentic Tools - Execute actions: perform semantic comparisons, insert generated text directly into notes, summaries, and outlines

Smart Formatting

• Grammar Check - Get writing improvement suggestions
• Zettelkasten Structure - Analyze note structure and get recommendations
• Flashcard Generation - Auto-generate flashcards from note content
• Tag Suggestions - AI-recommended tags based on content

Smart Organization

• Folder Recommendations - AI suggests where to file your notes
• Confidence Scoring - Multiple candidates ranked by relevance

Handwritten Notes (Vision)

• Watch Folder - Auto-process images in Inbox/Handwritten
• Multi-Page Support - Stitch up to 5 pages together
• OCR with Vision Models - Extract text from handwritten notes
• Image Enhancement - Pre-processing for better accuracy

Prerequisites

1. Rust & Cargo (latest stable)

   curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
   

2. wasm-pack

   cargo install wasm-pack
   

3. Node.js (v16+) - nodejs.org

4. Ollama - ollama.ai

   ollama pull nomic-embed-text
   ollama pull ministral-3:3b
   

Installation

Build from Source

### Install from Community Plugins

1. Open **Settings** > **Community Plugins**
2. Turn off **Safe Mode**
3. Click **Browse** and search for **Smart Vault Organizer**
4. Click **Install** -> **Enable**

### Build from Source

```bash
cd /path/to/your/vault/.obsidian/plugins
git clone https://github.com/zkdavis/obsidian-smart-vault.git smart-vault-organizer
cd smart-vault-organizer
npm install
npm run build

Install to Vault

Copy files

mkdir -p /path/to/vault/.obsidian/plugins/smart-vault-organizer
cp main.js manifest.json styles.css /path/to/vault/.obsidian/plugins/smart-vault-organizer/
cp -r pkg /path/to/vault/.obsidian/plugins/smart-vault-organizer/

Symlink (for development)

# macOS/Linux
ln -s /path/to/obsidian_app /path/to/vault/.obsidian/plugins/smart-vault-organizer

# Windows (PowerShell as Admin)
New-Item -ItemType SymbolicLink -Path "C:\vault\.obsidian\plugins\smart-vault-organizer" -Target "C:\path\to\obsidian_app"

Enable in Obsidian

1. Settings → Community Plugins → Disable Safe Mode
2. Reload plugins
3. Enable "Smart Vault Organizer"

Quick Start

1. Configure - Settings → Smart Vault Organizer

   • Verify Ollama endpoint (http://localhost:11434)
   • Set embedding model (nomic-embed-text)
   • Set LLM model (ministral-3:3b)

2. Scan Vault - Command Palette → Smart Vault: Scan vault for embeddings

3. Get Suggestions - Open any note and click the brain icon in the ribbon

Configuration

Setting	Description	Default
Ollama Endpoint	URL of Ollama server	http://localhost:11434
Embedding Model	Model for embeddings	nomic-embed-text
LLM Model	Model for reranking	ministral-3:3b
Chat Model	Model for Smart Chat	ministral-3:3b
Formatting Model	Model for Smart Formatting	ministral-3:3b
Organization Model	Model for Smart Organization	ministral-3:3b
Vision Model	Model for handwritten notes	ministral-3:3b
Similarity Threshold	Minimum score (0-1)	0.7
Max Suggestions	Suggestions to show	5
Max Content Length	Characters for embeddings	2000
LLM Reranking	Enable AI ranking	true
LLM Candidate Count	Candidates sent to LLM	5
LLM Temperature	Reranking randomness	0.3
Chat Temperature	Chat creativity	0.7
LLM Timeout	Request timeout (ms)	30000
LLM Concurrency	Parallel requests during scan	3
Thinking Mode	Chain-of-thought for qwen3/deepseek	false
Keyword Extraction	Extract keywords for matching	true
Handwritten Inbox	Watch folder for images	Inbox/Handwritten
Transcript Folder	Output folder for transcripts	Inbox

Documentation

See the docs/ folder for detailed guides:

• Link Suggestions - Embedding and reranking system
• Smart Chat - RAG-based Q&A
• Smart Formatting - Grammar, flashcards, tags
• Smart Organization - Folder recommendations
• Handwritten Notes - Vision OCR setup
• Configuration - All settings explained
• Architecture - Technical deep-dive
• Troubleshooting - Common issues

Development

Project Structure

src/
├── main.ts                 # Entry point
├── plugin/
│   ├── SmartVaultPlugin.ts # Core plugin
│   ├── cache/              # Cache management
│   └── scanning/           # Vault scanning, file processing
├── ui/
│   ├── LinkSuggestionView.ts # Main sidebar panel
│   └── tabs/               # Smart Chat, Formatting, Organization tabs
├── settings/               # Settings UI and types
├── llm/                    # LLM reranking service
├── suggest/                # Inline autocomplete
├── lib.rs                  # Rust: similarity calculations
├── llm.rs                  # Rust: LLM integration
└── links.rs                # Rust: link detection

Commands

npm run build         # Full build (WASM + TypeScript)
npm run build:wasm    # Build Rust/WASM only
npm run build:plugin  # Build TypeScript only
npm run dev           # Watch mode
npx tsc --noEmit      # Type check

Recommended Models

Model	Use Case	Size
nomic-embed-text	Embeddings (fast, quality)	768-dim
mxbai-embed-large	Embeddings (higher quality)	1024-dim
ministral-3:3b	All tasks (default)	3B

Troubleshooting

Plugin won't load

• Check Developer Console (Ctrl+Shift+I) for errors
• Verify all files exist: main.js, manifest.json, styles.css, pkg/

"Failed to initialize WASM module"

• Ensure pkg/ directory contains .wasm file
• Run npm run build again

No suggestions appearing

• Run vault scan first
• Check Ollama is running: ollama list
• Lower similarity threshold in settings

LLM timeout errors

• Increase timeout in settings (up to 120s)
• Use a faster model
• Reduce LLM candidate count

Contributing

Contributions welcome! Please ensure:

• Rust code passes cargo clippy
• TypeScript follows existing style
• Test with a real Obsidian vault

Support

If you find this plugin useful, please consider supporting development:

• GitHub Sponsors
• Ko-fi
• Buy Me a Coffee

License

MIT License - See LICENSE file

Acknowledgments

• Built with wasm-bindgen
• Embeddings via Ollama
• Inspired by the Obsidian community