obedding

Semantic search for your Obsidian vault using local embeddings

          │                    │
          ▼                    ▼
   ┌─────────────┐      ┌──────────────┐
   │  Obsidian   │      │   obedding   │
   │    Vault    │─────▶│              │
   │             │      │  ● Index     │
   │  *.md files │      │  ● Search    │
   │             │      │  ● Stats     │
   └─────────────┘      └──────┬───────┘
                              │
              ┌───────────────┼───────────────┐
              │               │               │
        ┌─────▼─────┐   ┌─────▼─────┐   ┌────▼─────┐
        │  LM Studio │   │   Ollama  │   │    MLX   │
        │  (DEFAULT) │   │           │   │ (NOT REC)│
        │  :1234     │   │   :11434  │   │  :28100  │
        └───────────┘   └───────────┘   └──────────┘

✨ Features

• 🔍 Semantic Search — Find notes by meaning, not just keywords
• 🏠 Local & Private — Everything runs on your machine, no external API calls
• 🔄 Multiple Backends — LM Studio (default), Ollama, or MLX
• ⚡ Incremental Indexing — Only re-index changed files
• 📝 YAML Support — Extracts metadata from frontmatter

🚀 Quick Start

1. Install a Backend

Choose one — LM Studio is recommended:

LM Studio (GUI, easy):

# Download from https://lmstudio.ai/
# Load an embedding model and start the server

Ollama (CLI, simple):

curl -fsSL https://ollama.ai/install.sh | sh
ollama pull qwen3-embedding:0.6b
ollama serve

2. Index Your Notes

npx obedding index --vault ~/.obsidian/Projects

3. Search

npx obedding search "what did we decide about caching?"

📖 Usage

# Index notes
npx obedding index --vault ~/.obsidian/Projects

# Incremental indexing (faster)
npx obedding index --vault ~/.obsidian/Projects --incremental

# Semantic search
npx obedding search "database optimization"

# Top 5 results
npx obedding search "API design" --top-k 5

# Filter by relevance
npx obedding search "redis" --min-score 0.6

# JSON output
npx obedding search "architecture" --json

# Show statistics
npx obedding stats

# Clear all embeddings
npx obedding clear --force

🎯 Backend Options

| Backend | Server | Model | Dims | Notes | |---------|--------|-------|------|-------| | LM Studio --backend lmstudio | :1234 | text-embedding-qwen3-embedding-0.6b | 1024 | ✅ Default, GUI | | Ollama --backend ollama | :11434 | qwen3-embedding:0.6b | 768 | ✅ Stable | | MLX --backend mlx | :28100 | Qwen3-Embedding-0.6B-4bit-DWQ | 2048 | ⚠️ Known issues |

📁 YAML Frontmatter

---
type: cli
repo: my-repo
context: feature-x
tags: [decision, architecture]
title: Cache Strategy Decision
---

📚 Documentation

• Architecture — System design and components
• Backends — Backend setup and comparison
• API Reference — Complete CLI documentation
• CLAUDE.md — Project context for Claude Code

🛠️ Development

git clone https://github.com/tuannvm/obedding.git
cd obedding
npm install
npm run build
npm link
obedding search "test"

❓ Troubleshooting

Server not running?

# LM Studio
curl http://localhost:1234/v1/models

# Ollama
curl http://localhost:11434/api/tags

No results?

# Check stats
npx obedding stats

# Re-index
npx obedding index --vault ~/.obsidian/Projects

📄 License

MIT