openclaw.integrate('obsidian')

/** "What did I write about RAG last month?" — searches 2000 notes in 2 seconds */

// Obsidian vs Notion — when to choose Obsidian

Choose Obsidian if:

• You want 100% offline notes

• You have an existing large vault (1000+ notes)

• You care about data ownership (plain Markdown)

• You want backlinks and graph view queries

Choose Notion if:

• You need team collaboration

• You rely on databases and views

• You want web-based access

how_it_works.md

🏗 Architecture: Local Vector Search

OpenClaw indexes your vault on startup

Reads all .md files, generates embeddings, stores in local vector DB (no cloud)

Query via Telegram / any chat interface

Semantic search finds the most relevant notes for your question

LLM synthesizes an answer with citations

References the source notes, links back to your vault

config.yaml

⚙️ Configuration

# openclaw/config.yaml

integrations:

obsidian:

enabled: true

vault_path: "~/Documents/MyVault"

index_on_startup: true

watch_changes: true

embedding_model: "nomic-embed-text"

example.log

📱 Example Queries

You

What did I write about RAG architectures?

OpenClaw

Found 4 relevant notes: • RAG vs Fine-tuning (2025-11-12) — argues RAG is better for factual recall • Vector DB comparison (2025-10-03) — Chroma vs Qdrant vs Weaviate • My RAG implementation notes (2025-12-01) — code snippets included • LLM limitations for knowledge (2026-01-15)

You

Create a new note: 'OpenClaw Obsidian demo', link it to RAG vs Fine-tuning

OpenClaw

✓ Created: OpenClaw Obsidian demo.md Added [[RAG vs Fine-tuning]] backlink

🚀 Related

→ Notion

// Team-friendly alternative

→ Research Assistant

// Save briefings to Obsidian vault

❓ FAQ

Q1. Does it modify my vault files?

Only if you enable write mode. By default, it reads your vault in read-only mode for search and context.

Q2. Does it work with Obsidian plugins?

Yes. It integrates via the Local REST API plugin. Dataview queries are also supported.

Q3. Is my vault data sent to the cloud?

Never. OpenClaw reads your local vault files directly. With local LLMs, nothing leaves your machine.

← Back to Integrations