Engram

A powerful search and knowledge graph plugin for Obsidian. Index your vault with SQLite FTS5 for instant full-text search, connect to Ollama for semantic search, and visualize your knowledge as an interactive graph.

Features

Full-Text Search (FTS5)

• Instant keyword search powered by SQLite FTS5 running entirely in your browser via WebAssembly
• Highlighted snippets with matched terms
• Filter by directory or tag
• Real-time incremental indexing — edits are reflected immediately

Semantic Search (Ollama)

• Find notes by meaning, not just keywords
• Uses local Ollama embeddings (bge-m3 by default) — your data never leaves your machine
• Cosine similarity ranking with percentage scores
• Optional — works without Ollama; enable when you want deeper search

Knowledge Graph

• Automatically extract entities (people, projects, organizations) and relationships from your notes
• Interactive network visualization with drag, zoom, and click-to-explore
• Filter by entity type or relationship type
• Detail panel with entity facts and connections

Dashboard

• Vault statistics at a glance: file count, word count, directory distribution
• Ollama connection status
• One-click actions: reindex, embed, extract graph

DB Explorer

• Browse the underlying SQLite database directly
• Inspect tables, columns, and rows
• Useful for debugging and advanced queries

Installation

From Community Plugins (Recommended)

1. Open Obsidian Settings → Community Plugins
2. Search for Engram
3. Click Install, then Enable

Manual Installation

1. Download main.js, manifest.json, styles.css, and sql-wasm.wasm from the latest release
2. Create a folder <vault>/.obsidian/plugins/obsidian-engram/
3. Copy the downloaded files into that folder
4. Restart Obsidian and enable the plugin in Settings → Community Plugins

Setting Up Semantic Search (Optional)

Semantic search requires Ollama running locally. This is entirely optional — the plugin works great with just keyword search and the knowledge graph.

1. Install Ollama: brew install ollama (macOS) or visit ollama.com
2. Start the server: ollama serve
3. Pull the embedding model: ollama pull bge-m3
4. In Engram settings, enable "Semantic Search" and click "Run Embedding"

The default model is bge-m3 (multilingual, 1024-dim). You can change this in settings.

Commands

Command	Description
Open Engram	Open the Engram panel
Reindex Vault	Re-scan all markdown files
Search	Jump to keyword search
Run Embedding	Generate embeddings for semantic search
Extract Graph	Extract entities and relationships
Load Vectors	Load embedding vectors into memory

Access commands via the Command Palette (Cmd/Ctrl + P) or use Cmd/Ctrl + K inside the Engram panel.

Settings

Setting	Default	Description
Auto-index on startup	On	Incrementally index changed files when Obsidian starts
Skip directories	node_modules, .git	Directories to exclude from indexing
Enable semantic search	Off	Connect to Ollama for embedding-based search
Ollama URL	http://localhost:11434	Ollama server address
Ollama model	bge-m3	Embedding model name
People directory	(empty)	Path to your people notes (e.g., People/) for graph extraction

How It Works

• Database: SQLite compiled to WebAssembly via sql.js-fts5, stored in the plugin folder as vault.db
• Indexing: Parses markdown files, extracts frontmatter/tags/wiki-links, and inserts into FTS5 virtual tables
• Semantic search: Calls Ollama's /api/embeddings endpoint locally, stores vectors in the database, and performs cosine similarity search in memory
• Graph extraction: Rule-based entity and relationship extraction from wiki-links, frontmatter, and heading patterns
• Real-time updates: Listens to Obsidian vault events (modify, delete, rename) for incremental indexing

Mobile Support

The plugin works on mobile devices for keyword search and knowledge graph features. Semantic search requires Ollama, which is a desktop-only service, so it is unavailable on mobile.

License

MIT