# Smart Connections - AI-Powered Note Discovery ## What It Does Smart Connections is an Obsidian plugin that uses AI vector embeddings to find semantically related notes in your vault. Instead of relying on keywords or manual links, it understands the *meaning* behind your notes and surfaces relevant connections automatically. **Key capabilities:** - **Semantic search** - Find notes by concept, not just keywords - **Automatic connections** - Discover related notes without manual linking - **AI chat** - Ask questions about your vault with context-aware responses - **Local processing** - Runs on your machine, works offline, zero cloud dependencies ## How It Works ### Vector Embeddings Explained Smart Connections converts each note into a mathematical representation called a "vector embedding." Notes with similar meanings have similar vectors, allowing the plugin to find conceptual relationships even when notes use different words. **Example:** - A note about "Second Brain" methodology - A note about "PKM (Personal Knowledge Management)" - A note about "Zettelkasten" These notes might not share keywords, but Smart Connections recognizes they're all about knowledge management systems and surfaces them together. ### The Technology - **Embedding Model**: TaylorAI/bge-micro-v2 (runs locally) - **Processing Speed**: ~17,000 tokens/second on modern hardware - **Storage**: Embeddings stored locally in `~/Library/Application Support/smart-connections/` - **No API keys required** - Everything runs on your machine ## Installation 1. Open Obsidian → Settings → Community plugins 2. Browse plugins and search "Smart Connections" 3. Install by Brian Petro 4. Enable the plugin 5. Initial embedding process begins automatically **First-time setup:** - Plugin scans your entire vault - Creates embeddings for all notes (one-time process) - Example: 5,900 sources embedded in ~3.25 seconds - Creates 9,618 embeddings total (includes headings, blocks) ## Core Features ### 1. Smart Connections Sidebar Access via: - Click connections icon in ribbon - Command palette: "Open Smart Connections view" - Hotkey: **⌥ C** (Alt+C) **The sidebar shows:** - Top related notes to your current file - Similarity scores (0.0 to 1.0) - Preview snippets of matched content - Quick actions for each result ### 2. Understanding Connections Each result shows: - **Score** (e.g., 0.89) - Higher = more related - **Note title** - The related note - **Preview snippet** - Context showing why it matches - **Expand/collapse** - Toggle full preview **Interacting with results:** - **Drag to add link** - Drag result into any note to create `[[link]]` - **Hover preview** - Hold ⌘/Ctrl while hovering to see Obsidian's built-in preview - **Click to open** - Click preview to open note for editing - **Refresh** - Re-embed current note and rerank results ### 3. Semantic Search (Lookup) Search by concept instead of keywords: - Command palette: "Smart Connections: Lookup" - Type a question or concept - Results ranked by semantic similarity **Example queries:** - "How do I organize my tasks?" (finds productivity notes) - "Ideas about creative workflows" (finds creativity + workflow notes) - "Notes about learning techniques" (finds education/study notes) ### 4. Smart Chat Interface AI-powered conversations with your vault as context. **Key features:** - **System prompt** - Set AI's role/style before sending messages - **Context builder** - Manually select which notes to include - **@ mentions** - Use `@` to reference specific notes - **Show connections** - Pull in semantically related notes automatically - **Show links** - Add notes linked to selected context items - **History** - Rename and revisit past conversations **Chat workflow:** 1. Click "Edit context" to build your knowledge base for the conversation 2. Type to search suggestions and add notes 3. Use "Show Connections" to auto-add related notes 4. Use "Show Links" to include linked notes 5. Click **X** to remove items or "New" to clear all 6. Ask your question 7. AI responds using only the context you've provided **Privacy note:** Chat uses your selected notes as context. You control exactly what information the AI can access. ## When to Use It ### Use Smart Connections When: - **Exploring a topic** - Find all related notes on a concept - **Making new connections** - Discover relationships you missed - **Research synthesis** - Gather notes across different areas - **Writing** - Find supporting ideas from existing notes - **Reviewing** - Rediscover forgotten relevant notes ### Use Traditional Search When: - Looking for exact phrases or quotes - Finding notes with specific tags - Searching for file names - Need regex or boolean operators Smart Connections complements Obsidian's built-in search—use both together. ## Performance Notes **Initial embedding:** - First run processes entire vault (one-time operation) - ~17,000 tokens/second on M1/M2/M3 Macs - Example: 5,900 sources in 3.25 seconds **Ongoing updates:** - New notes embedded automatically on save - Edited notes re-embedded when changed - Minimal performance impact during normal use **Storage:** - Embeddings stored locally - Typical vault: 50-200MB for embeddings database - Located at `~/Library/Application Support/smart-connections/` ## Pro Tips ### Organizing Context for Chat Build reusable context sets: 1. Create context with frequently-used notes 2. Use "Show Connections" to expand the set 3. Save as named context (future feature) ### Improving Connections - **Write clear headings** - Embeddings index by heading level - **Use descriptive block text** - Each block can be embedded separately - **Link related concepts** - Helps reinforce semantic relationships - **Regular reviews** - Use Smart Connections during weekly reviews to find orphan notes ### Integration with Other Workflows - Use Smart Connections to find notes, then link them manually - Combine with Dataview queries for structured + semantic search - Use chat for exploratory research, then synthesize into permanent notes ## Community and Support - **GitHub Repository**: https://github.com/brianpetro/obsidian-smart-connections - **Report bugs**: Open issue on GitHub - **Share workflows**: GitHub Discussions - **Smart Guide GPT**: Chat assistant for tips and advanced workflows ## Privacy and Philosophy Smart Connections follows these principles: - **Transparency** - Open source, auditable code - **User ownership** - Your data never leaves your machine - **Offline-first** - No internet required after installation - **Zero cloud dependencies** - No API keys, no external services **Mission:** "We build tools that turn raw notes into living knowledge." ## Related Articles - [[Vector Embeddings for Note Retrieval]] - Technical deep dive - [[Semantic Search vs Traditional Search]] - When to use each approach - [[Building Context for AI Chat]] - Advanced chat workflows - [[Obsidian Plugin Ecosystem]] - Overview of essential plugins ## See Also - [[Langchain]] - AI framework concepts relevant to Smart Connections - [[Vectorized Databases]] - Understanding how embeddings work - [[Second Brain Methodology]] - PKM context for using Smart Connections