Smart Connections is an Obsidian plugin that leverages AI embeddings to automatically link related notes and enable conversational interaction with your knowledge base. It's designed for Obsidian users seeking to enhance productivity and discover hidden connections within their notes, supporting both local and API-based AI models.

How It Works

The plugin analyzes your notes by creating numerical representations called "embeddings," which are stored locally. When you view a note, it uses these embeddings to suggest related content in real-time via the Smart View. The Smart Chat feature allows you to query your notes conversationally, using retrieved notes as context for AI models. This approach aims to surface relevant information efficiently without manual tagging or linking.

Quick Start & Requirements

• Install: Via Obsidian Community plugins.
• Prerequisites: Obsidian desktop. Local models require transformer.js (sandboxed in an iframe). API models require API keys and potential pre-paid credits (e.g., OpenAI).
• Setup: Out-of-the-box with a default local embedding model (BGE-micro). API key setup is required for API models.
• Docs: Getting Started

Highlighted Details

• Supports a wide range of AI models, including local options via Ollama and LM Studio, and cloud APIs like OpenAI, Gemini, and Claude.
• Features "Smart View" for real-time note suggestions and "Smart Chat" for conversational querying of your notes.
• Chat history is saved as markdown notes.
• Offers "block" level matching and "highlight to find" functionality for granular context.

Maintenance & Community

• Actively developed by Brian Petro, with community contributions encouraged.
• Community discussions available on GitHub.

Licensing & Compatibility

• The README mentions commercial use is welcome, with a future commercial license planned, similar to Obsidian. Specific license details are not explicitly stated but it is open-source.

Limitations & Caveats

• Currently desktop-only, with mobile support in beta.
• Response quality in Smart Chat depends on note content and organization; AI can occasionally provide incorrect responses.
• Initial embedding processing can be resource-intensive depending on vault size.