dive

What it does

Dive is a desktop application that provides a secure, local-first interface for running Ollama and Pi agentic AI models. It features an ultra-compact offline RAG (Retrieval-Augmented Generation) engine capable of indexing massive book libraries, making it ideal for researchers and knowledge workers who need to query large document collections privately.

Key Features

• Secure local-first architecture with no cloud dependencies
• Offline RAG engine for indexing and searching large book libraries
• PDF extraction and processing capabilities
• MCP (Model Context Protocol) tool integration for extended functionality
• Sandboxed custom skills for safe, isolated task execution
• macOS daemon support for background operations
• Support for EPUB and other document formats via Calibre integration

How to set up

Clone the repository from GitHub, ensure Ollama is installed and running locally on your system, configure your preferred model in Dive's settings, and start indexing your document library through the RAG interface. The application handles local processing entirely, with no external API calls required.