The author created a personal knowledge assistant using Obsidian, DuckDB, and MotherDuck that leverages backlinks from their notes to uncover connections and insights. This Retrieval-Augmented Generation (RAG) system operates locally with DuckDB as a vector database, while also syncing to MotherDuck for a web application that runs in the browser. The tool allows users to search for topics and reveals connected notes up to two hops away, alongside semantically similar content. The goal is to tap into forgotten ideas and hidden relationships in the author's extensive collection of 8,963 markdown notes. Building the web app involved using Claude Code and DuckDB's Vector Similarity Search Extension, with the author emphasizing the importance of a local-first approach. They wanted to maintain control over sensitive notes and avoid relying on external AI services that might compromise data integrity. The system processes markdown files, converting them into numerical vectors for semantic search, and respects the structure of the notes during chunking. The architecture allows for efficient searching and retrieval of related notes, providing meaningful insights without relying on large language models. The author’s project reflects a hands-on use of data engineering skills to enhance personal knowledge management. By integrating various tools and methods, they aimed to create a resource that streamlines the brainstorming process and helps generate new ideas from existing content. The outcome is a robust, locally controlled system that supports the author's creative work while providing a web app for easy sharing and exploration of their knowledge base.