Reor is an AI-driven desktop note-taking application that enhances personal knowledge management through features such as automatic linking of relevant notes, answering questions on notes, semantic search, and creating AI flashcards. All data is stored locally, and it supports a Markdown editor similar to Obsidian. The core assumption of the Reor project is that AI models for thinking tools should run locally by default. It leverages technologies like Ollama, Transformers.js, and LanceDB to enable the operation of large language models (LLM) and embedded models locally. It also supports connections to OpenAI or compatible APIs, such as Oobabooga.