What is Reor？
Reor is an open-source, privacy-focused AI-powered personal knowledge management (PKM) application that runs locally. It offers features such as automatic linking between related notes, semantic search, and Q&A capabilities. Users can edit notes using a Markdown editor similar to Obsidian. Leveraging technologies like Ollama, Transformers.js, and LanceDB, Reor enables local operation of large language models (LLMs) and embedding models, ensuring full data privacy. Reor also includes a local writing assistant to help users efficiently organize and retrieve knowledge. It supports multiple platforms, including macOS, Linux, and Windows, allowing users to download and install it easily.
Key Features of Reor
-
Automatic Linking of Related Notes: Reor can automatically detect and link notes with similar topics or content, eliminating the need for manual cross-referencing.
-
Semantic Search: Users can perform semantic searches without remembering the exact wording of a note—Reor returns the most relevant results based on meaning.
-
Q&A Functionality: Built-in LLMs allow users to ask questions, and the system generates answers based on the entire personal note corpus.
-
Local-First Design: All features are run and stored locally to ensure data privacy and control.
-
WYSIWYG Markdown Editing: Supports Markdown syntax with a “what-you-see-is-what-you-get” editing experience for streamlined note formatting.
-
Local Writing Assistant: Offers contextual suggestions from existing notes to support smoother and more coherent writing.
-
Support for Local Model Execution: Integrated with Ollama, allowing users to download and run models locally or connect to OpenAI-compatible APIs.
-
Cross-Platform Support: Available on macOS, Linux, and Windows, making it accessible across various devices.
Technical Principles Behind Reor
-
Vector Database: Reor chunks and embeds each note into an internal vector database. It uses vector similarity to automatically connect related notes.
-
LLM-Based Q&A: Uses large language models combined with Retrieval-Augmented Generation (RAG) to retrieve relevant content from notes and generate accurate answers.
-
Local Model Execution: Powered by Llama.cpp, Transformers.js, and LanceDB, Reor runs LLMs and embedding models locally to optimize performance and privacy.
-
Markdown Parsing and Editing: Notes are formatted in Markdown with WYSIWYG editing support for fast, structured note-taking.
-
Semantic Search Engine: Text is embedded into vectors, and similarity search is used to improve the accuracy and relevance of search results.
Official Links
-
Website: https://www.reorproject.org/
-
GitHub Repository: https://github.com/reorproject/reor
Application Scenarios
-
Personal Knowledge Management: Helps users efficiently organize and retrieve notes through automatic linking and semantic search.
-
Academic Research: Useful for researchers managing literature reviews, experimental notes, and references; it aids in writing and citing previous work.
-
Team Collaboration: Facilitates knowledge sharing and project documentation within teams, improving efficiency and enabling real-time collaboration.
-
Enterprise Knowledge Management: Supports management of internal documents, policies, and training materials, building a corporate knowledge base.
-
Creative Writing: Assists writers in managing ideas, drafts, characters, and plot points, while surfacing related content to inspire creativity.
