Introducing the Reor Project
What is Reor?
The Reor Project is an innovative AI-powered desktop application designed to enhance personal knowledge management. It offers a unique approach to note-taking by automatically linking related notes, assisting users with questions about their notes, providing semantic search capabilities, and even generating AI-powered flashcards. A standout feature of Reor is its commitment to local data storage, ensuring all your information is kept safe on your device. It employs an Obsidian-like markdown editor for easy note editing.
The Vision Behind Reor
The core idea of Reor is rooted in the belief that AI tools for thought should operate locally as a standard practice. To achieve this, Reor leverages the capabilities of leading technologies like Ollama, Transformers.js, and LanceDB, which together allow local execution of both large language models (LLMs) and embedding models. Here’s how it works:
- Every note created gets broken down into smaller parts and stored in an internal vector database.
- Notes that are related are automatically connected through vector similarity.
- The LLM-powered question and answer feature utilizes Retrieval-Augmented Generation (RAG) on the user's note repository.
- All data can be searched semantically, enabling intuitive access to information.
A Unique Method of Thought Augmentation
Reor can be visualized as an RAG application with dual sources of generation—the AI and the user. In the Q&A mode, the AI uses retrieved context from your notes to provide precise answers. Similarly, when in editor mode, users can activate a sidebar that displays notes related to the current topic, thus enhancing the thought process by encouraging cross-referencing with other ideas.
How to Get Started
Getting started with Reor is straightforward. Users can download the app from reorproject.org or through its releases page, with versions available for Mac, Linux, and Windows. Once downloaded, Reor can be installed like any standard application.
Local Model Operations
Reor’s interaction with Ollama allows users to download and run models directly within the app. This can be done by navigating to Settings and selecting "Add New Local LLM," where the desired model name can be entered for download. A catalog of models is available at the Ollama library. Additionally, there is support for connecting to an OpenAI-compatible API.
Note Importing and Integration
Reor operates from a single directory on your file system, chosen upon the first launch. To import notes from other applications, users will need to manually add markdown files to this directory. While there may be initial challenges with frontmatter parsing, future integrations with other apps are anticipated.
Building from the Source
For those interested in exploring Reor’s development side, setting up from the source requires installing Node.js. The steps include:
- Cloning the repository:
git clone https://github.com/reorproject/reor.git - Installing dependencies:
npm install - Running the development environment:
npm run dev - Building the app:
npm run build
Contribution Opportunities
The Reor Project is continuously expanding, and contributions are welcome. Whether you have a new feature idea, want to fix bugs, or improve stylistic elements, the Reor team is open to collaboration. Interested developers can begin by exploring the issues page and following the contributing guide.
Licensing and Further Information
Reor is released under the AGPL-3.0 license. For more detailed information, users can refer to the LICENSE file. The name "Reor" is derived from the Latin word meaning "to think," embodying the essence of its function as a thought-enhancing tool.