🧠 Re.mind

Re.mind is a context management CLI that lets users funnel conversational and technical data into a local Markdown vault, empowering LLMs to ground their responses by extracting exactly the information they need.

It transforms raw conversational data from ChatGPT, Claude, Gemini or Copilot into structured documents that can then be edited with tools like ObsidianMD. It creates a semantic index that AI agents can use to query documentation with millimeter precision without the need to copy and paste files to multiple locations.

---

🚀 Key Features

• 100% Local: Your data lives on your hard drive (Re.mind vault/).
• Simplicity: Based on Markdown and 100% compatible with Obsidian.
• Semantic Navigation (Dot Notation): Access any heading or knowledge block using efficient logical paths (project.folder.file.heading).
• Hybrid Tagging System: Native support for inline #hashtags that are automatically indexed globally.
• Context Friendly: Designed for an LLM to extract exact context blocks on demand, reducing token consumption.

---

📦 Installation

Re.mind is written in Python and distributed via PyPI. It requires Python 3.9 or higher.

pip install re.md

---

🛠️ Workflow and Commands

The core loop of Re.mind is encapsulated in writing (ingestion and indexing) and reading (querying and extraction).

1. Installation

• remind install Sets up the Re.mind vault base directory and deploys the .agents/skills/remind/SKILL.md file.

  $ remind install

• remind config --set-vault <path> Configures a custom location for your Re.mind vault globally. The CLI will remember this path across all your terminal sessions. It accepts both absolute and relative paths.

  $ remind config --set-vault "~/My_Custom_Vault"
  $ remind config --set-vault "./local_vault"

2. Data Management (Write)

• remind init <name> Initializes a new project notebook in your Vault. Creates the technical structure (.remind/) and generates the unique project slug.

  $ remind init "Trading Bot"

• remind import Scans the global import/ folder looking for .csv or .json exports (Google Takeout, Copilot, Claude or OpenAI) and generates clean, chronologically structured Markdown notebooks in a temporary inbox (_Inbox_). Smartly avoids duplicates.

  $ remind import

• remind index The core engine. Scans every project in the Vault, detects blocks, extracts inline hashtags (e.g., #architecture), and rebuilds the semantic map (map.index) and the auxiliary coordinate files (sidecars). Note: Always run this after manually reorganizing or editing your Markdown files in Obsidian.

• remind write <logical_path> <document_name> <temp_file> Creates a new node or overwrites an existing one. logical_path targets the project or a directory within it. document_name is the actual document name without the .md extension — wrap in quotes if it contains spaces (e.g. 'Meeting Notes'). temp_file is the path to the file holding the content.

• remind append <logical_path> <temp_file> Appends content from a temporary file to the end of an existing node.

3. Query and Extraction (Read)

• remind map [logical_path] Displays a visual tree with the structure of the indexed knowledge. With no arguments, it shows the global state of the Vault.

  $ remind map tradinbot4a2.fe

• remind tag <slug> list | remind tag <slug> <tag> Transversal search system. Lists all tags in a project sorted by popularity, or searches for a specific tag, returning the exact paths of the documents that contain it.

  $ remind tag tradinbot4a2 list
  $ remind tag tradinbot4a2 architecture

• remind me <paths> The key command for knowledge extraction. Breaks through the index layer to go directly to the hard drive and spit out the exact text via the terminal. Supports brace expansion {} syntax to extract multiple nodes simultaneously.

  $ remind me tradinbot4a2.fe.spec.{fsadx1a,fsbol2b}

---

🏗️ Internal Architecture

Re.mind abstracts complexity through a system of short deterministic identifiers (Slugs) and a coordinate map (Sidecars).

Every Markdown file has a hidden twin JSON file in the .remind/sidecars/ folder. This file acts as a spatial coordinate system: it maps the exact start and end lines of every heading and the tags the document contains, allowing the remind me command to extract surgical snippets without having to parse heavy text files in real-time.

---

🤝 Contributing

If you want to collaborate, clone the repository and install it in editable mode:

git clone https://github.com/cgpp5/Re.mind.git
cd Re.mind
pip install -e .

---

📄 License

This project is licensed under the GNU Affero General Public License v3.0. See the LICENSE file for details.