Dark Matter IDE

AI-Powered Code Editor with Built-in Local LLM Integration

Dark Matter is an open-source code editor forked from VS Code OSS, designed to provide local AI assistance directly within the development workflow. It requires no cloud APIs, no subscriptions, and ensures that no project data leaves the local machine.

---

Key Features

Built-in Local AI Chat

Dark Matter includes a fully integrated local LLM chat agent that works with any OpenAI-compatible backend. Users can connect to Ollama, llama.cpp, LM Studio, or any other compatible server to interact with AI directly within the editor.

• Zero Configuration: Works out of the box with a local Ollama instance (default) or any OpenAI-compatible backend.
• Multi-Backend Support: Connect to Ollama, llama-server (llama.cpp), LM Studio, Jan, vLLM, or any /v1/chat/completions-compatible endpoint.
• Model Flexibility: Compatible with Gemma, Llama, Mistral, CodeLlama, DeepSeek, and any model supported by your backend.
• 100% Private: All AI processing occurs locally on the user's hardware.
• Workspace Awareness: The AI agent understands the project structure and local file contents.
• Remote Server Support: Capability to connect to a local LLM backend running on any machine within the network.

Dynamic Context Awareness

Dark Matter is designed for deep project understanding. By default, it targets a 128k token context window (configurable up to 256k). Upon connecting to a model, the IDE automatically queries the model's native limits and dynamically scales the context to match the maximum supported resolution of your local model.

You can manually adjust the Maximum Context Window in Settings (localLLM.maxContextWindow) to better suit your GPU's VRAM capacity.

Extension Marketplace

Full access to the Open VSX Registry, allowing users to install thousands of extensions for language support, themes, debugging, and productivity.

Core Development Features

Dark Matter retains all standard features of the VS Code ecosystem, including IntelliSense, an integrated terminal, Git support, and advanced debugging.

---

Hardware Requirements

Important

GPU Memory (VRAM) Warning

Dark Matter requests a very large context window (256,000 tokens) to ensure the AI understands your entire project. This requires significant GPU VRAM.

Users must ensure their hardware has enough total VRAM to accommodate both the Model and the Context Window.

• Model Size: A 7B-9B model typically requires ~5-8GB of VRAM.
• Context Overhead: A 256k context window can add an additional 4GB to 8GB of VRAM overhead depending on the model architecture.

If you experience "100% CPU usage" or slow responses, it usually means the LLM backend has run out of GPU memory and is falling back to the CPU. In this case, consider using a smaller model or reducing project size.

---

Downloads / Releases

If you prefer not to build from source, you can download pre-compiled binaries for Windows and Linux directly from the GitHub Releases page.

• Windows Installer: Download DarkMatter-win32-x64-Setup.exe for a full installation with Start Menu shortcuts, file associations, and PATH integration.
• Windows Portable: Download DarkMatter-win32-x64.zip for a portable build — extract and run Dark Matter.exe.
• Linux (Debian/Ubuntu): Download the .deb package for an easy installation, or the .tar.gz archive for a portable build.

---

Building From Source

Prerequisites

If you want to build Dark Matter IDE from source, you will need to install the following based on your OS:

Windows

1. Node.js (v22.x LTS recommended)
2. Python 3.10+
3. C++ Build Tools — Build Tools for Visual Studio 2022 with the "Desktop development with C++" workload.
4. Inno Setup 6 (Optional) — Required only for .exe installer.
5. Git
6. A local LLM backend — e.g., Ollama, llama-server, or LM Studio

Linux (Debian/Ubuntu)

1. Node.js (v22.x LTS recommended)
2. Python 3.10+
3. Build Tools: sudo apt-get install build-essential g++ libx11-dev libxkbfile-dev libsecret-1-dev libkrb5-dev
4. Git
5. A local LLM backend — e.g., Ollama, llama-server, or LM Studio

Building from Source (Windows)

Dark Matter includes a PowerShell script to automate the build and packaging process.

# 1. Clone the repository
git clone https://github.com/abmina/dark-matter-ide.git
cd dark-matter-ide

# 2. Run the build script
# This will install dependencies, compile the source, and build the installer
.\build_exe.ps1

# If you don't have Inno Setup, you can build just the application:
.\build_exe.ps1 -SkipInstaller

The compiled output will be in the built/ directory (or VSCode-win32-x64/ if skipping the installer).

Building from Source (Linux)

# 1. Clone the repository
git clone https://github.com/abmina/dark-matter-ide.git
cd dark-matter-ide

# 2. Install dependencies
npm ci

# 3. Download Electron
npm run electron

# 4. Start the watch process (compilation)
npm run watch

# 5. Launch the IDE (in a separate terminal)
./scripts/code.sh

To package for Linux (creates .tar.gz and .deb):

npm run gulp -- vscode-linux-x64-min
npm run gulp -- vscode-linux-x64-build-deb

Using the AI Agent

1. Start your local LLM backend (e.g., ollama serve for Ollama, or launch llama-server / LM Studio).
2. Launch Dark Matter.
3. Open the Chat panel from the primary sidebar.
4. Select the desired model from the dropdown menu and begin the session.

Tip: You can configure the backend URL and model via the status bar AI icon or in Settings under localLLM.*.

---

Architecture

Dark Matter extends VS Code OSS with the following components:

Component	Description
Local LLM Chat Agent	Integrated chat participant managing local AI communication via any OpenAI-compatible backend.
Language Model Provider	Registers local models as first-class providers within the IDE's model picker.
Large Context Integration	Pre-configured 128k–256k context awareness for deep project understanding.
Custom Welcome UI	Professional startup experience tailored for AI-driven development.

---

Privacy and Security

Dark Matter is designed with a privacy-first architecture. All AI features are executed locally:

• No project data is transmitted to external servers.
• No third-party API keys or accounts are required.
• Telemetry and usage tracking are disabled by default.

---

Contributing

Contributions to the project are reviewed and welcomed.

1. Fork the repository.
2. Develop features in a dedicated branch (git checkout -b feature/name).
3. Commit changes with descriptive messages.
4. Submit a Pull Request for review.

---

License

Licensed under the MIT License. Dark Matter is a fork of Visual Studio Code - Open Source.

---

Acknowledgments

• The Microsoft VS Code team for the foundational platform.
• The Ollama team for democratizing local LLM access.
• The llama.cpp team for making high-performance local inference accessible.
• The Open VSX Registry for maintaining an open extension ecosystem.