GitHub - satiricalguru/Local-Mind: 🧠 LocalMind is a premium, private offline AI desktop companion built with Electron, React, and Tailwind CSS. Run local LLMs, Whisper speech-to-text, TTS voice synthesis, and image/audio/video generation 100% locally. Features custom cross-platform frameless window controls, dynamic themes, hardware tuning, and live system monitoring. · GitHub
Skip to content

Repository files navigation

LocalMind Logo

LocalMind

pnpm Electron React Tailwind CSS SQLite

Every model. Your machine.
LocalMind is a state-of-the-art, 100% private, and local desktop companion for running LLMs, Speech-to-Text (ASR), Text-to-Speech (TTS), image generators, audio, and video synthesis offline on your hardware.


📸 Screenshots

System Dashboard
System Dashboard - Live machine status and model configurations

Model Hub
Model Hub - Discover and install local models optimized for offline compute

Model Comparison
Model Comparison - Test prompts on two local models side-by-side

App Settings
App Settings - Theme selection, download limits, storage path, and hardware acceleration


✨ Features

  • 💻 Cross-Platform Frameless Window: Native macOS titlebar integration ( traffic lights hidden inset) and custom Windows/Linux window controls (Minimize, Maximize, Close).
  • 💬 Interactive Playgrounds:
    • Chat & Code: Chat with local models, streaming tokens with live performance metrics (Tokens Per Second, Time to First Token) and hardware acceleration indicators.
    • Image Gen: Create high-fidelity visual assets locally.
    • Audio & Video Gen: Synthesize soundtrack loops and visual clips offline.
    • Voice Transcription (STT): High-speed speech-to-text powered by local Whisper engines with interactive audio file uploaders and live microphone support.
    • Speech Synthesis (TTS): Convert text to high-quality audio files using offline voice styles.
  • Model Hub: One-click download and management of optimized model weights (GGUF, GGML, ONNX) directly from HuggingFace.
  • ⚙️ Advanced Settings:
    • App Interface Themes: GPU-Noir Dark Mode, system-matching settings, and a custom light Vanilla Light theme.
    • Download Tuning: Throttle simultaneous queue limits and redirect storage paths dynamically (e.g. ~/LocalMind/models).
    • Hardware Acceleration: Switch between Auto-Tuned (Metal/CUDA), forced CPU-only, or forced Vulkan acceleration backends.
  • 📊 Benchmarks & Monitor: Stream real-time CPU, RAM, GPU, VRAM, and temperature telemetry directly inside the sidebar heartbeat gauge.

🏗️ Repository Architecture

LocalMind is structured as a monorepo powered by pnpm and turborepo for rapid development and builds:

├── apps/
│   ├── electron/        # Electron Main Process (IPC Router, DatabaseManager, DownloadManager)
│   └── renderer/        # React + Vite + TanStack Router (App Playgrounds & Settings UI)
├── packages/
│   ├── registry-schema/ # HuggingFace model catalogs and metadata declarations
│   └── shared-types/    # TypeScript type definitions shared across Main/Renderer

🛠️ Technology Stack

  • Core Desktop: Electron (frameless native windows, preload context bridge)
  • Frontend Layer: React 18, Vite, TanStack Router (type-safe file-routing)
  • Styling: Tailwind CSS v4 (using clean CSS variables and premium Glassmorphic panel filters)
  • Database: SQLite (powered by better-sqlite3 with an automatic JSON-database fallback if SQLite bindings fail to load)
  • Package Manager: pnpm (workspace filters)
  • Build Engine: Turborepo

🚀 Getting Started

Prerequisites

  • Node.js >= 20.0.0
  • pnpm >= 10.0.0

Installation

Clone the repository and install all dependencies:

git clone https://github.com/your-username/localmind.git
cd localmind
pnpm install

Running Local Development

To launch the Vite frontend server and start the Electron application concurrently:

pnpm dev

Building the Project

To compile the packages and build the distribution assets:

pnpm build

Packaging for Release

To package the app as a standalone installer (.dmg on macOS, .exe on Windows) locally:

pnpm package

⚠️ Installation Troubleshooting

Since the GitHub Actions release binaries are open-source and compiled in a cloud CI environment without a paid commercial certificate, your operating system's security gatekeeper might display warnings:

🍎 macOS: "LocalMind" is damaged and can't be opened or "Unidentified Developer"

This is Gatekeeper's default warning for unsigned/un-notarized applications. To open it:

  1. Drag the LocalMind app to your /Applications folder.
  2. Open your terminal and run:
    xattr -cr /Applications/LocalMind.app
  3. Alternatively, you can Right-click (or Control-click) the app icon in Finder, select Open, and click Open in the confirmation dialog.

🪟 Windows: Windows protected your PC (SmartScreen)

  1. Click More info in the popup window.
  2. Click Run anyway to launch the installer.

🎨 Design Theme Customization

LocalMind implements a custom CSS-variable theme engine. The styles are defined inside apps/renderer/src/styles/index.css and can be overridden by targeting the data-theme attribute:

  • GPU-Noir Dark: Dark theme optimized for deep black OLED displays and neon accents.
  • Vanilla Light: Pure white background surfaces with soft borders and light glassmorphism.

🔒 Privacy & Security

All conversations, audio recordings, generations, settings, and downloaded models are stored completely locally on your machine:

  • Database path: ~/LocalMind/localmind.db
  • Models path: Configurable (defaults to ~/LocalMind/models)
  • No external telemetry, cloud tracking, or network callbacks.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

About

🧠 LocalMind is a premium, private offline AI desktop companion built with Electron, React, and Tailwind CSS. Run local LLMs, Whisper speech-to-text, TTS voice synthesis, and image/audio/video generation 100% locally. Features custom cross-platform frameless window controls, dynamic themes, hardware tuning, and live system monitoring.

Resources

Stars

4 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages