Intelligent File Organization with Privacy-First Local AI
Features • Download • Support • Quick Start • Documentation • Contributing • Changelog
StratoSort Core helps you organize messy files with local AI that runs on your machine. It analyzes content (not just filenames), suggests where files belong, and gives you semantic search with Knowledge OS and graph tools. Your data stays local, and you can start with a normal installer - no CLI setup required.
- Windows/macOS installers: GitHub Releases
- Step-by-step install help: Install Guide
- How to use the app: User Guide
- Want to help test? Beta Tester Guide
If you run into an issue or have an idea, these links are the fastest way to help:
- Issues tab: View all issues
- Report a bug: Open bug report template
- Request a feature: Open feature request issue
- Contributing guide: CONTRIBUTING.md
See StratoSort in action
Desktop: Video plays directly below | Mobile: Click the filename to watch
ElStratoSortShortDemo.mp4
- In-Process AI Engine — Embedded
node-llama-cppandOrama. No more background services to manage! - Zero-Setup Experience — Just install and run. Models are downloaded automatically.
- GPU Acceleration — Automatic detection of Metal (macOS), CUDA (Windows/Linux), or Vulkan.
- Performance Boost — Faster startup, lower memory footprint, and improved search latency.
See CHANGELOG.md for complete release notes.
StratoSort Core is the successor to the original StratoSort Stack (legacy repository). This repository represents a clean break starting from v2.0.0, focusing on a streamlined, in-process AI architecture. For versions prior to v2.0.0, or to view the legacy codebase, please visit the original repository.
| Feature | Description |
|---|---|
| Local AI Intelligence | Built-in AI (node-llama-cpp) to understand file content, not just filenames |
| Privacy-First Design | Zero data exfiltration. All processing happens locally on your device |
| Smart Folder Watcher | Real-time monitoring that automatically analyzes and sorts new files as they arrive |
| Image Understanding | Vision models and OCR categorize screenshots, photos, and scanned documents |
| Knowledge Graph | Interactive visualization of file relationships, clusters, and semantic connections |
| Semantic Search | Find files by meaning using Orama Vector Search and AI Re-Ranking |
| Safe Operations | Full Undo/Redo capability for all file moves and renames |
- Download the installer for Windows or macOS.
- Run it — allow the app if your OS shows a security prompt (see Install Guide).
- First launch — choose a model profile and approve the download (~2-5GB, one-time).
No terminal, Python, Docker, or API keys. See the full Install Guide for step-by-step instructions on both platforms and how to handle unsigned-app prompts.
| Requirement | Specification |
|---|---|
| Operating System | Windows 10/11 (64-bit), macOS 10.15+, or Linux |
| Memory | 8GB RAM minimum (16GB recommended for best performance) |
| Storage | ~2-5GB for AI models (depends on profile chosen) |
| GPU (Optional) | NVIDIA CUDA, Apple Metal, or Vulkan-compatible for acceleration |
git clone https://github.com/iLevyTate/StratoSortCore.git
cd StratoSortCore
npm ci
npm run devFirst Launch: The app automatically downloads required AI models (GGUF format) on first run. GPU acceleration is auto-detected.
Default Models (Base Small): Llama 3.2 3B (text), LLaVA Phi-3 Mini (vision), all-MiniLM-L6-v2
(embeddings). A "Better Quality" profile with larger models is available during setup. Change
defaults in src/shared/aiModelConfig.js. See docs/CONFIG.md for details.
Define categories with natural language descriptions. The Smart Folder Watcher monitors your downloads or designated folders, automatically analyzing new items and routing them based on content understanding.
StratoSort Core doesn't just read text files—it uses computer vision to interpret images and Tesseract OCR to extract text, enabling automatic organization of receipts, screenshots, and scanned PDFs.
Search implies meaning. The built-in ReRanker Service uses a compact LLM to evaluate results, surfacing conceptually relevant matches rather than simple keyword hits.
| Principle | Implementation |
|---|---|
| 100% Local Processing | No internet required after model download |
| Zero Telemetry | No data collection or tracking of any kind |
| Open Source | Full source code available for inspection |
| Secure by Default | Context isolation, input validation, path sanitization |
See SECURITY.md for the complete security policy.
| Document | Description |
|---|---|
| Install Guide | End-user install (Windows & Mac, no CLI) |
| User Guide | Feature walkthrough for everyday use |
| Beta Tester Guide | Testing + bug reporting for contributors |
| Getting Started | Developer setup and build guide |
| Architecture | System design and data flow |
| Graph Features | Knowledge Graph capabilities |
| IPC Contracts | IPC communication specifications |
| Release Guide | Release process and checks |
Contributions are welcome. Please see CONTRIBUTING.md for guidelines.
- Fork the repository
- Create a feature branch
- Make changes and verify with
npm test - Submit a Pull Request
StratoSort builds on ideas from the growing ecosystem of AI-powered file organization:
| Project | Description |
|---|---|
| llama-fs | Self-organizing filesystem with Llama 3; pioneered watch mode learning |
| Local-File-Organizer | Privacy-first organizer using Llama3.2 and LLaVA |
| ai-file-sorter | Cross-platform desktop app with preview and undo |
| Hazel | Industry standard Mac file automation |
| Sparkle | Mac AI organizer using GPT-4/Gemini |
StratoSort Personal Use License 1.0.0 (Based on PolyForm Noncommercial)
See LICENSE for details.
GitHub • Report Bug • Request Feature
Built with node-llama-cpp and Orama
© 2026 StratoSort Team