# 🎯 Logitum Adaptive Ring ### _Your Hardware Just Got Smarter_ **AI-powered, context-aware actions at your fingertips. Literally.** [![Made for Logitech](https://img.shields.io/badge/Made%20for-Logitech-00B8FC?style=for-the-badge&logo=logitech&logoColor=white)](https://logitech.com) [![HackaTUM 2025](https://img.shields.io/badge/HackaTUM-2025-FF6B35?style=for-the-badge&logo=hack-the-box&logoColor=white)](https://hack.tum.de/) [![Built with MCP](https://img.shields.io/badge/MCP-Powered-7C3AED?style=for-the-badge&logo=protocol&logoColor=white)](https://modelcontextprotocol.io)

[Features](#-key-features) • [How It Works](#-how-it-works) • [Installation](#-installation) • [Demo](#-demo) • [Architecture](#-architecture) ---

## 🤔 The Problem You know that feeling when you're drowning in tabs, constantly switching contexts, hunting through menus for features you _know_ exist somewhere? ``` ┌─────────────────────────────────────────────┐ │ Monday, 9:47 AM │ │ • 27 browser tabs open │ │ • Switched apps 184 times │ │ • Clicked through menus... so many menus │ │ • Lost in tool overload │ └─────────────────────────────────────────────┘ ``` **We've all been there.** It's like living in the 90s, but with more RAM. ## 💡 The Solution What if your Logitech device _just knew_ what you needed? What if it adapted to your workflow, learned your patterns, and brought the right tools to your fingertips **exactly when you need them**? ### Enter: **Logitum Adaptive Ring** 🎉 A Logitech plugin that transforms your MX device into an intelligent, self-adapting command center powered by **6,000+ AI tools** through the Model Context Protocol.

### ✨ Three Words: **Intelligence. Adaptation. Zero Config.**

--- ## 🚀 Key Features ### 🎯 **Instant Intelligence** Switch to VS Code? Get _Commit_, _Push_, _Debug_, _Run Tests_. Switch to Slack? Get _Send Message_, _@Mention_, _React_, _Search_. Switch to Chrome? Get _Bookmark_, _New Tab_, _History_, _Search_. **All automatically.** No configuration. No setup. Just works. ### 🧠 **Learns Your Workflow** The system tracks your behavior (with your permission, always): - Notices you run tests in VS Code 20 times a week - Automatically adds it to your actions ring - **Hardware that adapts to YOU** ### 🌐 **6,000+ Apps via MCP** Built on the [Model Context Protocol](https://modelcontextprotocol.io) — the open standard adopted by: - ✅ OpenAI (March 2025) - ✅ Google DeepMind (April 2025) - ✅ Microsoft & Anthropic - ✅ 6,480+ servers and growing ### 🔮 **Future-Proof** New MCP server published tomorrow? **You get it automatically.** No plugin updates. No configuration. It just appears. --- ## 🎬 How It Works ``` ┌─────────────────────────────────────────────────────────────────┐ │ 🔄 THE ADAPTIVE CYCLE │ └─────────────────────────────────────────────────────────────────┘ 1️⃣ You switch apps → Process Monitor detects it ↓ 2️⃣ Query MCP Registry → Find relevant servers ↓ 3️⃣ AI suggests workflows → "Commit", "Push", "Debug" ↓ 4️⃣ Actions Ring updates → Instantly available ↓ 5️⃣ You click action → MCP executes the tool ↓ 6️⃣ UI Monitor tracks patterns → "Hmm, they do this a lot..." ↓ 7️⃣ AI interprets behavior → "User runs tests frequently" ↓ 8️⃣ Vector clustering → Similar actions group together ↓ 9️⃣ Frequency threshold hit → "Want to add 'Run Tests'?" ↓ 🔟 Ring updated → Ring adapts permanently ♻️ REPEAT. IMPROVE. ADAPT. ``` --- ## 🏗️ Architecture ### Three-Tier Intelligence System ``` ┌─────────────────────────────────────────────────────┐ │ TIER 1: DISCOVERY (Instant) │ │ • MCP Registry lookup on app switch │ │ • AI suggests 6-8 common workflows │ │ • Actions Ring populates in <500ms │ └─────────────────────────────────────────────────────┘ ↓ ┌─────────────────────────────────────────────────────┐ │ TIER 2: OBSERVATION (Passive) │ │ • Windows UI Automation monitors clicks │ │ • Tracks focus changes, inputs, shortcuts │ │ • Stores interaction data (24-hour TTL) │ └─────────────────────────────────────────────────────┘ ↓ ┌─────────────────────────────────────────────────────┐ │ TIER 3: ADAPTATION (Intelligent) │ │ • Gemini AI interprets action sequences │ │ • VoyageAI embeddings cluster similar patterns │ │ • clustering algorithm identifies workflows │ └─────────────────────────────────────────────────────┘ ``` ### Tech Stack | Layer | Technology | Why We Love It | | ------------------ | --------------------------- | ------------------------------------------- | | 🎮 **Plugin** | Logitech Actions SDK (C#) | Native hardware integration | | 🤖 **AI Brain** | Google Gemini | Workflow understanding & suggestions | | 🧮 **Embeddings** | VoyageAI (`voyage-code-3`) | Semantic similarity detection | | 📊 **Clustering** | NN + Cosine Similarity | Pattern recognition | | 👀 **UI Tracking** | Windows UI Automation | Behavioral learning | | 🗄️ **Storage** | SQLite3 | Fast, local, private | | 🌐 **Ecosystem** | Model Context Protocol | 6,000+ tools, one standard | --- ## 🎨 What Makes This Special? ### vs. Traditional Solutions | Approach | Limitation | Logitum | | -------------------------- | ---------------------------------- | ----------------------------------------- | | **Keyboard Shortcuts** | Memorize 100+ combinations | ✅ Visual, contextual, max 8 at once | | **Command Palettes** | Still requires typing & searching | ✅ Physical button, instant | | **Claude Desktop** | Shows ALL tools, manual navigation | ✅ Shows RELEVANT tools, learns | | **Logitech Smart Actions** | Manual setup, no MCP, static | ✅ Auto MCP discovery, AI-powered, adapts | | **Other Plugins** | App-specific, limited scope | ✅ Universal (6k+ apps), future-proof | ### 🎯 The Killer Combo 1. **Physical Interface** - Tactile, instant, no context switching 2. **AI-Powered** - Understands context and intent 3. **Self-Learning** - Improves with your usage 4. **Open Standard** - MCP means infinite extensibility 5. **Zero Config** - Works immediately, improves automatically --- ## 🔧 Installation ### Prerequisites - **Hardware**: Logitech MX Master 3/4, Loupedeck CT/Live/Live S, or Razer Stream Controller - **OS**: Windows 10/11 (macOS support planned) - **Software**: Logitech Options+ / Loupedeck Software - **.NET**: 8.0 or higher ### Setup ```bash # Clone the repo git clone https://github.com/yourusername/logitum.git cd logitum/AdaptiveRingPlugin # Set up environment variables $env:GEMINI_API_KEY="your-gemini-api-key" $env:VOYAGEAI_API_KEY="your-voyageai-api-key" # Build the plugin cd src dotnet build -c Release # Plugin auto-deploys to: # C:\Users\\AppData\Local\Logi\LogiPluginService\Plugins\ ``` ### Get API Keys (Free Tiers Available!) - **Gemini**: [aistudio.google.com](https://aistudio.google.com) - **VoyageAI**: [voyageai.com](https://www.voyageai.com) --- ## 🎥 Demo [![image](https://github.com/user-attachments/assets/6c0e57cf-7e25-4b89-99f5-8502925e39d4)](https://www.youtube.com/watch?v=5bFkX00qs2A) --- ## 🧪 Development ### Project Structure ``` logitum/ ├── AdaptiveRingPlugin/ │ ├── src/ │ │ ├── Actions/ # Command implementations │ │ ├── Services/ # Core intelligence │ │ │ ├── ProcessMonitor.cs # App detection │ │ │ ├── MCPRegistryClient.cs # MCP integration │ │ │ ├── GeminiActionSuggestor.cs # AI suggestions │ │ │ ├── UIInteractionMonitor.cs # Behavior tracking │ │ │ ├── SemanticWorkflowProcessor.cs # Pattern analysis │ │ │ ├── VoyageAIClient.cs # Embeddings │ │ │ ├── VectorClusteringService.cs # Neighrest Neighbour Clustering │ │ │ └── ActionRankingService.cs # Smart sorting │ │ ├── Models/ # Data structures │ │ ├── Helpers/ # Utilities │ │ └── Scripts/ # Python AI services │ └── bin/Release/ # Compiled plugin ├── PLAN.md # Original hackathon plan ├── FEATURE_IMPLEMENTATION_SUMMARY.md # Technical docs └── README.md # You are here! 👋 ``` ### Build from Source ```bash # Debug build dotnet build # Release build (optimized) dotnet build -c Release # Watch mode (auto-rebuild) dotnet watch build ``` ### Run Tests ```bash # Unit tests dotnet test # Integration tests (requires hardware) dotnet test --filter Category=Integration ``` --- ## 🗄️ Database Schema ```sql -- Core tables apps → Registered applications app_actions → MCP actions per app mcp_servers → Known MCP registry entries -- Intelligence tables ui_interactions → Raw UI events (24h TTL) semantic_workflows → AI-interpreted patterns workflow_embeddings → Vector representations workflow_clusters → Grouped similar actions -- All stored locally in: -- C:\Users\\AppData\Local\Logitum\adaptive_ring.db ``` --- ## 🔐 Privacy & Security ### What We Track (If You Enable Learning) ✅ **App names** (e.g., "chrome.exe", "slack.exe") ✅ **Button clicks** (e.g., "clicked 'Commit'") ✅ **Window titles** (e.g., "Gmail - Inbox") ✅ **Action sequences** (e.g., "Type→Click→Submit") ### What We DON'T Track ❌ **Passwords** (masked as `******`) ❌ **Email/message content** ❌ **Browsing history URLs** ❌ **File paths or code** ### Where Data Lives 🏠 **100% Local** - Everything stored on your machine (and could stay there if you use a local LLM) 🔒 **No Cloud Sync** (unless you enable it) 🗑️ **24-Hour Auto-Delete** for raw UI events ### API Calls Only sent to **Gemini** (workflow interpretation) and **VoyageAI** (embeddings). **You control the API keys.** --- ## 🌟 Roadmap ### ✅ Completed (HackaTUM 2025) - [x] MCP Registry integration - [x] Gemini AI workflow suggestions - [x] Windows UI Automation tracking - [x] Vector embedding clustering - [x] Adaptive action learning - [x] SQLite3 with Vector Embeddings ### 🚧 In Progress - [ ] macOS support (via NSAccessibility) - [ ] Manual action editor UI - [ ] Export/import action profiles - [ ] Performance optimization (caching) ### 🔮 Future Vision - [ ] Multi-device sync (encrypted cloud) - [ ] Voice control integration - [ ] Cross-app workflow automation - [ ] Team collaboration (shared profiles) - [ ] Logitech Marketplace submission - [ ] Plugin SDK for custom actions --- ## 🤝 Contributing We'd love your help! Here's how: ### 🐛 Report Bugs Open an issue with: - OS version - Logitech device model - Steps to reproduce - Log files (found in `C:\Users\\AppData\Local\Logi\LogiPluginService\Logs\`) ### 💡 Suggest Features Open an issue with the `enhancement` label. ### 📝 Improve Docs Found a typo? Explanation unclear? PRs welcome! --- ## 🙏 Acknowledgments ### Built With - **[Logitech Actions SDK](https://logitech.github.io/actions-sdk-docs/)** - Hardware magic - **[Model Context Protocol](https://modelcontextprotocol.io)** - Open AI standard - **[Google Gemini](https://ai.google.dev/)** - Workflow intelligence - **[VoyageAI](https://www.voyageai.com)** - Semantic embeddings - **[MCP Registry](https://registry.modelcontextprotocol.io)** - Tool discovery ### Inspired By - The frustration of tab overload - The vision of tangible computing - The power of open standards - The community of 6,000+ MCP servers ### Special Thanks - **Logitech** - For making incredible hardware and opening it up to developers - **HackaTUM** - For the challenge that sparked this - **MCP Community** - For building the future of AI interoperability - **You** - For checking out this project! 🙌 --- ## 📞 Contact & Support ### Get Help - 💬 **GitHub Issues**: [github.com/mrsladoje/logitum/issues](https://github.com/yourusername/logitum/issues) - 📧 **Email**: mrsladoje@gmail.com ### Stay Updated ⭐ **Star this repo** to follow development 👀 **Watch** for release notifications 🔔 **Subscribe** to issues for discussions ---

## 🚀 Ready to Make Your Hardware Smarter? **[⬇️ Download Latest Release](https://github.com/mrsladoje/logitum/releases)** • **[🎥 Watch Demo](https://www.youtube.com/watch?v=5bFkX00qs2A)** --- ### _"The best way to predict the future is to build it."_ Made with ❤️ in 36 hours for HackaTUM 2025 **[Back to Top ↑](#-logitum-adaptive-ring)**

---

### 🌟 If this project helped you, give it a star! 🌟 [![GitHub stars](https://img.shields.io/github/stars/yourusername/logitum?style=social)](https://github.com/yourusername/logitum) [![GitHub forks](https://img.shields.io/github/forks/yourusername/logitum?style=social)](https://github.com/yourusername/logitum/fork) [![GitHub watchers](https://img.shields.io/github/watchers/yourusername/logitum?style=social)](https://github.com/yourusername/logitum)