[ { "awards": [ "1st Place Overall", "Logitech Challenge Special Mention", "JetBrains Challenge 4th Place" ], "categories": [ "1st Place Overall", "Logitech Challenge Special Mention", "JetBrains Challenge 4th Place" ], "description": "We decided to fill a major pain-point of the Logitech Master Series, and Logi Options+: The lack of Linux support. Being low-level engineers, and also linux users, we found that this ecosystem had little to no linux support, and any open source solutions were quite-frankly lacking. Logi Options+ provides a lot of customization and features that no available solutions on linux do, and Logitech's proprietary software does not currently support linux. Hence, we decided to reverse engineer the proto", "github_repo_count": 4, "github_repo_urls": [ "https://github.com/logilinux/.github", "https://github.com/logilinux/logilinux", "https://github.com/logilinux/logilinux-gui", "https://github.com/logilinux/logilinux-sdk" ], "github_repos_metadata": [ { "commit_count_default_branch": 5, "contributors_count": 1, "contributors_top": [ { "contributions": 5, "html_url": "https://github.com/ron0studios", "login": "ron0studios" } ], "created_at": "2025-11-23T08:29:08Z", "default_branch": "main", "description": null, "dirs_total_count": 1, "files_root_entries": [ { "path": "README.md", "size": 11, "type": "file" }, { "path": "profile", "size": 0, "type": "dir" } ], "files_total_count": 2, "first_commit_date_default_branch": "2025-11-23T08:29:08Z", "forks": 0, "has_ci": false, "has_contributing": false, "has_docker": false, "has_docs": false, "has_license_file": false, "has_notebooks": false, "has_readme_file": true, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/logilinux/.github", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 0, "issues_open": 0, "issues_total": 0, "languages_top": [], "last_commit_date_default_branch": "2025-11-23T09:28:25Z", "last_commit_oid_default_branch": "f5d7f28af7741ff08902bae0f1e2bfe3e543c78d", "latest_release_date": null, "latest_release_tag": null, "license_name": null, "license_spdx": null, "name_with_owner": "logilinux/.github", "owner": "logilinux", "parent_repo": null, "parent_url": null, "primary_language": null, "project_foreign_key": "lhp_7e3753aaa4e3a72c", "pull_requests_closed": 0, "pull_requests_merged": 0, "pull_requests_open": 0, "pull_requests_total": 0, "pushed_at": "2025-11-23T09:28:25Z", "readme_length": 11, "readme_text": "# logilinux", "readme_title": "logilinux", "releases_count": 0, "repo": ".github", "repo_name": ".github", "stars": 1, "topics": [], "updated_at": "2025-11-23T09:28:29Z", "url": "https://github.com/logilinux/.github", "watchers": 0 }, { "commit_count_default_branch": 22, "contributors_count": 3, "contributors_top": [ { "contributions": 15, "html_url": "https://github.com/ImArjunJ", "login": "ImArjunJ" }, { "contributions": 5, "html_url": "https://github.com/ron0studios", "login": "ron0studios" }, { "contributions": 2, "html_url": "https://github.com/GarThor", "login": "GarThor" } ], "created_at": "2025-11-22T11:20:49Z", "default_branch": "master", "description": null, "dirs_total_count": 10, "files_root_entries": [ { "path": ".gitignore", "size": 32, "type": "file" }, { "path": "CMakeLists.txt", "size": 242, "type": "file" }, { "path": "README.md", "size": 2823, "type": "file" }, { "path": "examples", "size": 0, "type": "dir" }, { "path": "lib", "size": 0, "type": "dir" }, { "path": "src", "size": 0, "type": "dir" }, { "path": "tools", "size": 0, "type": "dir" } ], "files_total_count": 43, "first_commit_date_default_branch": "2025-11-22T11:21:38Z", "forks": 2, "has_ci": false, "has_contributing": false, "has_docker": false, "has_docs": false, "has_license_file": false, "has_notebooks": false, "has_readme_file": true, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/logilinux/logilinux", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 2, "issues_open": 1, "issues_total": 3, "languages_top": [ { "language": "C++", "size": 129768 }, { "language": "CMake", "size": 4477 } ], "last_commit_date_default_branch": "2026-02-03T18:27:12Z", "last_commit_oid_default_branch": "b2d8d03aaabec2253a63cf5713235e04f2471d6e", "latest_release_date": null, "latest_release_tag": null, "license_name": null, "license_spdx": null, "name_with_owner": "logilinux/logilinux", "owner": "logilinux", "parent_repo": null, "parent_url": null, "primary_language": "C++", "project_foreign_key": "lhp_7e3753aaa4e3a72c", "pull_requests_closed": 0, "pull_requests_merged": 3, "pull_requests_open": 0, "pull_requests_total": 3, "pushed_at": "2026-02-03T18:27:12Z", "readme_length": 2822, "readme_text": "# LogiLinux - Logitech Device Library for Linux\n\nA C++ library and tools for interfacing with Logitech Creator devices on Linux, including the MX Dialpad and MX Keypad. This was made for LauzHack 2025.\n\n## What is This?\n\nLogiLinux provides a clean C++ library (liblogilinux) for working with Logitech input devices on Linux. It handles device discovery, event monitoring, and provides a type-safe API for building applications.\n\n## Quick Start\n\n### Building\n\n```bash\n./build.sh\n```\n\nOr manually:\n\n```bash\nmkdir build && cd build\ncmake ..\nmake -j$(nproc)\n```\n\n### Running Examples\n\n```bash\nsudo ./build/examples/dialpad-example\nsudo ./build/examples/mx-keypad-example \nsudo ./build/examples/volume-example\n```\n\n## Key Discovery\n\nThe MX Dialpad works as a standard Linux input device and sends events through the input subsystem. When you rotate the dial, it sends:\n\n- REL_HWHEEL (axis 6): Low-resolution steps (1-6 units per tick)\n- REL_MISC (axis 12): High-resolution angle (120 units per degree)\n- Positive values = clockwise, negative = counter-clockwise\n\n## Using the Library\n\n### Simple Example\n\n```cpp\n#include \n#include \n\nvoid onEvent(LogiLinux::EventPtr event) {\n if (auto rotation = std::dynamic_pointer_cast(event)) {\n std::cout << \"Rotated: \" << rotation->delta << \" steps\" << std::endl;\n }\n if (auto button = std::dynamic_pointer_cast(event)) {\n std::cout << \"Button \" << button->button_code << (button->pressed ? \" pressed\" : \" released\") << std::endl;\n }\n}\n\nint main() {\n LogiLinux::Library lib;\n auto device = lib.findDevice(LogiLinux::DeviceType::DIALPAD); // or MX_KEYPAD\n\n device->setEventCallback(onEvent);\n device->startMonitoring();\n\n return 0;\n}\n```\n\nCompile with:\n\n```bash\ng++ -std=c++17 myapp.cpp $(pkg-config --cflags --libs logilinux)\n```\n\nSee `lib/README.md` for full API documentation.\n\n## Prerequisites\n\n```bash\nsudo apt-get install build-essential cmake libudev-dev\nsudo dnf install gcc-c++ cmake libudev-devel\nsudo pacman -S base-devel cmake systemd\n```\n\n## Permissions\n\nTo avoid using sudo:\n\n```bash\nsudo bash -c 'cat > /etc/udev/rules.d/99-logitech.rules << EOF\nSUBSYSTEM==\"hidraw\", KERNEL==\"hidraw*\", ATTRS{idVendor}==\"046d\", ATTRS{idProduct}==\"bc00\", MODE=\"0666\"\nEOF'\n\nsudo udevadm control --reload-rules\nsudo udevadm trigger\n```\n\n## Supported Devices\n\n### MX Dialpad\n- Model: Logitech MX Dialpad \n- Vendor ID: `046d` (Logitech)\n- Product ID: `bc00`\n- Features: Rotation, button press, high-resolution scrolling\n- Protocol: HID++ 4.5\n\n### MX Keypad\n- Model: Logitech MX Creative Console / MX Keypad\n- Vendor ID: `046d` (Logitech) \n- Features: 9 programmable LCD buttons, navigation buttons\n- Capabilities: Button events, LCD image display, JPEG upload", "readme_title": "LogiLinux - Logitech Device Library for Linux", "releases_count": 0, "repo": "logilinux", "repo_name": "logilinux", "stars": 7, "topics": [], "updated_at": "2026-02-21T19:25:37Z", "url": "https://github.com/logilinux/logilinux", "watchers": 1 }, { "commit_count_default_branch": 19, "contributors_count": 2, "contributors_top": [ { "contributions": 13, "html_url": "https://github.com/ImArjunJ", "login": "ImArjunJ" }, { "contributions": 6, "html_url": "https://github.com/ron0studios", "login": "ron0studios" } ], "created_at": "2025-11-22T20:05:59Z", "default_branch": "master", "description": null, "dirs_total_count": 18, "files_root_entries": [ { "path": ".gitignore", "size": 1805, "type": "file" }, { "path": "README.md", "size": 10279, "type": "file" }, { "path": "index.html", "size": 664, "type": "file" }, { "path": "logilinux-ffi", "size": 0, "type": "dir" }, { "path": "logilinux.config.json", "size": 287, "type": "file" }, { "path": "logilinux.config.json.template", "size": 314, "type": "file" }, { "path": "package.json", "size": 1255, "type": "file" }, { "path": "pnpm-lock.yaml", "size": 97406, "type": "file" }, { "path": "postcss.config.js", "size": 81, "type": "file" }, { "path": "src-tauri", "size": 0, "type": "dir" }, { "path": "src", "size": 0, "type": "dir" }, { "path": "tailwind.config.js", "size": 2374, "type": "file" }, { "path": "tsconfig.json", "size": 696, "type": "file" }, { "path": "tsconfig.node.json", "size": 233, "type": "file" }, { "path": "vite.config.ts", "size": 776, "type": "file" } ], "files_total_count": 55, "first_commit_date_default_branch": "2025-11-23T00:30:40Z", "forks": 1, "has_ci": false, "has_contributing": false, "has_docker": false, "has_docs": false, "has_license_file": false, "has_notebooks": false, "has_readme_file": true, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/logilinux/logilinux-gui", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 0, "issues_open": 1, "issues_total": 1, "languages_top": [ { "language": "TypeScript", "size": 175094 }, { "language": "Rust", "size": 44604 }, { "language": "C++", "size": 13879 }, { "language": "CSS", "size": 4251 }, { "language": "C", "size": 3151 }, { "language": "JavaScript", "size": 2455 }, { "language": "CMake", "size": 938 }, { "language": "HTML", "size": 664 } ], "last_commit_date_default_branch": "2025-11-23T10:34:31Z", "last_commit_oid_default_branch": "59ece5eaf60eecd4e8c18be5db50cdd80dda420c", "latest_release_date": null, "latest_release_tag": null, "license_name": null, "license_spdx": null, "name_with_owner": "logilinux/logilinux-gui", "owner": "logilinux", "parent_repo": null, "parent_url": null, "primary_language": "TypeScript", "project_foreign_key": "lhp_7e3753aaa4e3a72c", "pull_requests_closed": 0, "pull_requests_merged": 0, "pull_requests_open": 0, "pull_requests_total": 0, "pushed_at": "2025-11-23T10:34:33Z", "readme_length": 9305, "readme_text": "# Logi## Features\n\n- 🎯 **Real-time Device Detection** - Automatic discovery of connected Logitech devices\n- 🖱️ **Interactive Device Visualization** - Realistic 3D-styled device models with live feedback\n- 🎨 **Modern Dark UI** - Logi Options+ inspired design with smooth animations\n- 🔘 **Button Press Detection** - Visual feedback for all button interactions\n- 🎡 **Dial & Wheel Monitoring** - Separate tracking for main dial and scroll wheel\n- ⚡ **Event System** - Real-time device event monitoring and handling\n\nA modern, Logi Options+ inspired desktop application for configuring Logitech devices on Linux.\n\nBuilt with **Tauri 2.0** (Rust backend) + **React 18/TypeScript** (frontend) + **logilinux** C++ library.\n\n## Features\n\n- 🎯 **Real-time Device Detection** - Automatic discovery of connected Logitech devices\n- 🖱️ **Interactive Device Visualization** - Realistic 3D-styled device models with live feedback\n- 🎨 **Modern Dark UI** - Logi Options+ inspired design with smooth animations\n- � **Button Press Detection** - Visual feedback for all button interactions\n- 🎡 **Dial & Wheel Monitoring** - Separate tracking for main dial and scroll wheel\n- ⚡ **Event System** - Real-time device event monitoring and handling\n- � **Permission Management** - Automated udev rules setup for device access\n\n## Supported Devices\n\n- ✅ **MX Dialpad Mouse** - Full support (dial, wheel, 4 buttons)\n- 🚧 **MX Creative Console** - Partial support (keypad visualization)\n\n## Prerequisites\n\n### System Dependencies\n\n**Ubuntu/Debian:**\n\n```bash\nsudo apt install build-essential curl wget file \\\n libwebkit2gtk-4.1-dev libssl-dev \\\n libayatana-appindicator3-dev librsvg2-dev \\\n cmake pkg-config libudev-dev\n```\n\n**Arch Linux:**\n\n```bash\nsudo pacman -S base-devel curl wget file \\\n webkit2gtk-4.1 openssl \\\n libayatana-appindicator librsvg \\\n cmake pkgconf systemd\n```\n\n**Fedora:**\n\n```bash\nsudo dnf install gcc-c++ curl wget file \\\n webkit2gtk4.1-devel openssl-devel \\\n libappindicator-gtk3-devel librsvg2-devel \\\n cmake pkgconfig systemd-devel\n```\n\n### Development Tools\n\n```bash\n# Install Rust (required for Tauri)\ncurl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh\nsource $HOME/.cargo/env\n\n# Install Node.js 20+ with pnpm\ncurl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.7/install.sh | bash\nsource ~/.bashrc # or ~/.zshrc\nnvm install 20\nnpm install -g pnpm\n```\n\n## Building & Running\n\n### 1. Clone the Repository\n\n```bash\ngit clone https://github.com/logilinux/logilinux-gui.git\ncd logilinux-gui\n```\n\n### 2. Install Dependencies\n\n```bash\n# Install frontend packages\npnpm install\n\n# Set up Python environment for custom actions (plugins)\n./setup_python.sh\n```\n\nThe Rust/C++ dependencies will be automatically built on first run.\n\n### 3. Run Development Build\n\n```bash\n# Run with Wayland compatibility fix (if needed)\nWEBKIT_DISABLE_COMPOSITING_MODE=1 pnpm tauri dev\n\n# Or just\npnpm tauri dev\n```\n\n### 4. Build Production Release\n\n```bash\n# Build production app (.deb package on Debian/Ubuntu)\npnpm tauri build\n\n# Output will be in: src-tauri/target/release/bundle/\n```\n\n## Project Structure\n\n```\nlogilinux-gui/\n├── src/ # React frontend (TypeScript)\n│ ├── components/ # UI components\n│ │ ├── layout/ # Layout components (Sidebar, etc)\n│ │ └── ui/ # shadcn/ui components\n│ ├── pages/ # Page components\n│ │ └── DevicesPage.tsx # Main device management page\n│ ├── lib/ # Utilities\n│ └── index.css # Global styles (device rendering CSS)\n│\n├── src-tauri/ # Rust backend (Tauri 2.0)\n│ ├── src/\n│ │ ├── commands/ # Tauri command handlers\n│ │ │ ├── devices.rs # Device discovery\n│ │ │ ├── events.rs # Event monitoring\n│ │ │ └── permissions.rs # Permission checking\n│ │ ├── logilinux/ # Rust FFI bindings\n│ │ │ ├── ffi.rs # Raw C bindings (auto-generated)\n│ │ │ └── mod.rs # Safe Rust wrappers\n│ │ └── main.rs # Tauri app entry point\n│ ├── build.rs # Build script (bindgen, CMake)\n│ ├── Cargo.toml # Rust dependencies\n│ └── tauri.conf.json # Tauri configuration\n│\n├── logilinux-ffi/ # C++ FFI wrapper library\n│ ├── CMakeLists.txt # CMake build config (fetches logilinux from GitHub)\n│ ├── include/\n│ │ └── logilinux_ffi.h # C-compatible header\n│ └── src/\n│ └── logilinux_ffi.cpp # C wrapper implementation\n│\n└── ref/ # Reference materials (not used in build)\n ├── logilinux/ # Local copy of logilinux library (reference only)\n └── mock_web/ # Design mockups\n```\n\n## Architecture\n\n```\n┌──────────────────────────────────┐\n│ React Frontend (TypeScript) │\n│ - DevicesPage.tsx │\n│ - Device visualization │\n│ - Event handling & state │\n└──────────┬───────────────────────┘\n │ Tauri IPC (invoke/listen)\n┌──────────▼───────────────────────┐\n│ Rust Backend (Tauri 2.0) │\n│ - discover_devices() │\n│ - start_device_monitoring() │\n│ - Event emission to frontend │\n└──────────┬───────────────────────┘\n │ FFI (bindgen)\n┌──────────▼───────────────────────┐\n│ C++ FFI Wrapper │\n│ - logilinux_ffi.cpp │\n│ - C-compatible interface │\n└──────────┬───────────────────────┘\n │ C++ API\n┌──────────▼───────────────────────┐\n│ logilinux Library (C++) │\n│ - Device communication │\n│ - Event callbacks │\n│ - HID++ protocol handling │\n└──────────────────────────────────┘\n```\n\n## How It Works\n\n1. **Frontend** (React) requests device discovery via Tauri command\n2. **Rust backend** calls C++ FFI wrapper functions\n3. **C++ wrapper** uses logilinux library to scan `/dev/input/*`\n4. **Device events** flow backwards through the stack:\n - C++ callback → Rust event handler → Tauri emit → React listener\n5. **UI updates** show real-time visual feedback (button presses, dial rotation)\n\n## Development Commands\n\n```bash\n# Frontend development (web preview only)\npnpm dev\n\n# Full Tauri app with hot reload\npnpm tauri dev\n\n# Build production binary\npnpm tauri build\n\n# Build specific bundle types\npnpm tauri build --bundles deb\npnpm tauri build --bundles appimage\n\n# Run Rust tests\ncd src-tauri && cargo test\n\n# Check Rust code quality\ncd src-tauri && cargo clippy\n\n# Clean build artifacts\nrm -rf src-tauri/target logilinux-ffi/build node_modules dist\n```\n\n## Troubleshooting\n\n### Device Not Detected\n\n1. **Check device files:**\n\n ```bash\n ls -l /dev/input/by-id/ | grep -i logitech\n ls -l /dev/input/event*\n ```\n\n2. **Test with console logs:**\n Open DevTools (Ctrl+Shift+I) and watch for device discovery logs.\n\n### Build Fails\n\n1. **Clean everything:**\n\n ```bash\n rm -rf src-tauri/target logilinux-ffi/build\n cargo clean\n pnpm tauri dev\n ```\n\n2. **Check CMake:**\n\n ```bash\n cmake --version # Should be 3.15+\n ```\n\n3. **Check bindgen dependencies:**\n ```bash\n sudo apt install libclang-dev # Ubuntu/Debian\n sudo pacman -S clang # Arch\n ```\n\n### Wayland Display Issues\n\nIf you see protocol errors:\n\n```bash\nWEBKIT_DISABLE_COMPOSITING_MODE=1 pnpm tauri dev\n```\n\n### Events Not Working\n\n1. **Check event monitoring started:**\n Look for \"✅ Device monitoring started\" in console\n\n2. **Check exclusive grab:**\n Look for \"Device grabbed exclusively\" message\n\n3. **Verify logilinux fetch:**\n ```bash\n ls logilinux-ffi/build/_deps/logilinux-src\n ```\n\n## Current Implementation Status\n\n### ✅ Completed Features\n\n- [x] Full Tauri 2.0 + React + TypeScript project setup\n- [x] Logi Options+ inspired dark UI (1100x700 fixed window)\n- [x] Realistic MX Dialpad device visualization with:\n - [x] Dot-textured casing\n - [x] 3D tactile buttons with visual press feedback\n - [x] Rotating main dial with indicator dot\n - [x] Rolling scroll wheel with ridges\n - [x] LED indicator\n- [x] C++ FFI wrapper (logilinux-ffi) auto-building via CMake\n- [x] Rust FFI bindings auto-generation via bindgen\n- [x] Device discovery polling (2-second intervals)\n- [x] Real-time event monitoring system:\n - [x] Button press/release detection (codes 275-278)\n - [x] Dial rotation tracking (accumulating angle)\n - [x] Wheel rotation tracking (separate from dial)\n - [x] C callback → Rust → Tauri emit → React flow\n- [x] Tauri 2.0 capabilities configuration\n- [x] Visual feedback animations (button scale, dial rotation, wheel roll)\n\n### 🚧 In Progress / Planned\n\n- [ ] Action mapping configuration\n- [ ] Profile management system\n- [ ] Settings persistence\n- [ ] Multi-device support\n- [ ] MX Creative Console full support\n- [ ] Custom button assignments\n- [ ] Application-specific profiles\n\n## Contributing\n\nContributions are welcome! Please see [PLAN_WEB.md](./PLAN_WEB.md) for the full development roadmap.\n\n## License\n\nGPL v3\n\n## Acknowledgments\n\n- Built on top of [logilinux](https://github.com/logilinux/logilinux) - The core C++ library for Logitech device communication\n- Powered by [Tauri](https://tauri.app/) - Rust-based desktop app framework\n- UI inspired by Logitech Options+ design language\n- Uses [shadcn/ui](https://ui.shadcn.com/) components", "readme_title": "Logi## Features", "releases_count": 0, "repo": "logilinux-gui", "repo_name": "logilinux-gui", "stars": 2, "topics": [], "updated_at": "2026-02-02T18:59:07Z", "url": "https://github.com/logilinux/logilinux-gui", "watchers": 1 }, { "commit_count_default_branch": 5, "contributors_count": 1, "contributors_top": [ { "contributions": 5, "html_url": "https://github.com/ron0studios", "login": "ron0studios" } ], "created_at": "2025-11-22T19:27:06Z", "default_branch": "master", "description": "Python support for logitech", "dirs_total_count": 7, "files_root_entries": [ { "path": ".gitignore", "size": 810, "type": "file" }, { "path": ".gitmodules", "size": 104, "type": "file" }, { "path": "LICENSE", "size": 1068, "type": "file" }, { "path": "README.md", "size": 6999, "type": "file" }, { "path": "examples", "size": 0, "type": "dir" }, { "path": "logilinux-driver", "size": 0, "type": "file" }, { "path": "logilinux", "size": 0, "type": "dir" }, { "path": "python", "size": 0, "type": "dir" }, { "path": "requirements.txt", "size": 31, "type": "file" }, { "path": "scripts", "size": 0, "type": "dir" }, { "path": "setup.py", "size": 2016, "type": "file" }, { "path": "src", "size": 0, "type": "dir" } ], "files_total_count": 25, "first_commit_date_default_branch": "2025-11-22T19:27:07Z", "forks": 1, "has_ci": false, "has_contributing": false, "has_docker": false, "has_docs": false, "has_license_file": true, "has_notebooks": false, "has_readme_file": true, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/logilinux/logilinux-sdk", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 0, "issues_open": 0, "issues_total": 0, "languages_top": [ { "language": "Python", "size": 38849 }, { "language": "C++", "size": 7459 }, { "language": "Shell", "size": 1776 } ], "last_commit_date_default_branch": "2025-11-22T23:46:48Z", "last_commit_oid_default_branch": "dc79207bb41e4805e05acdbbbb745d3b7a7144b7", "latest_release_date": null, "latest_release_tag": null, "license_name": "MIT License", "license_spdx": "MIT", "name_with_owner": "logilinux/logilinux-sdk", "owner": "logilinux", "parent_repo": null, "parent_url": null, "primary_language": "Python", "project_foreign_key": "lhp_7e3753aaa4e3a72c", "pull_requests_closed": 0, "pull_requests_merged": 0, "pull_requests_open": 0, "pull_requests_total": 0, "pushed_at": "2025-11-22T23:46:50Z", "readme_length": 6998, "readme_text": "# logilinux-sdk\n\nPythonic SDK for interfacing with Logitech Creator devices (MX Dialpad, MX Keypad) on Linux. Provides a clean plugin architecture inspired by the official C# SDK but with Python idioms.\n\n## Features\n\n- **Type-safe event system** - Dataclass-based events with full type hints\n- **Plugin architecture** - Create reusable plugins with Commands and Adjustments\n- **LCD display support** - Upload images to MX Keypad LCD buttons\n- **Clean lifecycle management** - Context managers and automatic cleanup\n- **Pythonic API** - Follows Python conventions while matching C# SDK naming\n\n## Installation\n\n### Prerequisites\n\n1. Build the C++ driver first:\n```bash\ncd logilinux-driver\n./build.sh\ncd ..\n```\n\n2. Install Python dependencies and the SDK:\n```bash\npip install -r requirements.txt\npip install -e .\n```\n\n3. Set library path and permissions (required for runtime):\n```bash\n# Option 1: Use helper script (handles sudo automatically)\n./scripts/run_with_lib.sh examples/simple_monitoring.py\n\n# Option 2: Manual with sudo\nexport LD_LIBRARY_PATH=$PWD/logilinux-driver/build/lib:$LD_LIBRARY_PATH\nsudo -E python examples/simple_monitoring.py\n\n# Option 3: Set up udev rules to avoid needing sudo (recommended)\nsudo bash -c 'cat > /etc/udev/rules.d/99-logitech-creative.rules << EOF\n# Logitech MX Creative Console - Dialpad\nSUBSYSTEM==\"hidraw\", ATTRS{idVendor}==\"046d\", ATTRS{idProduct}==\"bc00\", MODE=\"0666\"\n# Logitech MX Creative Console - Keypad \nSUBSYSTEM==\"hidraw\", ATTRS{idVendor}==\"046d\", ATTRS{idProduct}==\"c354\", MODE=\"0666\"\nEOF'\n\nsudo udevadm control --reload-rules\nsudo udevadm trigger\n\n# After udev rules, no sudo needed:\n./scripts/run_with_lib.sh examples/simple_monitoring.py\n```\n\n## Quick Start\n\n### Simple Event Monitoring\n\n```python\nfrom logilinux import Library, DeviceType, RotationEvent, ButtonEvent\n\nlib = Library()\ndevice = lib.find_device(DeviceType.DIALPAD)\n\ndef on_event(event):\n if isinstance(event, RotationEvent):\n print(f\"Dial rotated: {event.delta} steps\")\n elif isinstance(event, ButtonEvent):\n print(f\"Button {event.button_code} {'pressed' if event.pressed else 'released'}\")\n\ndevice.set_event_callback(on_event)\ndevice.start_monitoring()\n```\n\nRun with:\n```bash\n./scripts/run_with_lib.sh your_script.py\n```\n\n### Creating a Plugin\n\nThe SDK uses a plugin architecture matching the C# SDK patterns:\n\n```python\nfrom logilinux import Plugin, PluginCommand, PluginAdjustment\n\nclass MuteToggle(PluginCommand):\n \"\"\"Button command that toggles microphone mute.\"\"\"\n \n def __init__(self):\n super().__init__(\"Mic Toggle\", \"Mute/unmute microphone\", \"Audio\")\n self.is_muted = False\n \n def run_command(self, action_parameter=\"\"):\n self.is_muted = not self.is_muted\n # Perform actual mute action here\n self.action_image_changed() # Update button image\n \n def get_command_image(self, action_parameter=\"\", image_size=PluginImageSize.LARGE):\n # Return PIL Image for LCD button\n from PIL import Image, ImageDraw\n img = Image.new('RGB', (image_size.width, image_size.height))\n draw = ImageDraw.Draw(img)\n \n color = (200, 0, 0) if self.is_muted else (0, 200, 0)\n draw.rectangle([0, 0, image_size.width, image_size.height], fill=color)\n return img\n\n\nclass VolumeControl(PluginAdjustment):\n \"\"\"Dial adjustment that controls system volume.\"\"\"\n \n def __init__(self):\n super().__init__(\"Volume\", \"System volume\", \"Audio\", has_reset=True)\n self.volume = 50\n \n def apply_adjustment(self, action_parameter=\"\", diff=0):\n self.volume = max(0, min(100, self.volume + diff * 5))\n # Set system volume here\n self.adjustment_value_changed() # Update display\n \n def run_command(self, action_parameter=\"\"):\n # Reset to 50% when dial is pressed\n self.volume = 50\n self.adjustment_value_changed()\n \n def get_adjustment_value(self, action_parameter=\"\"):\n return f\"{self.volume}%\"\n\n\nclass MyPlugin(Plugin):\n \"\"\"Main plugin class.\"\"\"\n \n def load(self):\n # Initialize resources (API clients, timers, etc.)\n pass\n \n def unload(self):\n # Cleanup resources\n pass\n \n def get_commands(self):\n return [MuteToggle()]\n \n def get_adjustments(self):\n return [VolumeControl()]\n```\n\n### Running a Plugin\n\n```python\nfrom logilinux import PluginService, DeviceType\n\n# Create service and register plugin\nservice = PluginService()\nplugin = MyPlugin()\nservice.register_plugin(plugin)\n\n# Connect to devices\nkeypad = service.connect_device(DeviceType.MX_KEYPAD)\ndialpad = service.connect_device(DeviceType.DIALPAD)\n\n# Assign actions to buttons/dials\nmute_cmd = plugin.get_command('MuteToggle')\nservice.assign_button_action(keypad, 0, mute_cmd) # Grid button 0\n\nvolume_adj = plugin.get_adjustment('VolumeControl')\nservice.assign_dial_action(dialpad, volume_adj)\n\n# Run\nimport time\ntry:\n while True:\n time.sleep(0.1)\nexcept KeyboardInterrupt:\n service.shutdown()\n```\n\n## Architecture\n\nThe SDK follows a clean layered architecture:\n\n### C++ Bindings Layer (`_logilinux_native`)\n- Thin pybind11 wrapper over logilinux-driver\n- Direct 1:1 mapping to C++ API\n- No business logic\n\n### Core Layer (`logilinux.core`)\n- Pythonic wrappers: `Library`, `Device`, `MXKeypadDevice`\n- Type-safe events: `RotationEvent`, `ButtonEvent`, `DeviceEvent`\n- Context manager support for automatic cleanup\n\n### Plugin Layer (`logilinux.plugin`)\n- `Plugin` base class (matches C# SDK's Plugin)\n- `PluginService` for lifecycle management\n- Action routing and event handling\n\n### Actions Layer (`logilinux.actions`)\n- `PluginCommand` - Button press actions (matches C# SDK's PluginDynamicCommand)\n- `PluginAdjustment` - Dial rotation actions (matches C# SDK's PluginDynamicAdjustment)\n- Automatic state change notifications\n\n## API Overview\n\n### Core Classes\n\n- `Library()` - Main entry point for device discovery\n- `Device` - Base device interface\n- `MXKeypadDevice` - MX Keypad with LCD display support\n\n### Device Types\n\n- `DeviceType.DIALPAD` - MX Dialpad (dial + buttons)\n- `DeviceType.MX_KEYPAD` - MX Keypad (9 LCD buttons)\n\n### Events\n\n- `RotationEvent` - Dial/wheel rotation with delta and high-res info\n- `ButtonEvent` - Button press/release with button codes\n- `DeviceEvent` - Device connection/disconnection\n\n### Plugin System\n\n- `Plugin` - Base plugin class with load()/unload() lifecycle\n- `PluginCommand` - Button action with run_command() and get_command_image()\n- `PluginAdjustment` - Dial action with apply_adjustment() and get_adjustment_value()\n- `PluginService` - Service managing plugin lifecycle and device routing\n\n## Examples\n\nSee the `examples/` directory:\n\n- `simple_monitoring.py` - Basic event monitoring\n- `example_plugin.py` - Full plugin with commands and adjustments\n\n## Requirements\n\n- Python 3.7+\n- pybind11\n- Pillow (PIL)\n- Built logilinux C++ library (in logilinux-driver submodule)\n\n## License\n\nSee LICENSE file.", "readme_title": "logilinux-sdk", "releases_count": 0, "repo": "logilinux-sdk", "repo_name": "logilinux-sdk", "stars": 1, "topics": [], "updated_at": "2025-12-06T12:28:04Z", "url": "https://github.com/logilinux/logilinux-sdk", "watchers": 1 } ], "hackathon_location": "EPFL, Lausanne, Switzerland", "hackathon_name": "LauzHack 2025", "hackathon_year": 2025, "id": 1, "image_url": null, "project_fk": "lhp_7e3753aaa4e3a72c", "project_id": "1", "project_title": "logilinux", "project_uid": "id:1", "tags": [], "team": [ "Arjun Juneja", "Rudrrayan Manna" ], "title": "logilinux", "url": "https://github.com/logilinux/" }, { "awards": [ "2nd Place Overall" ], "categories": [ "2nd Place Overall" ], "description": "KnowTube is a learning companion that transforms YouTube videos into structured, active study materials. Instead of passively watching, learners can upload or link a video and instantly get a clean transcript, comprehension quizzes, and spaced‑repetition‑friendly flashcards generated from the key ideas in the content.\n\nBehind the scenes, KnowTube analyzes the video’s transcript to detect concepts, definitions, and relationships, then builds question sets and flashcards that encourage recall rath", "github_repo_count": 1, "github_repo_urls": [ "https://github.com/faresfawzi/kNOw-tube" ], "github_repos_metadata": [ { "commit_count_default_branch": 68, "contributors_count": 4, "contributors_top": [ { "contributions": 24, "html_url": "https://github.com/dominikglandorf", "login": "dominikglandorf" }, { "contributions": 23, "html_url": "https://github.com/abhinandshibu", "login": "abhinandshibu" }, { "contributions": 11, "html_url": "https://github.com/faresfawzi", "login": "faresfawzi" }, { "contributions": 10, "html_url": "https://github.com/spneshaei", "login": "spneshaei" } ], "created_at": "2025-11-22T09:06:28Z", "default_branch": "main", "description": null, "dirs_total_count": 23, "files_root_entries": [ { "path": ".gitignore", "size": 34, "type": "file" }, { "path": "KnowTubeDemo.mov", "size": 12270935, "type": "file" }, { "path": "README.md", "size": 923, "type": "file" }, { "path": "backend", "size": 0, "type": "dir" }, { "path": "docker-compose.yml", "size": 839, "type": "file" }, { "path": "frontend", "size": 0, "type": "dir" }, { "path": "kNOwtube-chrome-plugin", "size": 0, "type": "dir" }, { "path": "logitech-plugin", "size": 0, "type": "dir" }, { "path": "logo.pptx", "size": 33343, "type": "file" } ], "files_total_count": 117, "first_commit_date_default_branch": "2025-11-22T09:06:28Z", "forks": 0, "has_ci": false, "has_contributing": false, "has_docker": true, "has_docs": false, "has_license_file": false, "has_notebooks": false, "has_readme_file": true, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/faresfawzi/kNOw-tube", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 0, "issues_open": 0, "issues_total": 0, "languages_top": [ { "language": "Python", "size": 76656 }, { "language": "TypeScript", "size": 67031 }, { "language": "C#", "size": 17764 }, { "language": "Jupyter Notebook", "size": 5831 }, { "language": "HTML", "size": 2448 }, { "language": "CSS", "size": 2372 }, { "language": "JavaScript", "size": 2309 }, { "language": "Dockerfile", "size": 924 }, { "language": "Shell", "size": 114 } ], "last_commit_date_default_branch": "2025-11-28T17:18:37Z", "last_commit_oid_default_branch": "c888429665ce2318b419b12c328e0c7de9cafda1", "latest_release_date": null, "latest_release_tag": null, "license_name": null, "license_spdx": null, "name_with_owner": "faresfawzi/kNOw-tube", "owner": "faresfawzi", "parent_repo": null, "parent_url": null, "primary_language": "Python", "project_foreign_key": "lhp_7cab59a7c0895898", "pull_requests_closed": 0, "pull_requests_merged": 0, "pull_requests_open": 0, "pull_requests_total": 0, "pushed_at": "2025-11-28T17:19:07Z", "readme_length": 922, "readme_text": "# kNOw-tube\n\nYou can run in dev, with: `docker compose up --build`\n- hot reloading is on\n\nIf you install new dependencies, then need to rerun: \n1. `docker compose down`\n2. `docker compose up --build`\n\nTo use the extension in Chrome:\n1. Go to `chrome://extensions/`\n2. Enable \"Developer mode\" (top right)\n3. Click \"Load unpacked\"\n4. Select the `kNOwtube-chrome-plugin` folder\n5. The extension should now be loaded and ready to use!\n6. Use together client for the LLMs\n\n-- \n## Example for together\n\n``` \nresponse = client.chat.completions.create(\n model=model,\n messages=[\n {\"role\": \"user\", \"content\": prompt},\n ],\n temperature=temperature,\n # max_tokens=max_output_tokens,\n )\n```\n\n# Logitech plugin\n1. Install Logitech Options+ from https://www.logitech.com/de-ch/software/logi-options-plus.html\n2. open ./logitech-plugin/ExamplePlugin/Example.lplug4 with LogiPluginService", "readme_title": "kNOw-tube", "releases_count": 0, "repo": "kNOw-tube", "repo_name": "kNOw-tube", "stars": 1, "topics": [], "updated_at": "2026-03-03T13:56:18Z", "url": "https://github.com/faresfawzi/kNOw-tube", "watchers": 0 } ], "hackathon_location": "EPFL, Lausanne, Switzerland", "hackathon_name": "LauzHack 2025", "hackathon_year": 2025, "id": 2, "image_url": null, "project_fk": "lhp_7cab59a7c0895898", "project_id": "2", "project_title": "KnowTube", "project_uid": "id:2", "tags": [], "team": [ "Dominik Glandorf", "Fares Fawzi", "Seyed Parsa Neshaei", "Abhinand Shibu" ], "title": "KnowTube", "url": "https://github.com/faresfawzi/kNOw-tube/" }, { "awards": [ "3rd Place Overall", "Huawei Challenge 2nd Place" ], "categories": [ "3rd Place Overall", "Huawei Challenge 2nd Place" ], "description": "AI-powered presentation analysis tool that provides instant feedback on your speaking performance. Upload a video and get detailed insights on delivery, pacing, confidence, and areas to improve. Will also provide a generated audio giving an example of how you should ideally present.", "github_repo_count": 1, "github_repo_urls": [ "https://github.com/oscardef/showcAIse" ], "github_repos_metadata": [ { "commit_count_default_branch": 17, "contributors_count": 2, "contributors_top": [ { "contributions": 16, "html_url": "https://github.com/oscardef", "login": "oscardef" }, { "contributions": 1, "html_url": "https://github.com/kkorboe-epfl", "login": "kkorboe-epfl" } ], "created_at": "2025-11-22T12:08:47Z", "default_branch": "main", "description": "Project created at LauzHack 2025. A presentation assistant tool.", "dirs_total_count": 5, "files_root_entries": [ { "path": ".dockerignore", "size": 1723, "type": "file" }, { "path": ".gitignore", "size": 410, "type": "file" }, { "path": "Dockerfile.backend", "size": 2473, "type": "file" }, { "path": "Dockerfile.frontend", "size": 253, "type": "file" }, { "path": "IMPLEMENTATION_SUMMARY.md", "size": 8849, "type": "file" }, { "path": "README.md", "size": 13664, "type": "file" }, { "path": "TECHNICAL_SUMMARY.md", "size": 4018, "type": "file" }, { "path": "VOICE_CLONING_GUIDE.md", "size": 6167, "type": "file" }, { "path": "backend", "size": 0, "type": "dir" }, { "path": "docker-compose.yml", "size": 945, "type": "file" }, { "path": "frontend", "size": 0, "type": "dir" }, { "path": "test_voice_cloning.sh", "size": 3114, "type": "file" } ], "files_total_count": 33, "first_commit_date_default_branch": "2025-11-22T12:54:03Z", "forks": 0, "has_ci": false, "has_contributing": false, "has_docker": true, "has_docs": false, "has_license_file": false, "has_notebooks": false, "has_readme_file": true, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/oscardef/showcAIse", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 0, "issues_open": 0, "issues_total": 0, "languages_top": [ { "language": "JavaScript", "size": 81918 }, { "language": "Python", "size": 63208 }, { "language": "CSS", "size": 49120 }, { "language": "Shell", "size": 3114 }, { "language": "HTML", "size": 456 } ], "last_commit_date_default_branch": "2025-11-23T14:32:09Z", "last_commit_oid_default_branch": "edfaae813c69caad21fd5f33ec4fe1f0db93c66c", "latest_release_date": null, "latest_release_tag": null, "license_name": null, "license_spdx": null, "name_with_owner": "oscardef/showcAIse", "owner": "oscardef", "parent_repo": null, "parent_url": null, "primary_language": "JavaScript", "project_foreign_key": "lhp_1e65b245503245c2", "pull_requests_closed": 0, "pull_requests_merged": 0, "pull_requests_open": 0, "pull_requests_total": 0, "pushed_at": "2025-11-23T14:32:11Z", "readme_length": 13449, "readme_text": "# showcAIse 🎤\n\nAI-powered presentation analysis tool that provides instant feedback on your speaking performance. Upload a video and get detailed insights on delivery, pacing, confidence, and areas to improve.\n\n## ✨ Features\n\n- **Smart Analysis**: Detects strong moments and areas for improvement\n- **Confidence Scoring**: Advanced algorithm analyzing pace, fillers, sentiment, and language quality\n- **Segment-Specific Feedback**: Play and review individual moments with categorized insights\n- **🎤 Voice Cloning**: Generate improved presentations using your own voice (NEW!)\n- **Clean Professional UI**: Focused interface with dedicated Voice Clone tab\n- **Video Playback**: Segment-only player that isolates specific moments for focused review\n- **Actionable Recommendations**: Specific, prioritized suggestions for improvement\n\n## 🚀 Quick Start\n\n### Prerequisites\n- Docker Desktop ([Download](https://www.docker.com/products/docker-desktop))\n- Together AI API key ([Get one here](https://together.ai))\n- **Note**: Voice cloning requires ~2GB model download on first use (cached for future runs)\n\n### Setup & Run\n\n```bash\n# 1. Clone and navigate\ngit clone https://github.com/oscardef/showcAIse.git\ncd showcAIse\n\n# 2. Configure API key\ncd backend\ncp .env.example .env\n# Edit .env and add: TOGETHER_API_KEY=\"your-key-here\"\n\n# 3. Start with Docker\ncd ..\ndocker compose up --build\n```\n\n**That's it!** Visit http://localhost:3000\n\n### Docker Commands\n\n```bash\n# Start containers\ndocker compose up -d\n\n# View logs\ndocker compose logs -f\n\n# Stop containers\ndocker compose down\n\n# Rebuild after code changes\ndocker compose up --build\n\n# Pre-download TTS model (optional, for faster first voice clone)\ndocker compose exec backend python preload_tts.py\n```\n\n## 💻 Manual Setup (Without Docker)\n\n### Prerequisites\n- **Python 3.11.6** (Required for TTS library - 3.12+ not supported)\n- Node.js 18+\n- FFmpeg ([Install guide](https://ffmpeg.org/download.html))\n- ~2GB free space for TTS model\n\n### Backend Setup\n\n```bash\ncd backend\n\n# Use Python 3.11 (required for TTS)\npython3.11 -m venv venv\nsource venv/bin/activate # On Windows: venv\\Scripts\\activate\n\npip install -r requirements.txt\n\n# Configure API key\ncp .env.example .env\n# Edit .env: TOGETHER_API_KEY=\"your-key-here\"\n\n# Optional: Pre-download TTS model to avoid runtime delay\npython preload_tts.py\n\n# Run\npython main.py\n```\n\nBackend runs at http://localhost:8000\n\n### Frontend Setup\n\n```bash\ncd frontend\nnpm install\nnpm start\n```\n\nFrontend runs at http://localhost:3000\n\n## 📊 How It Works\n\n### Analysis Pipeline\n\n1. **Video Upload** → Extracts audio with FFmpeg\n2. **Transcription** → Together AI Whisper large-v3 model\n3. **Smart Segmentation** → Breaks transcript into natural speaking chunks\n4. **Confidence Scoring** → Analyzes each segment:\n - **Pacing** (±25 points): Optimal 130-160 WPM\n - **Filler Words** (±30 points): Strict thresholds (4%, 8%, 15%)\n - **Sentiment** (±20 points): Positive/negative/neutral tone\n - **Language Quality** (±15 points): Hedge words (\"I guess\", \"kind of\")\n5. **Moment Detection** → Identifies strong (≥70%) and weak (<50%) moments\n6. **Categorization** → Tags moments by type:\n - Strong: \"Confident language\", \"Clean delivery\", \"Perfect pacing\"\n - Weak: \"Excessive fillers\", \"Uncertain tone\", \"Poor pacing\"\n7. **Recommendations** → Generates specific, actionable improvement suggestions\n\n### Scoring System\n\n**Base Score**: 50 points (neutral starting point)\n\n**Adjustments:**\n- Good pacing (130-160 WPM): +25\n- Too fast/slow: -20\n- Minimal fillers (<4%): +30\n- Excessive fillers (>15%): -35\n- Positive sentiment: +20\n- Negative/uncertain tone: -25\n- Hedge words (>2): -15\n\n**Result**: 0-100 confidence score\n- **70-100**: Strong moment\n- **50-69**: Adequate\n- **0-49**: Needs improvement\n\n## 🎯 What You Get\n\n### Results Dashboard\n\n**1. Overview Tab**\n- Key metrics (words, WPM, fillers, duration)\n- Overall confidence score\n- Voice cloning quick-start button\n- Top 3 priority actions\n\n**2. Key Moments Tab**\n- Overview stats (performance score, strong/weak counts, duration)\n- **Strong Moments**: What made them effective with categories\n- **Weak Moments**: Specific issues + improvement suggestions\n- Segment-only video player (plays just that moment)\n\n**3. Sentiment Analysis Tab**\n- Overall sentiment and tone analysis\n- Sentiment trends (improving/declining/stable)\n- Actionable insights with severity levels\n- Negative moments to review\n- Best positive peaks\n\n**4. Delivery Metrics Tab**\n- Confidence calculation breakdown\n- Performance timeline chart\n- Weakest and strongest moments\n- Detailed speech analysis\n- Top filler words breakdown\n\n**5. Recommendations Tab**\n- Priority actions (top 3 critical improvements)\n- Additional suggestions with specific steps\n- Severity-based categorization\n\n**6. Voice Clone Tab** 🎤 (NEW!)\n- Generate improved presentation with your voice\n- Remove all filler words automatically\n- Replace uncertain language with confident phrasing\n- Maintain natural voice and speaking style\n- Download audio ready for video creation\n- View improved script with change summary\n- See metrics comparison (before/after)\n\n**7. Transcript Tab**\n- Full text with filler word highlighting\n- Easy reference for detailed review\n\n## 🎤 Voice Cloning Feature\n\n### How It Works\n\n1. **Upload & Analyze**: First, analyze your presentation video\n2. **Generate Clone**: Click \"Generate Improved Voice Clone\" button\n3. **AI Processing** (~1-2 minutes):\n - Extracts audio from your video\n - Analyzes transcript and generates improved script\n - Removes filler words (\"um\", \"uh\", \"like\", etc.)\n - Replaces uncertain language (\"I guess\" → \"I believe\")\n - Clones your voice using Coqui TTS XTTS v2 model\n - Generates clean audio output\n4. **Download**: Get WAV file ready for video creation\n\n### What Gets Improved\n\n**Removed:**\n- All filler words (um, uh, like, you know, so, actually, basically, literally)\n- Uncertain phrases (I think maybe, I guess, I don't know, kind of, sort of)\n\n**Replaced:**\n- \"I think maybe\" → \"I believe\"\n- \"I guess\" → \"I believe\"\n- \"probably\" → \"will\"\n- \"might be\" → \"is\"\n- \"could be\" → \"is\"\n- \"maybe\" → \"will\"\n\n**Maintained:**\n- Your natural voice characteristics\n- Speaking rhythm and cadence\n- Emotional tone\n- Core message and content\n\n### Technical Details\n\n**Model**: Coqui TTS XTTS v2\n- Multilingual voice cloning\n- High-quality synthesis\n- ~2GB model size\n- CPU-optimized (no GPU required)\n\n**Performance**:\n- First run: 2-3 minutes (includes model download)\n- Subsequent runs: 1-2 minutes (model cached)\n- Output: WAV format, ready for video\n\n**Caching**:\n- Docker: Persistent volumes (`tts_cache`)\n- Manual: `~/.local/share/tts/`\n- Model downloaded once, reused forever\n\n### Requirements\n\n- Python 3.11.6 (TTS doesn't support 3.12+)\n- ~2GB disk space for model\n- 2-4GB RAM during generation\n- Audio track in video (mono/stereo OK)\n\n### Troubleshooting Voice Cloning\n\n**\"Voice cloning failed\"**\n```bash\n# Check if video has audio\ndocker compose exec backend python -c \"from moviepy.editor import VideoFileClip; v = VideoFileClip('videos/YOUR_SESSION_ID.mp4'); print(v.audio)\"\n\n# Check TTS cache\ndocker compose exec backend ls -lh /root/.local/share/tts/\n\n# Re-download model if corrupted\ndocker compose exec backend rm -rf /root/.local/share/tts/\ndocker compose exec backend python preload_tts.py\n```\n\n**\"Model download too slow\"**\n```bash\n# Pre-download before first use\ndocker compose exec backend python preload_tts.py\n\n# Or during build (edit Dockerfile.backend, uncomment line):\n# RUN python preload_tts.py\n```\n\n**\"Out of memory\"**\n- Close other applications\n- Increase Docker memory limit (Docker Desktop → Settings → Resources)\n- Try shorter videos (<5 minutes)\n\n## 🏗️ Tech Stack\n\n**Backend**\n- FastAPI 0.121.3 (Python 3.11.6 for TTS compatibility)\n- Together AI Whisper API (transcription)\n- Coqui TTS XTTS v2 (voice cloning)\n- Transformers 4.33.0 (DistilBERT sentiment)\n- MoviePy 1.0.3 (audio extraction)\n- FFmpeg (media processing)\n\n**Frontend**\n- React 18.2\n- Clean, professional UI (no rounded corners, minimal nesting)\n- Segment-isolated video player\n\n**Deployment**\n- Docker & Docker Compose\n- Multi-container architecture\n- Hot reload for development\n\n## 📁 Project Structure\n\n```\nshowcAIse/\n├── backend/\n│ ├── main.py # FastAPI server\n│ ├── analyzer.py # Core analysis engine\n│ └── requirements.txt\n├── frontend/\n│ ├── src/\n│ │ ├── App.js # Main app component\n│ │ ├── Upload.js # Video upload interface\n│ │ ├── ResultsClean.js # Results dashboard\n│ │ ├── MomentsAnalysis.js # Moments display\n│ │ └── clean.css # Professional styling\n│ └── package.json\n├── docker-compose.yml # Container orchestration\n├── Dockerfile.backend\n└── Dockerfile.frontend\n```\n\n## 🔧 API Reference\n\n### POST /api/upload\nUpload video for analysis\n\n**Request**: `multipart/form-data` with video file\n\n**Response**:\n```json\n{\n \"session_id\": \"abc-123\",\n \"transcript\": \"Full transcription...\",\n \"word_count\": 150,\n \"wpm\": 145,\n \"duration\": 62.4,\n \"overall_confidence\": 72,\n \"key_clips\": {\n \"strong_moments\": [\n {\n \"segment_num\": 3,\n \"confidence\": 85,\n \"start_time\": 15,\n \"end_time\": 32,\n \"text\": \"Segment text...\",\n \"categories\": [\"Confident language\", \"Perfect pacing\"],\n \"metrics\": {\"wpm\": 145, \"fillers\": 0, \"sentiment\": 0.85}\n }\n ],\n \"weak_moments\": [\n {\n \"segment_num\": 7,\n \"confidence\": 42,\n \"issues\": [\"Excessive filler words\", \"Uncertain tone\"],\n \"suggestions\": [\"Reduce fillers by pausing...\", \"Use more confident language...\"],\n \"metrics\": {\"wpm\": 98, \"fillers\": 8, \"sentiment\": -0.45}\n }\n ]\n },\n \"recommendations\": [...],\n \"timeline\": [...]\n}\n```\n\n### POST /api/voice-clone/{session_id}\nGenerate improved presentation with voice cloning\n\n**Request**: POST to `/api/voice-clone/{session_id}` (no body required)\n\n**Response**:\n```json\n{\n \"status\": \"success\",\n \"audio_url\": \"/api/cloned-audio/{session_id}\",\n \"improved_script\": \"Cleaned and improved transcript...\",\n \"improvements\": {\n \"improvements\": [\n \"Removed 15 filler words\",\n \"Reduced script by 23 words (8.2%)\",\n \"Replaced uncertain language with confident phrasing\",\n \"Optimized sentence structure for clarity\"\n ],\n \"original_word_count\": 280,\n \"improved_word_count\": 257,\n \"original_wpm\": 167,\n \"target_wpm\": 145,\n \"estimated_duration_seconds\": 106.3\n }\n}\n```\n\n### GET /api/cloned-audio/{session_id}\nDownload cloned audio file\n\n**Response**: WAV audio file (audio/wav)\n\n## 🐛 Troubleshooting\n\n### Docker Issues\n\n```bash\n# Port conflicts (3000 or 8000 in use)\ndocker compose down\nlsof -ti:3000 -ti:8000 | xargs kill -9\ndocker compose up -d\n\n# Rebuild from scratch\ndocker compose down\ndocker compose up --build\n\n# Clean everything\ndocker system prune -a\n```\n\n### API Key Issues\n\n1. Verify key is correct (no extra spaces/newlines)\n2. Check `.env` file is in `backend/` directory\n3. Restart containers after changing `.env`:\n ```bash\n docker compose down\n docker compose up -d\n ```\n\n### FFmpeg Missing (Manual Setup)\n\n```bash\n# macOS\nbrew install ffmpeg\n\n# Ubuntu/Debian\nsudo apt-get install ffmpeg\n\n# Windows\n# Download from https://ffmpeg.org/download.html\n```\n\n### Module Errors\n\n```bash\n# Backend\ncd backend\npip install -r requirements.txt\n\n# Frontend\ncd frontend\nrm -rf node_modules\nnpm install\n```\n\n## 🎓 Algorithm Details\n\n### Timestamp Accuracy\n\nUses time-based calculation (0.4 seconds per word) instead of ratio-based positioning. Adds 1-second buffer for accuracy.\n\n```python\nstart_second = int((cumulative_words * 0.4)) - 1\nend_second = int((cumulative_words * 0.4)) + 1\n```\n\n### Segment-Only Video Player\n\n- Automatically starts at segment start time\n- Pauses at segment end time\n- Prevents seeking outside segment boundaries\n- Visual indicator shows segment timing\n- Loops back to start when finished\n\n### Hedge Word Detection\n\nIdentifies uncertain language:\n- \"kind of\", \"sort of\"\n- \"I guess\", \"I don't know\"\n- \"maybe\", \"probably\"\n- Penalty: -15 points if >2 occurrences\n\n### Filler Detection\n\nTracks common fillers:\n- \"um\", \"uh\", \"like\", \"you know\"\n- \"so\", \"actually\", \"basically\", \"literally\"\n\nThresholds:\n- <4%: No penalty (natural)\n- 4-8%: -10 points (mild)\n- 8-15%: -25 points (moderate)\n- >15%: -35 points (severe)\n\n## 📝 Development Notes\n\n### Recent Improvements (v2.1)\n\n**Algorithm Accuracy**\n- Changed confidence base from 70→50 (more realistic)\n- Added sentiment integration (±20 points)\n- Stricter filler thresholds\n- Hedge word detection\n\n**UI Redesign**\n- Reduced from 6 tabs → 3 tabs\n- Removed all rounded corners (flat professional design)\n- Eliminated card nesting\n- Prominent upload header\n- Segment-focused video player\n\n**Timestamp Fix**\n- Switched to time-based calculation\n- Added accuracy buffers\n- Segment-only playback with boundaries\n\n### Known Limitations\n\n1. Sentiment analysis per segment (may be slow for long videos)\n2. Hedge word detection is regex-based (context-independent)\n3. Timestamps accurate within ±1 second\n\n### Future Enhancements\n\n- Caching for sentiment analysis\n- Context-aware hedge word detection\n- User feedback loop for threshold tuning\n- Body language analysis (computer vision)\n- Comparative analytics (track improvement over time)\n\n---\n\n**Built with ❤️ for better presentations**\n\nVisit http://localhost:3000 to get started!", "readme_title": "showcAIse 🎤", "releases_count": 0, "repo": "showcAIse", "repo_name": "showcAIse", "stars": 0, "topics": [], "updated_at": "2025-11-23T14:32:14Z", "url": "https://github.com/oscardef/showcAIse", "watchers": 0 } ], "hackathon_location": "EPFL, Lausanne, Switzerland", "hackathon_name": "LauzHack 2025", "hackathon_year": 2025, "id": 3, "image_url": null, "project_fk": "lhp_1e65b245503245c2", "project_id": "3", "project_title": "showcAIse", "project_uid": "id:3", "tags": [], "team": [ "Fatumah Binta Sakho Doukouré", "Oscar de Francesca", "Ismail Nour", "Kristyn Dede Korboe" ], "title": "showcAIse", "url": "https://github.com/oscardef/showcAIse" }, { "awards": [ "Organizers' prize" ], "categories": [ "Organizers' prize" ], "description": "Birdy uses your webcam to monitor posture and detect hand gestures in real time. It runs in the background, alerts you when you slouch, and lets you control actions using simple gestures. Designed to improve comfort, reduce strain, and enhance productivity while you work.", "github_repo_count": 1, "github_repo_urls": [ "https://github.com/FacerOfGod/birdy" ], "github_repos_metadata": [ { "commit_count_default_branch": 12, "contributors_count": 1, "contributors_top": [ { "contributions": 12, "html_url": "https://github.com/FacerOfGod", "login": "FacerOfGod" } ], "created_at": "2025-11-22T22:38:16Z", "default_branch": "main", "description": "Birdy uses your webcam to monitor posture and detect hand gestures in real time. It runs in the background, alerts you when you slouch, and lets you control actions using simple gestures. Designed to improve comfort, reduce strain, and enhance productivity while you work.", "dirs_total_count": 5, "files_root_entries": [ { "path": ".gitignore", "size": 54, "type": "file" }, { "path": "README.md", "size": 5546, "type": "file" }, { "path": "assets", "size": 0, "type": "dir" }, { "path": "css", "size": 0, "type": "dir" }, { "path": "index.html", "size": 16229, "type": "file" }, { "path": "js", "size": 0, "type": "dir" }, { "path": "main.js", "size": 22702, "type": "file" }, { "path": "package-lock.json", "size": 29399, "type": "file" }, { "path": "package.json", "size": 1363, "type": "file" }, { "path": "server.py", "size": 6233, "type": "file" }, { "path": "server.spec", "size": 660, "type": "file" }, { "path": "splash.html", "size": 9368, "type": "file" } ], "files_total_count": 33, "first_commit_date_default_branch": "2025-11-22T22:49:18Z", "forks": 0, "has_ci": false, "has_contributing": false, "has_docker": false, "has_docs": false, "has_license_file": false, "has_notebooks": false, "has_readme_file": true, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/FacerOfGod/birdy", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 0, "issues_open": 0, "issues_total": 0, "languages_top": [ { "language": "JavaScript", "size": 94154 }, { "language": "HTML", "size": 25597 }, { "language": "CSS", "size": 22828 }, { "language": "Python", "size": 6893 } ], "last_commit_date_default_branch": "2025-12-07T17:40:14Z", "last_commit_oid_default_branch": "fb5fe5edbc80aec1d5f99c5a227a67faadcdf33f", "latest_release_date": null, "latest_release_tag": null, "license_name": null, "license_spdx": null, "name_with_owner": "FacerOfGod/birdy", "owner": "FacerOfGod", "parent_repo": null, "parent_url": null, "primary_language": "JavaScript", "project_foreign_key": "lhp_0f317f06405544aa", "pull_requests_closed": 0, "pull_requests_merged": 0, "pull_requests_open": 0, "pull_requests_total": 0, "pushed_at": "2025-12-07T17:40:23Z", "readme_length": 5365, "readme_text": "# Birdy - AI Posture & Gesture Control\n\nBirdy is an intelligent desktop assistant that uses your webcam to monitor your posture and control your computer with hand gestures. Built with Electron, MediaPipe, and Python\n\n## How a project born out of giving up ended up winning a prize\n\nThis project was built for [Lauzhack EPFL 2025](https://lauzhack.com/). A huge room full of people who look like they’ve been preparing for this moment for weeks. Teams of four forming battle plans, whiteboards covered with diagrams, AI agents everywhere.\n\nMe?\n_A few hours in, and I was already lost._\n\nMy team and I originally planned to attack the SBB challenge… until we realized none of us were actually excited about it. Then came hours of brainstorming, trying to chase an idea that felt meaningful, original, or at least buildable. Nothing clicked. Every new idea died within minutes.\n\nSo eventually, we did the unthinkable:\n_We gave up._\n\nBut here’s the thing, once we freed ourselves from the pressure to “produce something impressive,” Lauzhack suddenly became fun again. I decided to stay anyway, just to enjoy the atmosphere. Hackathons have this special kind of energy that keeps you coding even when you have no idea why.\n\nSo I started hopping around GitHub, poking into random repositories like someone browsing channels with no plan at 3 a.m. And then I stumbled on a project showing how AI/ML could detect body posture and head position. That caught my eye.\n\n_“Why not? Let’s try it,” I thought._\n\nOne experiment became two. Then ten. Suddenly I was building things with no intention of submitting anything. Just pure curiosity.\n\nAnd then it hit me.\n_“This… actually has potential.”_\n\nThe submission deadline was midnight. At 23:55, I was still debating with myself. Should I submit? Should I pretend this isn’t happening?\n\nA teammate who was also deep in his own fun side-project just looked at me and said:\n_“Dude, just sign up. You can always dip later if you don’t want to present.”_\n\nSo at 23:59, literally one minute before the deadline, I hit submit.\n\nAnd that was it. Suddenly I was motivated again. Not because I had planned it, not because I felt ready, but because Birdy had become something that mattered, not only to me, but potentially to anyone who sits in front of a computer all day.\n\n## Example \n\n\"image\"\n\n## Features\n\n### 🧘 Posture Monitoring\n* **Real-time Analysis**: Detects slouching, leaning, and distance from the screen.\n* **Smart Alerts**: Visual notifications when you need to sit up straight or take a break.\n* **Standing Detection**: Automatically detects when you stand up.\n\n### ⏱️ Smart Timer\n* **Sitting Timer**: Tracks how long you've been sitting.\n* **Auto-Pause/Resume**: The timer automatically stops when you stand up or leave the desk, and resumes when you sit back down.\n* **Break Reminders**: Reminds you to stretch after prolonged sitting.\n\n### 🖐️ Gesture Control\nControl your computer without a mouse using intuitive hand gestures.\n\n| Action | Gesture | Description |\n| :--- | :--- | :--- |\n| **Toggle Cursor** | **Fist → Peace** | Make a fist, then a peace sign to activate/deactivate the virtual cursor. |\n| **Move Cursor** | **Peace Sign** | Move your hand while holding a peace sign to move the cursor. |\n| **Click** | **Pinch** | Pinch your thumb and index finger quickly to click. |\n| **Drag** | **Pinch (Hold)** | Pinch and hold to drag items. Release to drop. |\n| **Switch Desktop** | **Swipe** | Swipe left or right with an open hand (when cursor is inactive). |\n| **Task View** | **Fist → Open Hand** | Hold a fist, then open your hand to toggle Task View. |\n| **Volume Control** | **Index Finger** | Place your index finger near your ear and move up/down to adjust volume. |\n\n### 🖥️ Compact Mode\n* Switch to a minimalist \"Compact Mode\" to keep Birdy unobtrusive while you work.\n* Always-on-top floating window with essential controls.\n\n## Prerequisites\n\n1. **Node.js**: [Download & Install](https://nodejs.org/)\n2. **Python 3.x**: [Download & Install](https://www.python.org/)\n\n## Installation\n\n1. **Clone the repository**:\n ```bash\n git clone https://github.com/yourusername/birdy.git\n cd birdy\n ```\n\n2. **Install Node.js dependencies**:\n ```bash\n npm install\n ```\n\n3. **Install Python dependencies**:\n ```bash\n pip install websockets pyautogui pycaw comtypes\n ```\n\n## Usage\n\nBirdy requires both the Python server (for system control) and the Electron app to be running.\n\n1. **Start the Python Server**:\n Open a terminal in the project directory and run:\n ```bash\n python server.py\n ```\n *Keep this terminal open.*\n\n2. **Start the Application**:\n Open a new terminal in the project directory and run:\n ```bash\n npm start\n ```\n\n3. **Calibration**:\n * Sit in your normal working posture.\n * Click the **\"Calibrate\"** button.\n * Birdy will now monitor your posture relative to this baseline.\n\n## Troubleshooting\n\n* **\"Server not connected\"**: Ensure `server.py` is running in a separate terminal.\n* **Gestures not working**: Make sure your hand is clearly visible to the camera. Good lighting helps!\n* **Cursor drifting**: Recalibrate by toggling the cursor off and on (Fist -> Peace).", "readme_title": "# Birdy - AI Posture & Gesture Control", "releases_count": 0, "repo": "birdy", "repo_name": "birdy", "stars": 4, "topics": [], "updated_at": "2026-01-11T16:51:29Z", "url": "https://github.com/FacerOfGod/birdy", "watchers": 0 } ], "hackathon_location": "EPFL, Lausanne, Switzerland", "hackathon_name": "LauzHack 2025", "hackathon_year": 2025, "id": 4, "image_url": null, "project_fk": "lhp_0f317f06405544aa", "project_id": "4", "project_title": "Birdy", "project_uid": "id:4", "tags": [], "team": [ "Sylvain Pichot" ], "title": "Birdy", "url": "https://github.com/FacerOfGod/birdy" }, { "awards": [ "Huawei Challenge 1st Place" ], "categories": [ "Huawei Challenge 1st Place" ], "description": "Mobile app which provides real-time obstacle detection and environmental awareness to enhance independence for visually impaired individuals.", "github_repo_count": 1, "github_repo_urls": [ "https://github.com/MarognaLorenzo/lauzhack" ], "github_repos_metadata": [ { "commit_count_default_branch": 19, "contributors_count": 2, "contributors_top": [ { "contributions": 15, "html_url": "https://github.com/MarognaLorenzo", "login": "MarognaLorenzo" }, { "contributions": 4, "html_url": null, "login": "marelaG" } ], "created_at": "2025-11-22T14:07:20Z", "default_branch": "main", "description": null, "dirs_total_count": 36, "files_root_entries": [ { "path": ".gitignore", "size": 225, "type": "file" }, { "path": ".idea", "size": 0, "type": "dir" }, { "path": "README.md", "size": 1943, "type": "file" }, { "path": "app", "size": 0, "type": "dir" }, { "path": "build.gradle.kts", "size": 269, "type": "file" }, { "path": "gradle.properties", "size": 1346, "type": "file" }, { "path": "gradle", "size": 0, "type": "dir" }, { "path": "gradlew", "size": 5766, "type": "file" }, { "path": "gradlew.bat", "size": 2674, "type": "file" }, { "path": "settings.gradle.kts", "size": 533, "type": "file" } ], "files_total_count": 71, "first_commit_date_default_branch": "2025-11-22T14:13:35Z", "forks": 1, "has_ci": false, "has_contributing": false, "has_docker": false, "has_docs": false, "has_license_file": false, "has_notebooks": false, "has_readme_file": true, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/MarognaLorenzo/lauzhack", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 0, "issues_open": 0, "issues_total": 0, "languages_top": [ { "language": "Kotlin", "size": 37293 } ], "last_commit_date_default_branch": "2025-11-25T08:39:09Z", "last_commit_oid_default_branch": "ad5386da24a5921cb4c42d8565b23a4dd429ecc1", "latest_release_date": null, "latest_release_tag": null, "license_name": null, "license_spdx": null, "name_with_owner": "MarognaLorenzo/lauzhack", "owner": "MarognaLorenzo", "parent_repo": null, "parent_url": null, "primary_language": "Kotlin", "project_foreign_key": "lhp_3a3b3b91929ca019", "pull_requests_closed": 0, "pull_requests_merged": 0, "pull_requests_open": 0, "pull_requests_total": 0, "pushed_at": "2025-11-25T08:39:33Z", "readme_length": 1943, "readme_text": "# PawPilot - Your AI Assistant\n\nWelcome to PawPilot, an intelligent Android application developed for LauzHack. Pilot is designed to be your personal AI companion, using computer vision and text to speech to understand your environment and help you navigate it.\n\n## Key Features\n\n- **Visual Analysis:** Pilot can periodically capture images from your device's camera to analyze and understand your surroundings.\n- **Intuitive & Responsive UI:** The app features a clean, minimalistic interface with a central pulsating button that visually communicates its current state: inactive (gray), capturing images (blue).\n- **Haptic Feedback:** Enjoy subtle vibrations that provide tactile confirmation of your actions, making the experience more intuitive.\n\n## How to Use\n\n1. **Start/Stop Visual Analysis:** Tap the central button to begin or end the visual analysis. When active, the button will turn blue and pulsate.\n3. **View Transcription:** The recorded audio will be automatically transcribed, and the resulting text will be read to the user outloud.\n\n## Setup for Developers\n\n1. **Clone the Repository:**\n ```bash\n git clone \n ```\n2. **Open in Android Studio:** Open the cloned project in Android Studio.\n3. **Add Your API Key:** To enable audio transcription, you'll need to add your Together AI API key. Open the `MainActivity.kt` file and replace the placeholder with your actual key:\n ```kotlin\n // In app/src/main/java/com/example/lauzhack/MainActivity.kt\n \n // ... inside the handleLongPressEnd function\n val text = TogetherTranscription.transcribeAudio(\n filePath = outputFile,\n apiKey = \"YOUR_API_KEY_HERE\" // <-- Add your key here\n )\n ```\n4. **Build and Run:** Build and run the app on an Android device or emulator.\n\n---\n\n*This project was proudly created during the LauzHack event, showcasing the power of combining modern Android development with cutting-edge AI.*", "readme_title": "PawPilot - Your AI Assistant", "releases_count": 0, "repo": "lauzhack", "repo_name": "lauzhack", "stars": 0, "topics": [], "updated_at": "2025-11-25T08:39:36Z", "url": "https://github.com/MarognaLorenzo/lauzhack", "watchers": 0 } ], "hackathon_location": "EPFL, Lausanne, Switzerland", "hackathon_name": "LauzHack 2025", "hackathon_year": 2025, "id": 5, "image_url": null, "project_fk": "lhp_3a3b3b91929ca019", "project_id": "5", "project_title": "PawPilot", "project_uid": "id:5", "tags": [], "team": [ "Lorenzo Marogna", "Marcelina Gancewska", "Wissam Pheng" ], "title": "PawPilot", "url": "https://github.com/MarognaLorenzo/lauzhack" }, { "awards": [ "Huawei Challenge 3rd Place" ], "categories": [ "Huawei Challenge 3rd Place" ], "description": "Improving inference of LLM by doing better KV caching", "github_repo_count": 1, "github_repo_urls": [ "https://github.com/reinattwijaya/Sparse-Attention-Zoo" ], "github_repos_metadata": [ { "commit_count_default_branch": 56, "contributors_count": 4, "contributors_top": [ { "contributions": 50, "html_url": "https://github.com/andresnowak", "login": "andresnowak" }, { "contributions": 3, "html_url": null, "login": "sayantan0013" }, { "contributions": 2, "html_url": null, "login": "sathvikb" }, { "contributions": 1, "html_url": null, "login": "reinattwijaya" } ], "created_at": "2025-11-22T16:10:51Z", "default_branch": "main", "description": null, "dirs_total_count": 3, "files_root_entries": [ { "path": ".gitignore", "size": 193, "type": "file" }, { "path": ".python-version", "size": 5, "type": "file" }, { "path": "LICENSE", "size": 1069, "type": "file" }, { "path": "README.md", "size": 2839, "type": "file" }, { "path": "configs", "size": 0, "type": "dir" }, { "path": "datatrove", "size": 0, "type": "dir" }, { "path": "eval.py", "size": 16971, "type": "file" }, { "path": "eval_govreport.py", "size": 10550, "type": "file" }, { "path": "infer.py", "size": 2218, "type": "file" }, { "path": "main.py", "size": 15247, "type": "file" }, { "path": "main_inference.py", "size": 888, "type": "file" }, { "path": "pyproject.toml", "size": 1198, "type": "file" }, { "path": "src", "size": 0, "type": "dir" }, { "path": "tokenizer_dataset.sh", "size": 457, "type": "file" }, { "path": "train.sh", "size": 4568, "type": "file" }, { "path": "train_two_stage.sh", "size": 5696, "type": "file" }, { "path": "uv.lock", "size": 144739, "type": "file" } ], "files_total_count": 28, "first_commit_date_default_branch": "2025-10-09T18:38:07Z", "forks": 0, "has_ci": false, "has_contributing": false, "has_docker": false, "has_docs": false, "has_license_file": true, "has_notebooks": false, "has_readme_file": true, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/reinattwijaya/Sparse-Attention-Zoo", "is_archived": false, "is_fork": true, "is_private": false, "issues_closed": 0, "issues_open": 0, "issues_total": 0, "languages_top": [ { "language": "Python", "size": 107165 }, { "language": "Shell", "size": 10721 } ], "last_commit_date_default_branch": "2025-11-23T09:19:09Z", "last_commit_oid_default_branch": "3e2329cbbcf1384f939bacb4eaf7c76917a72ac5", "latest_release_date": null, "latest_release_tag": null, "license_name": "MIT License", "license_spdx": "MIT", "name_with_owner": "reinattwijaya/Sparse-Attention-Zoo", "owner": "reinattwijaya", "parent_repo": "andresnowak/Sparse-Attention-Zoo", "parent_url": "https://github.com/andresnowak/Sparse-Attention-Zoo", "primary_language": "Python", "project_foreign_key": "lhp_b8a47a79858bbbe6", "pull_requests_closed": 0, "pull_requests_merged": 0, "pull_requests_open": 0, "pull_requests_total": 0, "pushed_at": "2025-11-23T09:19:40Z", "readme_length": 2839, "readme_text": "# Sparse Attention Zoo\n\n## Currently Implemented\n\n### DeepSeek-V3.2 Dynamic Sparse Attention (DSA)\n\nImplementation of the Dynamic Sparse Attention mechanism from [DeepSeek-V3.2](https://github.com/deepseek-ai/DeepSeek-V3.2-Exp/blob/main/DeepSeek_V3_2.pdf).\n\n**Key Features:**\n- **Lightning Indexer**: Lightweight attention indexer that selects relevant tokens\n- **Two-Stage Training**:\n - **Dense Warm-up Stage**: Train only the indexer with frozen main model using dense attention\n - **Sparse Training Stage**: Train both indexer and main model with sparse attention pattern\n- **Efficient Token Selection**: Uses top-k selection from indexer scores\n- **KL Divergence Alignment**: Aligns indexer distribution with main attention distribution\n\n**Architecture:**\n- Separate indexer network per attention layer\n- Partial RoPE application for position encoding\n- Detached indexer input from main computational graph\n- Multi-head indexer with configurable heads and dimensions\n\n## Installation\n\n```bash\nuv sync\n```\n\nand NGC 25.06 pytorch container\n\n## Usage\n\n### Key Arguments\n\n**Model Configuration:**\n- `--model_name`: Base LLaMA model to use\n- `--index_top_k`: Number of tokens to select per query (default: 2048)\n- `--index_num_heads`: Number of indexer heads (default: 16)\n- `--rope_head_dim`: Dimension for RoPE in indexer (default: 32)\n- `--index_head_dim`: Head dimension for indexer (default: 64)\n\n**Training Configuration:**\n- `--warmup_stage`: Enable dense warm-up stage (freeze main model)\n- `--batch_size`: Training batch size\n- `--learning_rate`: Learning rate\n- `--num_epochs`: Number of training epochs\n- `--gradient_accumulation_steps`: Gradient accumulation steps\n\n**Data Configuration:**\n- `--dataset_name`: HuggingFace dataset name\n- `--dataset_config`: Dataset configuration\n- `--max_train_samples`: Maximum number of training samples\n- `--dataset_offset`: Offset for dataset samples\n\n**Logging:**\n- `--wandb_project`: W&B project name\n- `--wandb_run_name`: W&B run name\n- `--log_every`: Logging frequency (steps)\n\n## Architecture Details\n\n### Indexer Network\n- Query projection: `hidden_size, num_heads * head_dim`\n- Key projection: `hidden_size, head_dim`\n- Weight projection: `hidden_size, num_heads`\n- Partial RoPE on first `rope_head_dim` dimensions\n- LayerNorm on keys\n\n### Loss Functions\n- **Warm-up Stage**: KL divergence between indexer and aggregated attention scores\n- **Sparse Stage**:\n - Main model: Cross-entropy language modeling loss\n - Indexer: KL divergence on selected top-k tokens only\n\n## References\n\n- [DeepSeek-V3.2 Paper](https://github.com/deepseek-ai/DeepSeek-V3.2-Exp/blob/main/DeepSeek_V3_2.pdf)\n- DeepSeek-V3.2: \"We first use a short warm-up stage to initialize the lightning indexer. In this stage, we keep dense attention and freeze all model parameters except for the lightning indexer.\"", "readme_title": "Sparse Attention Zoo", "releases_count": 0, "repo": "Sparse-Attention-Zoo", "repo_name": "Sparse-Attention-Zoo", "stars": 0, "topics": [], "updated_at": "2025-11-23T09:19:44Z", "url": "https://github.com/reinattwijaya/Sparse-Attention-Zoo", "watchers": 0 } ], "hackathon_location": "EPFL, Lausanne, Switzerland", "hackathon_name": "LauzHack 2025", "hackathon_year": 2025, "id": 6, "image_url": null, "project_fk": "lhp_b8a47a79858bbbe6", "project_id": "6", "project_title": "LLM Inference Optimizer", "project_uid": "id:6", "tags": [], "team": [ "Reinatt Hansel Wijaya", "Sathvik Bhagavan" ], "title": "LLM Inference Optimizer", "url": "https://github.com/reinattwijaya/Sparse-Attention-Zoo" }, { "awards": [ "Logitech Challenge Winner" ], "categories": [ "Logitech Challenge Winner" ], "description": "Using haptic feedback in different ways to help visually impaired people navigate and find buttons, input fields, etc. There is a large market segment that we can sell mice to and thus capture new users.", "github_repo_count": 1, "github_repo_urls": [ "https://github.com/arthurbrunborg/MXSensePlugin" ], "github_repos_metadata": [ { "commit_count_default_branch": 16, "contributors_count": 2, "contributors_top": [ { "contributions": 9, "html_url": "https://github.com/arthurbrunborg", "login": "arthurbrunborg" }, { "contributions": 7, "html_url": "https://github.com/BhuvanVankadari", "login": "BhuvanVankadari" } ], "created_at": "2025-11-22T16:20:19Z", "default_branch": "main", "description": "MX Sense LauzHack", "dirs_total_count": 8, "files_root_entries": [ { "path": ".gitignore", "size": 6870, "type": "file" }, { "path": ".vscode", "size": 0, "type": "dir" }, { "path": "ACCESSIBILITY_README.md", "size": 4751, "type": "file" }, { "path": "HAPTIC_SETTINGS_GUIDE.md", "size": 5006, "type": "file" }, { "path": "IMPLEMENTATION_SUMMARY.md", "size": 4788, "type": "file" }, { "path": "MXSensePlugin.sln", "size": 1396, "type": "file" }, { "path": "TEXT_TO_SPEECH_GUIDE.md", "size": 5134, "type": "file" }, { "path": "src", "size": 0, "type": "dir" } ], "files_total_count": 28, "first_commit_date_default_branch": "2025-11-22T17:29:24Z", "forks": 0, "has_ci": false, "has_contributing": false, "has_docker": false, "has_docs": false, "has_license_file": false, "has_notebooks": false, "has_readme_file": false, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/arthurbrunborg/MXSensePlugin", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 0, "issues_open": 0, "issues_total": 0, "languages_top": [ { "language": "C#", "size": 67477 } ], "last_commit_date_default_branch": "2025-11-23T10:26:00Z", "last_commit_oid_default_branch": "25d2abc7ca6f1f2eb4fad28a8182a6d1ac22d927", "latest_release_date": null, "latest_release_tag": null, "license_name": null, "license_spdx": null, "name_with_owner": "arthurbrunborg/MXSensePlugin", "owner": "arthurbrunborg", "parent_repo": null, "parent_url": null, "primary_language": "C#", "project_foreign_key": "lhp_aef1e68e2c0abd86", "pull_requests_closed": 0, "pull_requests_merged": 0, "pull_requests_open": 0, "pull_requests_total": 0, "pushed_at": "2025-11-23T10:26:53Z", "readme_length": null, "readme_text": null, "readme_title": null, "releases_count": 0, "repo": "MXSensePlugin", "repo_name": "MXSensePlugin", "stars": 1, "topics": [], "updated_at": "2025-11-23T17:51:43Z", "url": "https://github.com/arthurbrunborg/MXSensePlugin", "watchers": 0 } ], "hackathon_location": "EPFL, Lausanne, Switzerland", "hackathon_name": "LauzHack 2025", "hackathon_year": 2025, "id": 7, "image_url": null, "project_fk": "lhp_aef1e68e2c0abd86", "project_id": "7", "project_title": "MX Sense", "project_uid": "id:7", "tags": [], "team": [ "Arthur Brunborg" ], "title": "MX Sense", "url": "https://github.com/arthurbrunborg/MXSensePlugin" }, { "awards": [ "Logitech Challenge 2nd Place" ], "categories": [ "Logitech Challenge 2nd Place" ], "description": "Exploring 3D Images with seamless integration using Logitech's equipment", "github_repo_count": 1, "github_repo_urls": [ "https://github.com/22cav/lauzhack2025" ], "github_repos_metadata": [ { "commit_count_default_branch": 21, "contributors_count": 4, "contributors_top": [ { "contributions": 14, "html_url": "https://github.com/22cav", "login": "22cav" }, { "contributions": 5, "html_url": "https://github.com/mattiabianco27", "login": "mattiabianco27" }, { "contributions": 1, "html_url": "https://github.com/faltaa", "login": "faltaa" }, { "contributions": 1, "html_url": "https://github.com/lprunotto", "login": "lprunotto" } ], "created_at": "2025-11-22T13:18:04Z", "default_branch": "lauzhack2025", "description": "Control Blender in real-time with hand gestures", "dirs_total_count": 19, "files_root_entries": [ { "path": ".gitignore", "size": 485, "type": "file" }, { "path": "ExamplePlugin", "size": 0, "type": "dir" }, { "path": "README.md", "size": 9544, "type": "file" }, { "path": "blender_addon", "size": 0, "type": "dir" }, { "path": "config", "size": 0, "type": "dir" }, { "path": "core", "size": 0, "type": "dir" }, { "path": "docs", "size": 0, "type": "dir" }, { "path": "gestures", "size": 0, "type": "dir" }, { "path": "handlers", "size": 0, "type": "dir" }, { "path": "inputs", "size": 0, "type": "dir" }, { "path": "logo", "size": 0, "type": "dir" }, { "path": "main_orchestrator.py", "size": 11309, "type": "file" }, { "path": "outputs", "size": 0, "type": "dir" }, { "path": "requirements.txt", "size": 112, "type": "file" }, { "path": "tests", "size": 0, "type": "dir" } ], "files_total_count": 76, "first_commit_date_default_branch": "2025-11-22T16:11:59Z", "forks": 0, "has_ci": false, "has_contributing": false, "has_docker": false, "has_docs": true, "has_license_file": false, "has_notebooks": false, "has_readme_file": true, "has_tests": true, "homepage_url": null, "input_url": "https://github.com/22cav/lauzhack2025", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 0, "issues_open": 0, "issues_total": 0, "languages_top": [ { "language": "Python", "size": 120043 } ], "last_commit_date_default_branch": "2025-11-23T18:48:58Z", "last_commit_oid_default_branch": "71b7f3f392302ee9edc00adfa408255d7be64d94", "latest_release_date": null, "latest_release_tag": null, "license_name": null, "license_spdx": null, "name_with_owner": "22cav/3DX", "owner": "22cav", "parent_repo": null, "parent_url": null, "primary_language": "Python", "project_foreign_key": "lhp_d08efd008f175913", "pull_requests_closed": 0, "pull_requests_merged": 2, "pull_requests_open": 0, "pull_requests_total": 2, "pushed_at": "2025-11-25T15:41:38Z", "readme_length": 8666, "readme_text": "
\n\n![Gesture Control System](logo/logo.jpg)\n\n# Gesture Control System\n\n**Real-time hand gesture recognition for controlling Blender's 3D viewport**\n\n[![Python 3.8+](https://img.shields.io/badge/python-3.8+-blue.svg)](https://www.python.org/downloads/)\n[![MediaPipe](https://img.shields.io/badge/MediaPipe-0.10+-green.svg)](https://google.github.io/mediapipe/)\n[![Platform](https://img.shields.io/badge/platform-macOS%20%7C%20Linux%20%7C%20Windows-lightgrey.svg)](https://github.com)\n\n
\n\n---\n\n## Overview\n\nA production-ready gesture recognition system that enables intuitive control of Blender's 3D viewport through hand gestures. Built with MediaPipe for robust hand tracking and featuring platform-specific optimizations for macOS, Linux, and Windows.\n\n### Core Features\n\n- 🎯 **4 Core Gestures** - Pinch, V-gesture, Open Palm, Closed Fist\n- 🔄 **Viewport Control** - Rotate and pan Blender's 3D viewport naturally\n- 🎬 **Animation Control** - Play/stop timeline with hand gestures\n- 🖥️ **Cross-Platform** - Optimized for macOS (main-thread), Linux/Windows (threaded)\n- ⚡ **Low Latency** - Real-time processing with smoothing filters\n- 🎛️ **Configurable** - YAML-based configuration for sensitivity and mappings\n\n---\n\n## Quick Start\n\n### Prerequisites\n\n```bash\n# Python 3.8 or higher\npython --version\n\n# Install dependencies\npip install -r requirements.txt\n```\n\n### Running the System\n\n**1. Start the gesture engine:**\n\n```bash\npython main_orchestrator.py --config config/blender_mode.yaml --debug\n```\n\n**2. In Blender:**\n- Install the addon from `blender_addon/gesture_control_addon.py`\n- Enable \"Gesture Control Center\" in Preferences → Add-ons\n- Open the sidebar (N key) → Gesture tab\n- Click **\"Connect Only\"** to link with the running engine\n\n**3. Control Blender with gestures!**\n\n---\n\n## Gestures\n\nThe system recognizes 4 core hand gestures for Blender control:\n\n| Gesture | Description | Action |\n|---------|-------------|--------|\n| 🤏 **Pinch** | Thumb + index finger touching | **Rotate viewport** - Move hand while pinching to orbit camera |\n| ✌️ **V-Gesture** | Index + middle fingers extended | **Pan viewport** - Move hand to pan camera position |\n| 🖐️ **Open Palm** | All fingers extended | **Play animation** - Start timeline playback |\n| ✊ **Closed Fist** | All fingers closed | **Stop animation** - Pause timeline |\n\n### Gesture Details\n\n**Pinch (Rotation Mode)**\n- Pinch thumb and index finger together\n- Move your hand to rotate the viewport\n- Automatic orbit around scene center\n- Release to exit rotation mode\n\n**V-Gesture (Navigation Mode)**\n- Extend index and middle fingers (peace sign)\n- Keep ring and pinky fingers closed\n- Move your hand to pan the viewport\n- Release to exit navigation mode\n\n**Animation Control**\n- Open palm to start playback\n- Closed fist to stop\n- Works independently of viewport modes\n\n---\n\n## Architecture\n\n```\n┌─────────────┐\n│ Camera │\n└──────┬──────┘\n │\n ▼\n┌─────────────────┐\n│ MediaPipe │\n│ Hand Tracking │\n└──────┬──────────┘\n │\n ▼\n┌─────────────────┐\n│ Gesture │\n│ Detector │\n└──────┬──────────┘\n │\n ▼\n┌─────────────────┐\n│ Filters & │\n│ Validators │\n└──────┬──────────┘\n │\n ▼\n┌─────────────────┐\n│ Event Bus │\n└──────┬──────────┘\n │\n ▼\n┌─────────────────┐\n│ Gesture │\n│ Handlers │\n└──────┬──────────┘\n │\n ▼\n┌─────────────────┐\n│ Blender Output │\n│ (Socket) │\n└──────┬──────────┘\n │\n ▼\n┌─────────────────┐\n│ Blender Addon │\n│ (Viewport) │\n└─────────────────┘\n```\n\n### Key Components\n\n**Gesture Detection** (`gestures/`)\n- MediaPipe-based hand landmark tracking\n- Confidence validation and quality checks\n- Smoothing filters for stable detection\n\n**Event System** (`core/`)\n- Central EventBus for message routing\n- Type-safe event handling\n- Modular handler registration\n\n**Handlers** (`handlers/`)\n- `blender_viewport_handler.py` - Viewport rotation and panning\n- `blender_animation_handler.py` - Timeline control\n- Configurable sensitivity and behavior\n\n**Blender Integration** (`blender_addon/`)\n- Socket-based communication (port 8888)\n- Real-time viewport manipulation\n- Simplified rotation and pan functions\n\n---\n\n## Configuration\n\nEdit `config/blender_mode.yaml` to customize behavior:\n\n```yaml\n# Gesture detection settings\ninputs:\n gesture:\n enabled: true\n camera_index: 0\n show_preview: true\n min_confidence: 0.6\n filter_window: 3\n\n# Blender output\noutputs:\n blender:\n enabled: true\n host: localhost\n port: 8888\n```\n\n### Sensitivity Tuning\n\nIn the Blender addon panel, adjust:\n- **Rotation Sensitivity** - Controls viewport rotation speed (default: 0.5)\n- **Pan Sensitivity** - Controls viewport panning speed (default: 0.1)\n\n---\n\n## Project Structure\n\n```\nlauzhack2025/\n├── config/ # YAML configuration files\n├── core/ # Event system and orchestration\n│ ├── event_system.py # EventBus implementation\n│ ├── gesture_handler.py # Handler base classes\n│ └── launcher.py # Application launcher\n├── gestures/ # Gesture recognition\n│ ├── detector.py # Main detection engine\n│ ├── filters.py # Smoothing filters\n│ ├── validators.py # Quality validation\n│ └── library/\n│ └── navigation.py # Core gesture definitions\n├── handlers/ # Gesture handlers\n│ ├── blender_viewport_handler.py\n│ └── blender_animation_handler.py\n├── inputs/ # Input modules\n│ └── gesture_input_production.py\n├── outputs/ # Output modules\n│ └── blender_output.py # Blender socket communication\n├── blender_addon/ # Blender addon\n│ └── gesture_control_addon.py\n├── main_orchestrator.py # Main entry point\n└── requirements.txt # Python dependencies\n```\n\n---\n\n## Platform Support\n\n### macOS\n- **Main-thread camera mode** - Required for camera permissions\n- Camera window displays in foreground\n- Launched via Terminal.app for proper access\n\n### Linux / Windows\n- **Threaded camera mode** - Background processing\n- Standard OpenCV camera access\n- Preview window optional\n\nThe system automatically detects your platform and uses the appropriate mode.\n\n---\n\n## Troubleshooting\n\n**Camera doesn't open?**\n- Check camera permissions in System Preferences (macOS)\n- Try a different camera: `--camera-index 1`\n- Ensure no other app is using the camera\n\n**Gestures not detected?**\n- Improve lighting conditions\n- Position hand clearly in frame\n- Adjust `min_confidence` in config (lower = more sensitive)\n- Check debug output with `--debug` flag\n\n**Blender not responding?**\n- Verify addon is installed and enabled\n- Check port 8888 is available: `lsof -i :8888`\n- Ensure \"Connect Only\" button was clicked in Blender\n- Check Blender's system console for errors\n\n**Viewport movement too fast/slow?**\n- Adjust sensitivity in Blender addon panel\n- Modify `sensitivity` in handler config\n- Fine-tune in real-time without restarting\n\n---\n\n## Development\n\n### Testing\n\n```bash\n# Run all tests\npython -m pytest tests/ -v\n\n# Test specific components\npython -m pytest tests/test_gestures.py\npython -m pytest tests/test_handler_system.py\n```\n\n### Adding Custom Gestures\n\n1. Define gesture in `gestures/library/navigation.py`:\n```python\n@register(\"navigation\")\nclass MyGesture(Gesture):\n @property\n def name(self) -> str:\n return \"MY_GESTURE\"\n \n def detect(self, landmarks, context):\n # Detection logic\n return GestureResult(name=self.name, confidence=0.9)\n```\n\n2. Add handler in `handlers/`\n3. Configure mapping in YAML\n\n---\n\n## Technical Details\n\n**Gesture Detection Pipeline:**\n1. MediaPipe extracts hand landmarks (21 points per hand)\n2. Landmarks filtered through smoothing window (reduces jitter)\n3. Quality validator checks landmark visibility and confidence\n4. Gesture detector matches against registered patterns\n5. Confidence validator ensures stable detection\n6. Event published to EventBus\n7. Handlers process and route to outputs\n\n**Movement Tracking:**\n- Pinch: Tracks midpoint of thumb/index, calculates deltas\n- V-Gesture: Tracks midpoint of index/middle, applies smoothing\n- Sensitivity multipliers: Rotation (20x), Navigation (150x)\n- Deadzone filtering to ignore micro-movements\n\n---\n\n## Requirements\n\n```\nmediapipe>=0.10.0\nopencv-python>=4.8.0\nPyYAML>=6.0\nnumpy>=1.24.0\n```\n\n---\n\n## License\n\nMIT License - See LICENSE file for details.\n\n---\n\n
\n\n**Built for LauzHack 2025**\n\nMade with ❤️ by the gesture control team\n\n
", "readme_title": "Gesture Control System", "releases_count": 0, "repo": "lauzhack2025", "repo_name": "3DX", "stars": 4, "topics": [ "blender-addon", "gesture-recognition", "landmarks" ], "updated_at": "2025-11-25T19:56:45Z", "url": "https://github.com/22cav/3DX", "watchers": 0 } ], "hackathon_location": "EPFL, Lausanne, Switzerland", "hackathon_name": "LauzHack 2025", "hackathon_year": 2025, "id": 8, "image_url": null, "project_fk": "lhp_d08efd008f175913", "project_id": "8", "project_title": "3DX", "project_uid": "id:8", "tags": [], "team": [ "Matteo Caviglia" ], "title": "3DX", "url": "https://github.com/22cav/lauzhack2025" }, { "awards": [ "Logitech Challenge 3rd Place" ], "categories": [ "Logitech Challenge 3rd Place" ], "description": "JARVIS is an intelligent multimodal assistant that creates seamless human-AI interaction through voice and vision. Speak, and the system automatically captures your webcam feed, transcribes your speech, and sends both to GPT for contextual understanding. The AI responds with natural spoken audio, creating a fluid conversation that understands both what you're saying and what you're looking at. Built entirely with OpenAI's latest APIs, JARVIS demonstrates the future of accessible, hands-free AI a", "github_repo_count": 1, "github_repo_urls": [ "https://github.com/AaronDinesh/lauzhack-2025" ], "github_repos_metadata": [ { "commit_count_default_branch": 109, "contributors_count": 4, "contributors_top": [ { "contributions": 49, "html_url": "https://github.com/AaronDinesh", "login": "AaronDinesh" }, { "contributions": 30, "html_url": "https://github.com/CarlOwOs", "login": "CarlOwOs" }, { "contributions": 20, "html_url": "https://github.com/GKrafft2", "login": "GKrafft2" }, { "contributions": 10, "html_url": "https://github.com/CHUKLA", "login": "CHUKLA" } ], "created_at": "2025-11-22T10:46:30Z", "default_branch": "main", "description": null, "dirs_total_count": 19, "files_root_entries": [ { "path": ".gitignore", "size": 5201, "type": "file" }, { "path": ".gitmodules", "size": 80, "type": "file" }, { "path": ".python-version", "size": 5, "type": "file" }, { "path": "README.md", "size": 0, "type": "file" }, { "path": "assistant_plan.py", "size": 6746, "type": "file" }, { "path": "assistant_plan_validator.py", "size": 12527, "type": "file" }, { "path": "audio", "size": 0, "type": "dir" }, { "path": "camera", "size": 0, "type": "dir" }, { "path": "jarvis-mx", "size": 0, "type": "dir" }, { "path": "main.py", "size": 7239, "type": "file" }, { "path": "mx_server.py", "size": 34799, "type": "file" }, { "path": "pyproject.toml", "size": 1254, "type": "file" }, { "path": "sam3", "size": 0, "type": "file" }, { "path": "sam_local", "size": 0, "type": "dir" }, { "path": "tool_executor.py", "size": 10073, "type": "file" }, { "path": "ui", "size": 0, "type": "dir" }, { "path": "uv.lock", "size": 544013, "type": "file" }, { "path": "yolo_pc_part", "size": 0, "type": "dir" } ], "files_total_count": 66, "first_commit_date_default_branch": "2025-11-22T10:46:30Z", "forks": 1, "has_ci": false, "has_contributing": false, "has_docker": false, "has_docs": false, "has_license_file": false, "has_notebooks": false, "has_readme_file": true, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/AaronDinesh/lauzhack-2025", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 0, "issues_open": 0, "issues_total": 0, "languages_top": [ { "language": "Python", "size": 119932 }, { "language": "TypeScript", "size": 51226 }, { "language": "JavaScript", "size": 20402 }, { "language": "Shell", "size": 1981 }, { "language": "CSS", "size": 1402 } ], "last_commit_date_default_branch": "2025-11-23T10:56:42Z", "last_commit_oid_default_branch": "5c61d83312c25b0481a252469ae948c3346e34ea", "latest_release_date": null, "latest_release_tag": null, "license_name": null, "license_spdx": null, "name_with_owner": "AaronDinesh/lauzhack-2025", "owner": "AaronDinesh", "parent_repo": null, "parent_url": null, "primary_language": "Python", "project_foreign_key": "lhp_b34be2c511fb277f", "pull_requests_closed": 0, "pull_requests_merged": 15, "pull_requests_open": 0, "pull_requests_total": 15, "pushed_at": "2025-11-23T10:57:00Z", "readme_length": null, "readme_text": null, "readme_title": null, "releases_count": 0, "repo": "lauzhack-2025", "repo_name": "lauzhack-2025", "stars": 2, "topics": [], "updated_at": "2025-11-27T19:23:09Z", "url": "https://github.com/AaronDinesh/lauzhack-2025", "watchers": 0 } ], "hackathon_location": "EPFL, Lausanne, Switzerland", "hackathon_name": "LauzHack 2025", "hackathon_year": 2025, "id": 9, "image_url": null, "project_fk": "lhp_b34be2c511fb277f", "project_id": "9", "project_title": "JARVIS", "project_uid": "id:9", "tags": [], "team": [ "Aaron Dinesh", "Carlos Hurtado", "Guillaume Krafft", "Krish Chawla" ], "title": "JARVIS", "url": "https://github.com/AaronDinesh/lauzhack-2025" }, { "awards": [ "AXA Challenge Winner" ], "categories": [ "AXA Challenge Winner" ], "description": "LeaseCare empowers Swiss tenants by leveraging Generative AI and Vision Language Models to provide complete legal protection throughout the rental lifecycle. In under two minutes, the platform analyzes lease agreements against the Swiss Law to identify unfair clauses while simultaneously managing photo documentation to defend against unjustified damage claims. Whether you are signing a new contract or managing a move-out, LeaseCare acts as your personal legal expert, adapting to any rental conte", "github_repo_count": 1, "github_repo_urls": [ "https://github.com/ChenghengLi/leaseCare-Public" ], "github_repos_metadata": [ { "commit_count_default_branch": 3, "contributors_count": 1, "contributors_top": [ { "contributions": 3, "html_url": "https://github.com/ChenghengLi", "login": "ChenghengLi" } ], "created_at": "2025-11-23T10:31:15Z", "default_branch": "main", "description": null, "dirs_total_count": 13, "files_root_entries": [ { "path": ".env.example", "size": 804, "type": "file" }, { "path": ".firebaserc", "size": 57, "type": "file" }, { "path": ".gitignore", "size": 469, "type": "file" }, { "path": ".vscode", "size": 0, "type": "dir" }, { "path": "DEPLOYMENT.md", "size": 4921, "type": "file" }, { "path": "README.md", "size": 40123, "type": "file" }, { "path": "SECURITY-CHECKLIST.md", "size": 4922, "type": "file" }, { "path": "create-public-repo.sh", "size": 6141, "type": "file" }, { "path": "env.d.ts", "size": 38, "type": "file" }, { "path": "firebase.json", "size": 146, "type": "file" }, { "path": "firestore.indexes.json", "size": 327, "type": "file" }, { "path": "firestore.rules", "size": 1070, "type": "file" }, { "path": "index.html", "size": 329, "type": "file" }, { "path": "package-lock.json", "size": 190890, "type": "file" }, { "path": "package.json", "size": 1076, "type": "file" }, { "path": "postcss.config.js", "size": 91, "type": "file" }, { "path": "public", "size": 0, "type": "dir" }, { "path": "src", "size": 0, "type": "dir" }, { "path": "storage.rules", "size": 614, "type": "file" }, { "path": "tailwind.config.js", "size": 259, "type": "file" }, { "path": "tsconfig.app.json", "size": 289, "type": "file" }, { "path": "tsconfig.json", "size": 139, "type": "file" }, { "path": "tsconfig.node.json", "size": 414, "type": "file" }, { "path": "vercel.json", "size": 385, "type": "file" }, { "path": "vite.config.ts", "size": 379, "type": "file" } ], "files_total_count": 61, "first_commit_date_default_branch": "2025-11-23T10:40:36Z", "forks": 1, "has_ci": false, "has_contributing": false, "has_docker": false, "has_docs": false, "has_license_file": false, "has_notebooks": false, "has_readme_file": true, "has_tests": false, "homepage_url": null, "input_url": "https://github.com/ChenghengLi/leaseCare-Public", "is_archived": false, "is_fork": false, "is_private": false, "issues_closed": 0, "issues_open": 0, "issues_total": 0, "languages_top": [ { "language": "JavaScript", "size": 1051309 }, { "language": "Vue", "size": 199828 }, { "language": "TypeScript", "size": 159917 }, { "language": "Shell", "size": 6141 }, { "language": "CSS", "size": 3979 }, { "language": "HTML", "size": 329 } ], "last_commit_date_default_branch": "2025-11-23T10:44:47Z", "last_commit_oid_default_branch": "9f51ce24df2d9993557665faf160cee706a7a362", "latest_release_date": null, "latest_release_tag": null, "license_name": null, "license_spdx": null, "name_with_owner": "ChenghengLi/leaseCare-Public", "owner": "ChenghengLi", "parent_repo": null, "parent_url": null, "primary_language": "JavaScript", "project_foreign_key": "lhp_023f4589123a2432", "pull_requests_closed": 0, "pull_requests_merged": 0, "pull_requests_open": 0, "pull_requests_total": 0, "pushed_at": "2025-11-23T10:44:47Z", "readme_length": 31328, "readme_text": "# LeaseCare Switzerland - AI-Powered Tenant Protection Platform\n\nA comprehensive tenant protection platform leveraging multi-agent AI systems, computer vision, and Swiss legal expertise to help tenants manage lease agreements, document property conditions, and defend against unfair damage claims.\n\n## 🌐 Live Demo\n\n**Deployed on Vercel**: https://leasecare.chlc.top/\n\n---\n\n## 🚀 Features\n\n### Core Capabilities\n- **🤖 Multi-Agent AI Pipeline**: 3-stage defense analysis system with specialized AI agents\n- **📄 Smart Document Analysis**: NLP-powered lease agreement parsing with risk assessment\n- **🎯 Auto Asset Detection**: ML-based classification (property, vehicles, equipment)\n- **📸 Guided Photo Documentation**: Timestamped evidence collection with metadata\n- **💬 Live Lease Chat**: Context-aware AI assistant with clickable law citations\n- **🔍 Checkout Comparison**: Computer vision damage detection with before/after analysis\n- **⚖️ Defense Report Generation**: Automated legal defense with Swiss CO compliance\n- **🇨🇭 Swiss Law Integration**: Real-time legal explanations via OpenJustice API\n\n---\n\n## 🏗️ Architecture Overview\n\n### System Architecture\n\n```\n┌─────────────────────────────────────────────────────────────┐\n│ Frontend Layer │\n│ Vue 3 + TypeScript + Tailwind CSS + Vite │\n└─────────────────────────────────────────────────────────────┘\n │\n ▼\n┌─────────────────────────────────────────────────────────────┐\n│ State Management │\n│ Pinia Store (Lease Data, User Context, Evidence) │\n└─────────────────────────────────────────────────────────────┘\n │\n ▼\n┌─────────────────────────────────────────────────────────────┐\n│ Service Layer │\n│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │\n│ │ Together AI │ │ OpenJustice │ │ Firebase │ │\n│ │ Service │ │ Service │ │ Service │ │\n│ └──────────────┘ └──────────────┘ └──────────────┘ │\n└─────────────────────────────────────────────────────────────┘\n │\n ▼\n┌─────────────────────────────────────────────────────────────┐\n│ External Services │\n│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │\n│ │ Together AI │ │ OpenJustice │ │ Firebase │ │\n│ │ (Llama-4) │ │ API │ │ (Storage) │ │\n│ └──────────────┘ └──────────────┘ └──────────────┘ │\n└─────────────────────────────────────────────────────────────┘\n```\n\n### Data Flow\n\n```\nUser Upload → Document Parser → AI Analysis → Review Stage\n ↓\n Guided Intake → Photo Storage\n ↓\n Live Chat ← Context Retrieval\n ↓\n Checkout → Damage Detection\n ↓\n Defense Pipeline → Report Generation\n```\n\n## 🛠️ Technology Stack\n\n### Frontend Framework\n- **Vue 3.5+** - Composition API with `