v2.0.2

What's Changed

• Refactor model load retries and timeouts for GPU layers by @avikeid2007 in #33

Full Changelog: v2.0.0...v2.0.2

v2.0.0

Release Notes

✨ New Features & Major Updates

• WinUI 3 Modernization: Completely rebuilt the desktop application framework—migrating from WPF to WinUI 3 / Windows App SDK. Enjoy native Windows 11 design idioms, smoother animations, and better performance!
• Vision Models Support: Added capability to run multimodal models natively. You can now chat about your images securely and locally.
• Modernized "RAG as a Service" Interface: Completely redesigned the Document Management and Local API view using native WinUI data controls, adding empty states, and "RUNNING" badge indicators.
• Model Catalog Enhancements: Added new "Expand All" and "Collapse All" quick actions to easily manage large lists of downloaded and available models.
• Quick Links: Added convenient buttons in the Settings view to rate the app on the Microsoft Store, request features, and star the GitHub repository.

🐛 Bug Fixes & Improvements

• Global Navigation Button Fix: Fixed an overlapping hit-test bug where the custom window drag-region intercepted clicks, allowing the hamburger menu toggle to work reliably again.
• Chat "Enter to Send" Fix: Addressed an event routing issue specific to WinUI 3 where pressing 'Enter' would leave trailing newlines instead of instantly sending the message.
• Smart Theme Sync: Markdown rendering in chat bubbles now natively inherits your dynamic system theme via the parent visual tree—ensuring flawless text legibility across both Dark and Light modes.
• Iconography Overhaul: Replaced standard Unicode text icons throughout the application (such as in the Settings page) with proper, scalable Segoe Fluent theme icons.

What's Changed

• Avi/add new win UI by @avikeid2007 in #28
• feat: Update MSIX and release workflows to disable temporary store ce… by @avikeid2007 in #30
• Avi/add new win UI by @avikeid2007 in #31
• Refactor XAML files for improved readability and consistency by @avikeid2007 in #32

Full Changelog: 1.0.9...v2.0.0

v1.0.10

What's Changed

• Avi/update nugets by @avikeid2007 in #26
• fix: Update SDK version in global.json and add KaiROS.AI.WinUI soluti… by @avikeid2007 in #27
• Avi/wpf version by @avikeid2007 in #29

Full Changelog: v1.0.8...v1.0.10

1.0.9

What's Changed

• Avi/update nugets by @avikeid2007 in #26
• fix: Update SDK version in global.json and add KaiROS.AI.WinUI soluti… by @avikeid2007 in #27
• Avi/wpf version by @avikeid2007 in #29

Full Changelog: v1.0.8...1.0.9

v1.0.8

What's Changed

• Avi/upload doc feature by @avikeid2007 in #24
• update by @avikeid2007 in #25

Full Changelog: v1.0.6...v1.0.8

v1.0.6

What's Changed

• Add Vulkan backend, Arc detection, and upgrade LLamaSharp by @avikeid2007 in #14
• Deply As Powershell by @avikeid2007 in #15
• feat: introduce hardware detection service, model manager, and model … by @avikeid2007 in #16
• Avnish/cert fix by @avikeid2007 in #17
• Change logo image source in README by @avikeid2007 in #18
• feat: Implement initial mobile chat application with LLM integration,… by @avikeid2007 in #19
• Avnish/play store changes by @avikeid2007 in #20
• feat: Introduce mobile version with dedicated assets, updated README,… by @avikeid2007 in #21
• feat: Add GitHub Actions workflow to build and release Windows and An… by @avikeid2007 in #22
• Pages redesign by @avikeid2007 in #23

Full Changelog: 1.0.5...v1.0.6

v1.0.5

Release Notes - KaiROS AI (v1.0.5)

🚀 Highlights

📦 Massive Model Catalog Expansion

The model catalog has been expanded to 31 pre-configured models across 9 major organizations.

• New Organizations: Qwen, Google (Gemma 3), Meta (LLaMA 3.1/3.2), Microsoft (BitNet, Phi-3), MistralAI, OpenAI (GPT-oss), RWKV, Stability AI, and SmolLM2.
• UI Redesign: Features collapsible organization sections and a dedicated "Downloaded Models" section for quick access.
• Advanced Filtering: New filters for Organization, Model Family (e.g., Qwen, Gemma), and Hardware Variant (CPU-Only, GPU-Recommended).

⚡ Hardware & Performance

• Next-Gen GPU Support: Fixed VRAM detection for modern GPUs (RTX 50/40/30 series and AMD RX 7000/6000). Correctly identifies 16GB VRAM on systems like the RTX 5070 Ti.
• LLamaSharp 0.25.0 Upgrade: Updated the core runtime to support the latest GGUF formats, including Qwen 3 and Gemma 3 models.
• Smart Fallback Loading: If a model fails to load into GPU memory, the app now automatically retries with 50% layers, then 25%, and finally falls back to CPU mode to ensure a successful load.

🔌 Developer & API Enhancements

• Enhanced Local API: The built-in REST API is now more robust for developer integrations (VS Code Continue, custom scripts).
• Custom Model Icons: Local and custom-added models now feature a standard KaiROS logo and unified metadata display.
• Improved Documentation: Updated README.md and GitHub Pages with detailed developer guides and high-resolution screenshots of the new RAG and Catalog features.

🛠️ Internal Improvements

• Package Updates: All LLamaSharp backends (CUDA 12, CPU, Vulkan) updated to 0.25.0.
• UI Polish: Added high-quality organization logos and model badges for better scannability.
• Robustness: Improved error handling for LoadWeightsFailedException with detailed logging.

🔜 Coming Soon

• Microsoft Store: The app is currently being prepared for its official Microsoft Store debut!

---

Run powerful LLMs locally with KaiROS AI — Private. Fast. Free.

What's Changed

• feat: implement LLM model management, including a model catalog, cust… by @avikeid2007 in #11
• New feature and design enhancement for the model catalog, hardware detection, and user experience by @avikeid2007 in #12
• Add Chat Interface section to README with screenshot by @avikeid2007 in #13

Full Changelog: 1.0.4...1.0.5

v1.0.4

Release Notes - KaiROS AI v1.0.4

🚀 New Features

Custom Model Support

• ➕ Add your own .gguf models from local files or download URLs
• 📦 SQLite database stores custom model entries persistently
• 🗑️ Delete custom models with one click

Execution Backend Selection (Now Working!)

• 🎛️ Choose between CPU, CUDA, DirectML, or NPU in Settings
• ✅ Selection now properly applies when loading models
• 📊 Loading text shows actual selected backend

API Mode Enhancements

• 🌐 Added internetClient and internetClientServer capabilities
• 🔌 Improved API stability

RAG Document Support

• 📄 Enhanced debug logging for document loading
• 🔍 Better context retrieval tracking
• 📝 Support for PDF, Word, and text files

---

🐛 Bug Fixes

• Fixed: Execution Backend UI wasn't applying selection
• Fixed: "Loading on GPU" text showed regardless of backend selection
• Fixed: Radio buttons for backend selection weren't working
• Fixed: MessageBox and OpenFileDialog ambiguity errors

---

📦 Technical Changes

• Added Microsoft.Data.Sqlite for custom model persistence
• Updated IHardwareDetectionService with SetSelectedBackend() method
• Added comprehensive debug logging for RAG pipeline
• Manifest now includes network capabilities

---

Full Changelog: v1.0.3 → v1.0.4

What's Changed

• docs: add privacy policy and download instructions to README. by @avikeid2007 in #3
• Avnish/msix support by @avikeid2007 in #4
• feat: Add MSIX packaging, configure application startup with dependen… by @avikeid2007 in #5
• feat: add project website and GitHub Pages deployment workflow by @avikeid2007 in #6
• Avnish/v2 by @avikeid2007 in #7
• Avnish/api mode by @avikeid2007 in #9
• feat: implement LLM model management, including a model catalog, cust… by @avikeid2007 in #10

Full Changelog: 1.0.0...1.0.4

v1.0.0 - Initial Release 🚀

KaiROS AI v1.0.0 🧠

The first official release of KaiROS AI – A powerful local AI assistant for Windows!

✨ Features

• 🤖 Local LLM Execution - Run AI models privately on your own hardware
• 🚀 GPU Acceleration - CUDA 12 & Vulkan support for NVIDIA/AMD/Intel GPUs
• 📦 Model Catalog - Pre-configured models from Hugging Face (TinyLlama, Phi-3, Mistral, LLaMA)
• ⬇️ Download Manager - Pause, resume, and manage model downloads
• 💬 Chat Interface - Modern UI with streaming responses
• 💾 Chat History - Save and restore chat sessions (SQLite)
• 📊 Performance Stats - Real-time tokens/sec and memory usage
• 🎨 Beautiful Dark Theme - Modern gradient-based UI design
• ⚙️ Hardware Detection - Automatic CPU/GPU/NPU detection

📥 Installation

1. Download KaiROS.AI-1.0.0-Setup.exe
2. Run the installer
3. Launch KaiROS AI from Start Menu
   Silent Install:

KaiROS.AI-1.0.0-Setup.exe /VERYSILENT /SUPPRESSMSGBOXES

💻 System Requirements

• Windows 10/11 (x64)
• 4+ GB RAM (8+ GB recommended)
• Optional: NVIDIA GPU with CUDA 12 for acceleration

📋 Included Models

Model	Size	RAM Required
TinyLlama 1.1B	0.8 GB	2 GB
Phi-3 Mini 3.8B ⭐	2.2 GB	4 GB
Mistral 7B ⭐	4.4 GB	8 GB

---

Made with ❤️ for local AI enthusiasts