v2.0.24

HomeGenie 🧞 v2.0.24

🎉 What's new in this release

This update enhances HG AI's capabilities with a powerful new local model, upgrades default cloud assistants, and resolves critical logic and UI bugs to deliver a more stable and reliable automation experience.

🧠 AI & Widget Genie Enhancements

• New Local AI Model: Added Qwen3.5-9B-DeepSeek-V4-Flash to the list of Local AI models (Lailama).
• Gemini Upgrades: Updated all Gemini-based AI Genies to default to gemini-flash-latest (currently pointing to Gemini 3.5 Flash).
• Smarter Widget Genie: Improved the default system instructions for the Widget Genie for more precise and reliable real-time widget generation.

⚙️ System, Logic & UI Fixes

• Robust Program ID Generation: Resolved a critical bug where deleting intermediate programs could cause a program ID collision on subsequent additions. The generation engine has been modernized with robust LINQ checks to guarantee strictly unique and sequential IDs.
• Smart Custom Widget Filtering: Fixed a bug where the custom widget module selector used a loose "OR" matching logic. It now enforces a strict "AND" logic, ensuring modules are only shown if they satisfy both type and field filters simultaneously.
• UI Polish: Resolved minor user interface issues to ensure a smoother look and feel.

Full Changelog: v2.0.23...v2.0.24

v2.0.23

HomeGenie 🧞 v2.0.23

🎉 What's new in this release

This update brings an exciting enhancement to our AI integration, alongside a massive under-the-hood modernization of the HomeGenie core engine to ensure better stability, performance, and future-proofing.

🧠 Smarter Gemini AI Assistant

The Gemini AI integration now has a sense of time! We have updated the AI's system prompt and provided it with a new internal tool to read the current date and time.

⚙️ System & Stability Improvements

We've updated almost all the foundational libraries that power HomeGenie to bring you a smoother experience:

• Fully powered by .NET 10: The core home automation engine has been updated and fully optimized for the modern .NET 10 framework, guaranteeing Long Term Support (LTS), maximum efficiency, and top-tier performance.

Full Changelog: v2.0.22...v2.0.23

v2.0.22

HomeGenie 🧞 v2.0.22

This release brings significant improvements to the AI automation engine, enhanced UI components, and important infrastructure stability fixes.

🚀 Key Highlights

AI & Automation Enhancements

We have completely overhauled the integration with Google Gemini to make the AI assistant faster, cheaper to run, and more reliable:

• Gemini 3 Support & Function Calling: The AI (Program 930) now uses a "pull-based" tool-use architecture. Instead of processing the entire house state with every message, it now queries the status only when needed.
• Improved Efficiency: This change drastically reduces token consumption and fixes communication errors (HTTP 400) when interacting with Gemini 3+ models.
• Optimized Context Engine: Added a new status endpoint (Program 578) to provide the AI with precise, on-demand data, resulting in faster and more accurate responses.
• Refined Prompting: The base System Prompt has been improved for better contextual reasoning and more "natural" home interaction.

UI & Widget Updates

• New Widget Examples: Updated and polished the library of example custom widgets, including:
  • dimmer, weather-widget, time-clock, and display-temperature.
• UI Fixes: Resolved a minor bug in the media-center-dialog for a smoother user experience.

Infrastructure & Fixes

• UPnP/SSDP: Integrated a patch for the UPnP library to ensure proper resource disposal, preventing potential memory leaks and improving network stability.

---

Note for Developers:
If you are using the AI automation programs, ensure both Program 930 (Gemini Intent Handler) and Program 578 (Context Engine) are updated to this version to ensure full compatibility with the new Function Calling implementation.

Full Changelog: v2.0.21...v2.0.22

v2.0.21

HomeGenie 🧞 v2.0.21 - Release Notes

🚀 New Features & Enhancements

• LLamaSharp Upgrade: Upgraded core LLamaSharp dependency from v0.26.0 to v0.27.0. This update brings significant performance improvements and internal fixes.
• Support for New LLMs: Added native support for the latest generation of Large Language Models:
  • QWen3.5
  • Gemma4
• Expanded Model List: These new models have been added to the default model selection list, allowing you to easily switch to the latest state-of-the-art architectures directly from the settings.

🛠 Technical Note

• This version aligns with the latest LLamaSharp v0.27.0 release, ensuring better compatibility and extended capabilities for your local AI automation tasks.

Full Changelog: v2.0.20...v2.0.21

v2.0.20

HomeGenie 🧞 v2.0.20 - Release Notes

This update introduces a major connectivity feature for mobile users, alongside important UI and AI tool improvements.

🚀 What's New

• HomeGenie Panel Notifications Program:
  We have introduced a new, dedicated automation program that allows HomeGenie Server to send real-time push notifications directly to the HomeGenie Panel Android app. You can now easily route device state changes and security alarms to your smartphone with full privacy and zero-friction configuration.
  (Note: Make sure to update your HomeGenie Panel app to the latest version to enjoy this feature!)

• Widget Genie AI Prompt Improvements:
  The internal prompt for "Widget Genie" has been refined and optimized. The AI now understands context better, resulting in more accurate, reliable, and better-looking custom widget generation.

🛠 Updates & Maintenance

• Monaco Editor Update:
  The built-in code editor has been updated to Monaco v0.55.1, bringing the latest syntax highlighting improvements and a smoother scripting experience for developers and power users.

Full Changelog: v2.0.19...v2.0.20

v2.0.19

HomeGenie 🧞 v2.0.19 - Release Notes

🚀 Improvements & Fixes

• Lailama AI Responsiveness: Fixed an issue where the AI assistant would occasionally fail to respond to the first user input after startup. The initialization (warmup) process has been refined to ensure the model context is fully ready to process messages immediately.
• AI Chat: Fixed bug in microhone input handling

Full Changelog: v2.0.18...v2.0.19

v2.0.18

HomeGenie 🧞 v2.0.18 - Release Notes

This release introduces significant improvements to the widget engine, focusing on portability, robustness, and internal dependencies.

🚀 Key Features & Changes

• Dynamic Module Generation: Implemented automatic generation of .module.js files for all widgets. A new template system located at ./data/widgets/.template.module.js is now used to ensure every widget is correctly initialized and portable.
• Widget Export Improvements: Widget export (ZIP archives) now includes the auto-generated .module.js file, ensuring that exported widgets are fully self-contained and ready to use when imported into other instances.
• Legacy Support: Automatically generated .module.js files for all pre-existing example widgets to ensure full compatibility with the new module system.

🛠 Dependency Upgrades

• Zuix.js upgraded to v1.2.7: This version includes a critical fix for Shadow DOM context resolution, improving stability when widgets are moved, re-attached, or dynamically orchestrated in complex dashboards (e.g., Angular-based environments).
• Jint engine upgraded to v4.8.0: Updated the internal scripting engine to the latest stable version for improved performance and compatibility.

Full Changelog: v2.0.17...v2.0.18

v2.0.17

HomeGenie 🧞 v2.0.17 - Release Notes

We are excited to bring you the latest improvements for Widget Genie and Automan, making your coding and debugging workflow smoother than ever.

🚀 What's New

• Enhanced Widget Genie Intelligence: We've updated the system prompts to provide more accurate and context-aware assistance.
• Context-Aware Fixes: Widget Genie now understands your code at a deeper level. You can simply ask for information or fixes for any specific program by referencing its number.
• Smart Error Resolution: Never get stuck on a build error again. When the Widget Genie chat is open and a compilation error occurs, you can now click the "Ask AI" button to automatically fix the error.
• Expanded Knowledge Base: We have integrated API statistics and data logger documentation into Widget Genie's system prompt.

Full Changelog: v2.0.16...v2.0.17

v2.0.16

HomeGenie 🧞 2.0.16: Smarter Agents, Stable Workflows

This release is all about intelligence and refinement. We've taken the groundbreaking capabilities of Gemini Automan and made them smarter, more stable, and more deeply integrated into the HomeGenie ecosystem.

🧠 A New Brain for Your Agents: Pure Text Processing

Until now, Automan was a master code generator. Now, it's also an oracle. We've introduced the new Prompt.Ask API, which allows AI-generated programs to ask questions and get simple text answers (like summaries, translations, or data analysis) without generating new code. This unlocks incredible new possibilities for creating true AI data pipelines right inside your automations.

⛓️ More Robust Agentic Workflows

We've supercharged Automan's ability to orchestrate complex tasks:

• Shared Workspace: Sub-agents can now securely pass large amounts of data to each other using a dedicated shared folder (SharedAutomanWorkspace), enabling complex workflows like our new "Autonomous News Agency" tutorial.
• Smarter Lifecycle Management: We've fine-tuned the AI's understanding of HomeGenie's architecture. It now correctly implements the two-phase Setup block, preventing common restart loop bugs and ensuring generated programs are stable and efficient by default.

🛠️ UI Fixes & Documentation

• We've squashed a bug in the local-ai-chat example widget, improving its reliability.
• The "Learn with AI" section on HomeGenie website has been updated with new, in-depth tutorials that showcase these advanced multi-agent capabilities. Go check them out and see how to build your own digital workforce!

This update lays the foundation for even more complex and intelligent agentic systems in the future.

Happy Automating! 🚀

Full Changelog: v2.0.15...v2.0.16

v2.0.15

HomeGenie 🧞 2.0.15: Hardware Acceleration & Desktop Experience

This release represents a massive leap forward for Local AI performance and everyday usability. We've completely overhauled our build system to unleash the full power of your GPU, while transforming HomeGenie from a standard background server into a seamless, native-feeling desktop app.

⚡ Unleash Your GPU (Dedicated AI Builds)

Running Local AI (LLMs and Computer Vision) is now faster than ever. To isolate native dependencies and provide the best out-of-the-box performance, we are now shipping hardware-specific accelerated builds:

• CUDA 12 Edition (-cuda12): Massive, fully-loaded builds packed with NVIDIA acceleration. Perfect for turbocharging LLamaSharp, YoloSharp, and ONNX Runtime. If you have a modern NVIDIA GPU, this is your weapon of choice.
• Vulkan / DirectML Edition (-vulkan): Broad cross-vendor GPU support (AMD, Intel, NVIDIA). It leverages DirectML on Windows and Vulkan on Linux for highly efficient, accelerated AI processing.
• macOS Edition: The standard macOS build now comes with native Apple GPU acceleration (Metal) included and enabled by default for blazing-fast local AI out of the box!
• Standard Editions: The classic, lightweight builds you know and love, perfect for Raspberry Pi (ARM/ARM64) and standard CPU-only execution on Windows/Linux.

🎨 The "Native App" Experience

We've bridged the gap between web dashboard and desktop software:

• PWA Auto-Launch: By default, launching HomeGenie will now automatically open its Progressive Web App (PWA) in a dedicated, chromeless browser window (--start-browser). It looks and feels exactly like a native desktop app.
• Single-Instance Magic: Already got HomeGenie running in the background? No problem. Launching the executable again won't cause annoying "port already in use" errors. Thanks to our new global Mutex architecture, it will smartly detect the running instance and simply open a new UI window connected to it!
• Note for power users: Need a headless server? Just remove the --start-browser flag from your startup script.

---

📦 Which file should I download?

With so many new options, here is a quick guide to choosing the right file for your setup:

• NVIDIA GPU Owners (Win/Linux): Download ...-cuda12.zip for maximum local AI performance.
• AMD/Intel GPU Owners (Win/Linux): Download ...-vulkan.zip for cross-platform hardware acceleration.
• Raspberry Pi / Mini PCs: Download ...linux-arm64.zip or ...linux-arm.zip.
• macOS Users: Download ...osx-x64.zip (Native Apple GPU support is built-in!).
• Standard CPU-only (Win/Linux): Download the base ...win-x64.zip or ...linux-x64.zip.

Full Changelog: v2.0.14...v2.0.15