While Local AI offers privacy and offline capabilities, Cloud AI provides unmatched reasoning power and general knowledge. HomeGenie Server 2.0 integrates seamlessly with Google Gemini, allowing you to offload complex tasks—like natural language interpretation or code generation—to one of the world's most advanced AI models.

Why Cloud?

• Superior Intelligence: Models like Gemini 2.5 Pro/Flash have a much larger context window and reasoning capability
  than any model that can run on a Raspberry Pi or a workstation.
• Zero Hardware Load: Your server doesn't need heavy RAM or NPU. The heavy lifting is done by Google's data centers.

Powered by Gemini

Integrating Google's multimodal AI into your automation.

Open media

1. Gemini Intent Handler

The Intent Handler is a program that transforms natural language into HomeGenie commands. Unlike old-school voice assistants that require rigid phrases ("Turn on light 1"), Gemini understands Intent and Context.

How it works

1. You speak or type a command (via HomeGenie Panel or a chat widget).
2. The program captures the State of your Home (list of devices, names, and current status).
3. It sends this context + your command to the Gemini API.
4. Gemini replies with a precise JSON command structure.
5. HomeGenie executes the action.

Example Scenario:

• User says: "It's too dark in here for reading."
• Gemini thinks: The user implies they need light. I see a 'Living Room Lamp' and a 'Reading Light' in the
  device list. I should turn them on.
• Action: Turns on the relevant lights.

Code Example (C#)

This snippet demonstrates how to construct a "Context-Aware" prompt for Gemini.

// 1. Gather Context: Get a list of all controllable devices
var devices = Modules.InDomain("HomeAutomation.HomeGenie").List;
string deviceContext = JsonConvert.SerializeObject(devices.Select(d => new { d.Name, d.Address, d.Type }));

// 2. Build the System Prompt
string systemPrompt = $@"
You are a smart home assistant. 
Here is the list of available devices (JSON): {deviceContext}.
Interpret the user request and output a JSON array of actions.
Format: [ { ""Address"": ""..."", ""Command"": ""Control.On"" } ]
Do not output markdown or explanations. Just JSON.";

// 3. Call Gemini API (Conceptual)
var response = await GeminiClient.GenerateContentAsync(systemPrompt, userRequest);

// 4. Execute Actions
dynamic actions = JsonConvert.DeserializeObject(response);
foreach(var action in actions) {
    Modules.WithAddress((string)action.Address).Command((string)action.Command).Submit();
}

2. Widget Genie (Coding Assistant)

Widget Genie is your personal pair programmer inside HomeGenie. It leverages the coding capabilities of Cloud AI to help you build custom Widgets, write Automation Scripts, or debug errors.

The "Platform Architect" Persona

Widget Genie isn't just a generic chatbot. It is initialized with a specialized System Prompt that acts as a "HomeGenie Platform Architect". This means it comes pre-loaded with deep knowledge of the system's internal mechanics:

• zuix.js Expert: It knows the specific syntax for HomeGenie widgets (e.g., using @if directives instead of Angular/Vue style interpolation).
• API Knowledge: It knows the exact C# helper classes (Modules, Program, Api) and doesn't invent non-existent methods.
• Best Practices: It automatically handles common pitfalls, such as the "Comma Trap" (parsing decimal numbers correctly in different locales) and proper resource disposal.

Capabilities

• Generate Widgets: Describe a UI ("I need a red button that turns on the fan and shows a spinning icon when active") and the Widget Genie will write the HTML, CSS, and JavaScript code for you, respecting the Shadow DOM isolation.
• Fix Code: Paste an error log or a broken script, and Widget Genie will suggest the correction based on the official API.
• Hardware Advice: Ask about integrating specific hardware (e.g., "How do I connect an ESP32-S3 camera?") and it will recommend the correct HomeGenie Mini firmware and configuration.

Widget Genie

Asking the AI to generate a custom thermostat widget code.

Open media

Local vs. Cloud: What to choose?