ivx-ai-integration

Files (1)

• SKILL.md

SKILL.md (first lines)

# IntelliVerseX AI Integration

## Overview

The SDK's AI module provides seven subsystems for adding intelligence to games — from voice-driven hosts to real-time content moderation. All subsystems share a common provider abstraction so you can swap between managed (IntelliVerseX cloud), commercial (OpenAI, Azure, Anthropic), or self-hosted (Ollama, vLLM, LiteLLM) backends.

---

## When to Use

Ask your AI agent any of these:

* "Add an AI voice host for my trivia game"
* "Set up NPC dialog with GPT-4o"
* "Generate trivia questions using AI"
* "Add content moderation to player chat"
* "Configure Ollama as a self-hosted LLM provider"
* "Set up AI player profiling"
* "Add AI hints to my puzzle game"

---

## What the Agent Does

```mermaid
flowchart TD
    A[You: "Add AI to my game"] --> B[Agent loads ivx-ai-integration skill]
    B --> C[Creates IVXAIConfig]
    C --> D[Configures provider + model]
    D --> E{Which subsystem?}
    E -->|Voice Host| F[IVXAIVoiceServices]
    E -->|NPC Dialog| G[IVXAINPCDialogManager]
    E -->|Content Gen| H[IVXAIContentGenerator]
    E -->|Moderation| I[IVXAIModerator]
    E -->|Profiling| J[IVXAIProfiler]
    E -->|Assistant| K[IVXAIAssistant]
    E -->|Sessions| L[IVXAISessionManager]
```

---

## 1. Configuration

### IVXAIConfig ScriptableObject

Create via **Create > IntelliVerseX > AI Configuration**.

| Field | Description |
|-------|-------------|
| `Provider` | `IntelliVerseX`, `OpenAI`, `AzureOpenAI`, `Anthropic`, `Custom` |
| `BaseUrl` | API endpoint (auto-filled for named providers, required for Custom) |
| `Model` | Model identifier (e.g. `gpt-4o`, `claude-sonnet-4-20250514`, `llama3`) |
| `MaxTokens` | Default max response tokens |
| `Temperature` | Default sampling temperature (0.0–2.0) |
| `EnableVoice` | Toggle voice streaming subsystem |
| `VoiceId` | TTS voice identifier |
| `EnableMockMode` | Return canned responses without API calls (for testing) |

### API Key Injection (Secure)

**Never** hardcode API keys in config assets or source code.

```csharp
// Inject at runtime — typically from a secure config service or environment
IVXAIConfig.Instance.SetApiKey(SecureConfigService.GetKey("AI_API_KEY"));
```

For the **IntelliVerseX** managed provider, the SDK uses your `GameId` for authentication — no separate API key needed.

---

## 2. Provider Options

| Provider | Models | Voice | Cost |
|----------|--------|-------|------|
| **IntelliVerseX (Managed)** | Managed fleet (GPT-4o, Claude, etc.) | Yes | Per-token, billed to your project |
| **OpenAI** | GPT-4o, GPT-4o-mini, o1, o3 | Yes (TTS API) | Your OpenAI billing |
| **Azure OpenAI** | Same as OpenAI, deployed in your Azure tenant | Yes | Your Azure billing |
| **Anthropic** | Claude Opus, Sonnet, Haiku | No (text only) | Your Anthropic billing |
| **Custom** | Any OpenAI-compatible endpoint (Ollama, vLLM, LiteLLM) | Varies | Self-hosted |

### Self-Hosted Example (Ollama)

```csharp
IVXAIConfig.Instance.Provider = AIProvider.Custom;
IVXAIConfig.Instance.BaseUrl = "http://localhost:11434/v1"; // Ollama
IVXAIConfig.Instance.Model = "llama3:8b";
```

---

## 3. The Seven AI Subsystems

### 3.1 IVXAISessionManager

Manages conversation sessions with context windows.

```csharp
using IntelliVerseX.AI;

var session = await IVXAISessionManager.Instance.CreateSessionAsync(
    persona: "quiz_host",
    systemPrompt: "You are a witty trivia host."
);

string response = await session.SendAsync("Ask me a science question.");
```

### 3.2 IVXAINPCDialogManager

Drives in-game NPC conversations with persona configs.

```csharp
var npcConfig = new NPCDialogConfig
{
    PersonaName = "Professor Oak",
    SystemPrompt = "You are a wise professor who gives hints about nature.",
    ContextWindowSize = 10,
    MaxResponseTokens = 150,

From intelli-verse-x/Intelli-verse-X-SDK, MIT.

More from intelli-verse-x/Intelli-verse-X-SDK