🔌 NeuroVerse Plugin System Overview : Continues to Optimise

The NeuroVerse plugin system empowers developers to extend the assistant's capabilities using dynamic, runtime-loaded APK plugins. These plugins can run AI requests, interact with Android APIs, and even define custom actions, all through a simple, well-documented architecture.

🎯 Project Goal

Develop a flexible, APK-based plugin system for NeuroVerse, enabling third-party features and dynamic automation tools to be loaded and executed at runtime.

🛠 Core Features

• APK-based Plugin Format: Each plugin is packaged as an APK and bundled with a manifest.json.
• Native Android Support: Plugins use full Android API access (e.g., Context, Activity, View).
• No Custom Toolchain: Build plugins using Android Studio with Kotlin or Java.
• Plugin API Interface: All plugins implement a shared interface from plugin-api.jar.
• Dynamic Loading: Uses DexClassLoader to load the plugin’s main class and invoke Plugin.run(Context).
• AI-Driven Execution: Commands are routed to matching plugins using a JSON schema and AI-based intent matching.

🧱 Plugin Architecture

1. Shared Interface

Each plugin extends the Plugin base class:

open class Plugin(protected val context: Context) {
    open fun getName(): String = "none"
    open fun onStart() {}
    open fun submitAiRequest(prompt: String): JSONObject = JSONObject()
    open fun onAiResponse(response: JSONObject) {}
    open fun onStop() {}
}

2. Manifest File Format (manifest.json)

{
  "name": "AppLauncherPlugin",
  "description": "Launches any installed app based on user request",
  "main": "com.example.plugin.AppLauncherPlugin",
  "plugin-api-version": "1.0",
  "permissions": ["LAUNCH_APP"]
}