Unity OBB Mounter (Debug & AI Testing Tool)

This package provides a utility to pack StreamingAssets into an OBB (Opaque Binary Blob) expansion file and mount it directly to the Android filesystem.

---

🧠 The Problem

On Android, Unity typically handles large builds by splitting them into an APK and an OBB expansion file. However, this creates a significant technical hurdle for developers using native libraries:

1. Unity's Default OBB is an Unmounted Archive

By default, Unity treats the OBB like a standard zip archive. While Unity's internal UnityWebRequest or Resources APIs can "see" inside it via the jar:file:// protocol, the Android OS does not actually mount the file into the filesystem. It remains a single compressed blob that the OS views as a single file.

2. Native Libraries Cannot Access "Internal" Paths

Most native AI engines and C++ plugins (ONNX Runtime, Llama.cpp, TensorFlow Lite) require a standard absolute file path (e.g., /storage/emulated/0/...).

• Native Failure: They cannot "reach inside" a compressed OBB or APK.
• No Abstraction: They do not have access to Unity's internal asset management layer.
• Result: Even if your model is inside the OBB, passing a path to a native plugin will return a File Not Found error because the path isn't "real" to the Linux kernel.

3. The "Double Storage" Penalty

To bypass this, developers usually copy models from the OBB to Application.persistentDataPath at runtime. This results in:

• Wasted Space: A 2GB model now takes up 4GB of user storage (one copy in the OBB, one in the data folder).
• Slow First Launch: Users must wait minutes for the "extraction" process to finish before the AI is ready.

This tool mounts the OBB as a virtual filesystem, providing a direct, raw path to your assets that native C++ libraries can read immediately with zero copying and zero storage overhead.

---

🚀 Comparison

Feature	Standard Unity OBB	OBB Mounter (This Tool)
Access Method	Unity-only APIs (WebRequest)	Standard File I/O (File.Open)
Native Compatibility	❌ No	✅ Yes (Direct Path)
Disk Overhead	2x (if extracted for native)	✅ 1x (Zero Copy)
Initial Load Time	Slow (Extraction required)	✅ Instant

---

⚠️ WARNING: DEBUG TOOL ONLY

• OBB is Obsolete: Google Play now uses Play Asset Delivery (PAD) for production.
• Intended Use: This is a debugging/testing utility to help AI developers iterate quickly without writing complex extraction logic.
• Compatibility: OBB mounting relies on the Android StorageManager API, which can vary across device manufacturers and Android versions.

---

💻 Usage

1. Setup: Drag and drop the AndroidObbMount prefab into your initial scene.
2. Persistence: Keep the prefab alive (or use DontDestroyOnLoad) if you are changing scenes, or the OBB file will be unmounted.
3. Implementation: Use a coroutine to wait for the mount point to resolve. If mounting fails, the tool falls back to the standard Application.streamingAssetsPath.

using System.Collections;
using UnityEngine;
using System.IO;

public class MyModelLoader : MonoBehaviour
{
    IEnumerator Start()
    {
        // Wait until the mount point is resolved by the prefab
        while (string.IsNullOrEmpty(AndroidObbMount.AndroidObbMount.mountPoint))
        {
            yield return null;
        }

        string rootPath = AndroidObbMount.mountPoint;
        string modelPath = Path.Combine(rootPath, "my_ai_model.onnx");

        Debug.Log("Accessing model at: " + modelPath);
        
        // Pass modelPath directly to your native AI plugin (C++/Rust/etc)
        MyNativePlugin.Init(modelPath);
    }
}