🗂️🤖 How would you build a Lovable-style file system?

I’m building something similar to **Lovable**, but with one important detail:
👉 an **AI agent** will need to **access, create, and edit files and folders** all the time.

The idea is that **it shouldn’t matter which server is running**, everything must work the same 🔄

### 📁 What the agent (and the user) needs to do:

📤 **Upload files**
📂 **Create folders and subfolders**
📜 **List files and directories**
✏️ **Edit existing files**
🤖 **AI navigating and modifying everything on its own**

Nothing can rely on local disk, because:
⚖️ there are multiple servers
🔀 any request can hit any server

---

### 🤔 My real questions are:

🗄️ Store everything in **object storage** (like **Amazon S3 / MinIO**) and treat folders as “virtual”?
🧱 Keep the **file tree in the database** and only store blobs in S3?
✏️ How do you handle **frequent edits made by the AI**?
🔒 Is **locking/versioning** worth it, or just overwrite?
🔗 Better to generate **S3 pre-signed URLs** or always go through the backend?

---

### 🎯 The goal is simple:

📁 A reliable file system
🤖 Friendly for an AI agent to work with
🔄 Fully server-independent
🚀 Scales without becoming a headache

If you’ve built something like this before, have tips, pitfalls, or references…
drop them here 🙌🔥