AIFactory Hub — High-Level Architecture

This document describes the architectural vision at a high level. Detailed implementation is proprietary.

---

Layers

Layer	Purpose	Visibility
QML Frontend	User interface	Public (this doc)
UI Shell	Module registry, theming, bridge	Public (concept)
Modules	4 workspaces	Public (concept)
Engines	Local processing (Whisper, FFmpeg)	Public (names)
AI Contracts	Provider-independent interfaces	Concept public, code private
Gateway	Facade for external systems	Concept public
Integrations	Ollama, Gemini, ComfyUI, YouTube, Telegram	Public (names)
Core	Config, logging, storage, events	Concept public

---

Communication Model

Need	Mechanism
External operation	Gateway (composition) + injected contract (runtime)
Local technical operation	Engine
Internal cross-module request	Service contract
“Something happened”	EventBus

---

Design Principles

1. Dependency direction — Higher-level components depend on lower-level contracts.
2. AI contracts are compile-time only — Not runtime layers.
3. Hybrid AI — Local-first with cloud fallback.
4. DI over singletons — Constructor injection preferred.
5. EventBus for events, contracts for requests — No module imports another module.
6. Secrets in OS keychain — Never in config files.
7. Data discipline — raw/ immutable · assets/ reusable · cache/ reconstructible.
8. UI thread responsive — Long operations run off the main thread.

---

Hybrid AI Routing

Simple tasks → Local (Ollama · Qwen · fast, private) Complex tasks → Cloud (Gemini · powerful, requires internet) Fallback → Auto (if local fails, route to cloud)

Full details are proprietary.

---

For licensing inquiries: see ../contact/.