Trợ Lý KOC - AI Video Generator Platform

Trợ Lý KOC is a SaaS platform enabling KOCs/Reviewers to automatically generate high-quality AI videos. The system integrates advanced AI technologies (LivePortrait, IDM-VTON, SVD-XT, MimicMotion, FaceFusion) to provide features like Talking Head, Virtual Try-on, and Face Swap.

🚀 Key Features

• Talking Head: Animate portrait photos with lip-sync from audio.
• Virtual Try-On: Virtual clothing try-on, compositing garments onto models.
• Image to Video: Convert static images into short videos.
• Motion Transfer: Transfer body motion from a reference video to a character image.
• Face Swap: Swap faces in existing videos.

---

🛠️ Technology Stack

Backend (.NET 10)

• Architecture: Modular Monolith
• Framework: ASP.NET Core 10
• Communication: MassTransit (RabbitMQ), SignalR (Real-time updates)
• Database: SQL Server (EF Core), Redis (Caching)
• Scheduling: Quartz.NET

Frontend (Next.js 14)

• Framework: Next.js (App Router), React, TypeScript
• Styling: Tailwind CSS, Shadcn/UI
• State Management: Zustand, React Query

AI Worker (Python 3.11)

• Framework: PyTorch, CUDA
• Features: GPU-accelerated inference, Queue-based processing
• Libraries: Diffusers, InsightFace, FFMPEG

Infrastructure

• Containerization: Docker, Docker Compose
• Storage: MinIO (S3 Compatible)
• Gateway: Cloudflare Tunnel

---

⚙️ Installation Guide (Local Development)

Prerequisites

• Docker Desktop & Git
• NVIDIA GPU (VRAM >= 8GB) & NVIDIA Container Toolkit (for AI Worker)
• .NET 10 SDK (Optional, if running backend outside docker)
• Node.js 20+ (Optional, if running frontend outside docker)

Step 1: Clone Repository

git clone https://github.com/your-repo/tro-li-koc.git
cd tro-li-koc

Step 2: Environment Configuration

Copy the .env.example file to .env in the docker/ directory:

# In docker/ folder
cp .env.example .env

Note: Update environment variables such as CLOUDFLARE_TUNNEL_TOKEN if necessary.

Step 3: Start the System

Run the entire system using Docker Compose:

cd docker
docker-compose up -d --build

Services will be available at:

• Frontend: http://localhost:3000
• Backend API: http://localhost:5500 (Swagger: http://localhost:5500/swagger)
• RabbitMQ: http://localhost:15679 (User: admin, Pass: admin123)
• MinIO: http://localhost:9031 (User: minioadmin, Pass: minioadmin123)

---

📂 Project Structure

tro-li-koc/
├── docker/                 # Docker configuration (Compose, Env)
├── docs/                   # Documentation & ADRs
├── scripts/                # Helper scripts
├── src/
│   ├── ai-worker/          # Python AI Worker service
│   ├── backend/            # .NET 10 Modular Monolith API
│   │   ├── src/
│   │   │   ├── TroLiKOC.API/           # Main Entry Point
│   │   │   ├── TroLiKOC.SharedKernel/  # Shared code
│   │   │   └── Modules/                # Domain Modules (Identity, Jobs, etc.)
│   └── frontend/           # Next.js Web App
└── README.md

📚 Detailed Documentation (ADRs)

Key architectural decisions are recorded in docs/adr:

• ADR 001: Database Choice (SQL Server + Redis)
• ADR 002: MassTransit Version
• ADR 004: AI Worker Architecture
• ADR 005: RabbitMQ RabbitMQ Routing
• ADR 006: Quartz.NET Configuration

---

📝 License

This project is proprietary software. All rights reserved.