OpenGenie AI Stack

正體中文 | 日本語 | 한국어

A modular, self-hosted AI infrastructure framework for AMD, NVIDIA, and ARM64 hardware. Deploy a full-stack private AI appliance — LLM inference, RAG pipeline, workflow automation, and observability — on your own hardware in minutes.

---

Features

• Multi-GPU Support — AMD ROCm, NVIDIA CUDA, ARM64 (Apple Silicon, Jetson, Ampere)
• 12-Phase Methodology — structured, independently deployable modules from driver setup to monitoring
• LLM Inference — Ollama + OpenWebUI with always-ready VRAM optimization and Lemonade native engine
• RAG Pipeline — Qdrant vector DB + Docling document processor + Mosquitto MQTT
• Workflow Automation — n8n in queue mode with Redis and distributed workers
• Observability — Grafana + Prometheus + Loki + cAdvisor + DCGM Exporter (GPU metrics)
• One-Click Backup — timestamped backup and restore for all persistent data
• Auto Hardware Tuning — HWI Advisor auto-detects hardware and generates optimal config

---

Quick Start

Prerequisites

• Ubuntu 22.04 / 24.04 LTS
• Docker Engine + Docker Compose v2
• GPU drivers installed (ROCm / CUDA / NVIDIA Container Toolkit)
• sudo access

1. Clone

git clone https://github.com/TigerAI-Taiwan/OpenGenie-AI-Stack.git
cd OpenGenie-AI-Stack

2. Choose your stack

Hardware	Directory
NVIDIA GPU	deployments/nvidia-compose-stack/
AMD ROCm GPU	deployments/amd-compose-stack/
ARM64 (Apple Silicon / Jetson / Ampere)	deployments/arm64-compose-stack/

cd deployments/amd-compose-stack   # or nvidia / arm64

3. Configure

cp .env.example .env
# Edit .env — replace all CHANGE_ME values with your own credentials
nano .env

4. Hardware calibration (recommended)

sudo bash master-deploy.sh init

This auto-detects your CPU/GPU and writes a tuning profile to tiger-tuning.env.

5. Deploy

# Full deployment (all phases)
sudo bash master-deploy.sh all

# Or deploy individual phases
sudo bash 02-database-postgres-pgadmin/deploy.sh
sudo bash 03-ai-interface-ollama-openwebui-redis/deploy.sh

6. Verify

sudo bash master-deploy.sh test

---

12-Phase Architecture

Phase	Layer	Components
00	HWI Advisor	Auto hardware calibration, tuning profile
00	Foundation	Driver setup, Docker, Node-RED
01	Infrastructure	Portainer, WebSSH
02	Database	PostgreSQL 17, pgAdmin 4
03	AI Interface	Ollama, OpenWebUI, Redis
04	Automation	n8n (queue mode + workers)
05	RAG Stack	Qdrant, Docling, Mosquitto
06	AI Core Engine	Lemonade inference engine
07	Validation	Health checks, benchmark scripts
08	Backup & Recovery	1-click backup, restore, VRAM purge
09	Monitoring & Alerts	tiger-monitor, MQTT alerting
10	Observability	Grafana, Prometheus, Loki, cAdvisor
11	Lifecycle	What's Up Docker (WUD)

---

Service Ports (Default)

Service	Port
OpenWebUI	8080
n8n	5678
Grafana	3000
Portainer	9000
pgAdmin	8000
Qdrant	6333
Ollama	11434
Lemonade	8080
WUD	3838

---

Repository Structure

deployments/
├── amd-compose-stack/          # AMD ROCm stack
├── nvidia-compose-stack/       # NVIDIA CUDA stack
└── arm64-compose-stack/        # ARM64 stack
    ├── 00-pre-flight-advisor/
    ├── 01-infra-webssh-portainer/
    ├── 02-database-postgres-pgadmin/
    ├── 03-ai-interface-ollama-openwebui-redis/
    ├── 04-automation-n8n/
    ├── 05-rag-stack-docling-qdrant-mosquitto/
    ├── 06-ai-core-lemonade/
    ├── 07-validation-stack/
    ├── 08-backup-recovery/
    ├── 09-monitoring-alerting/
    ├── 10-observability-grafana/
    ├── 11-lifecycle-wud/
    ├── 12-commercial-gateway/
    ├── 13-landing-portal/
    ├── master-deploy.sh
    └── .env.example

---

Contributing

Contributions are welcome! See CONTRIBUTING.md for branch naming, commit format, and PR guidelines.

Please use the issue templates to report bugs or request features.

---

License

MIT © 2026 TigerAI-Taiwan