Portfolio Project LUMINAL

Self-Hosted AI Automation Platform

Workflow automation, local LLM inference, and Midnight Media Assistant — all in a containerized, GPU-accelerated environment.

View on GitHub

Key Features

Midnight Media Assistant

Custom AI assistant with 7 Python tools for intelligent media library access. Built with anti-hallucination safeguards.

Workflow Automation

Visual workflow builder with n8n for creating automated processes with custom AI nodes and integrations.

GPU Accelerated

NVIDIA GPU passthrough for accelerated AI inference. Docker Secrets for secure credential management.

Zero Trust SSO

Cloudflare Access with Google OAuth for secure authentication. Automatic user provisioning and policy-based access control.

Custom AI Assistant

Midnight Media Assistant

A custom AI assistant built on OpenWebUI with advanced prompt engineering, RAG-enabled knowledge, and real-time tool integration for media library management.

Integrated Tools
🎬 Plex 📥 Radarr 📺 Sonarr 📊 Tautulli 💬 Bazarr ⬇️ SABnzbd 🎯 Overseerr
Key Capabilities
  • Real-time library queries with accurate dates
  • Anti-hallucination prompt engineering
  • Episode synopses and detailed metadata
  • Actor/director search across library
☁️ Proxmox Cluster: Marz
Host: glados
dev
/root/LUMINAL/
OpenWebUI + Midnight
→ gemma3:12b
→ 7 Python tools
→ RAG Knowledge
← HTTP APIs →
Host: virgil
docker
/root/HELIOS/
Plex Radarr Sonarr Tautulli Bazarr SABnzbd Overseerr

System Components

🤖 AI Services
n8n
n8n Workflow Automation
OpenWebUI
OpenWebUI AI Chat Interface with RAG
Ollama
Ollama Local LLM Inference
🔧 Infrastructure
Qdrant
Qdrant Vector Database
Docker
Docker Container Orchestration
NVIDIA
NVIDIA GPU GPU Acceleration
Cloudflare
Cloudflare Access Zero Trust SSO

Available LLM Models

llama3.1:8b

4.9 GB

Fast general-purpose model

Fast Inference

gemma3:12b

8.1 GB

High-performance for complex tasks

Midnight Default

gpt-oss:20b

~20 GB

Maximum reasoning capability

Advanced

Example Deployment

Service endpoints for a typical LUMINAL deployment

n8n Automation
:5678

Workflow automation

OpenWebUI + Midnight
:3000

AI chat interface

Qdrant
:6333

Vector database

Ollama API
:11434

LLM inference API

Screenshots & Demos

OpenWebUI Midnight Chat

OpenWebUI Midnight Chat

n8n Workflow

n8n Workflow Automation

Qdrant Dashboard

Qdrant Vector Database

Home Assistant

Home Assistant Integration

Technical Skills Demonstrated

Infrastructure & DevOps

Docker, GPU passthrough, orchestration

AI & Prompt Engineering

LLM tools, RAG, anti-hallucination

Security & SSO

Docker Secrets, Cloudflare Zero Trust

Python Development

API integration, tool building