authentik
Cyberpunk Authentik: A Neon Glassmorphism Theme
Transform Authentik default UI into a cyberpunk glassmorphism portal with AI-generated backgrounds, glowing app cards, and smooth animations.
~/authors/
ai
Setting Up Qwen3-TTS with GPU Acceleration on Ubuntu Server
A step-by-step guide to setting up Qwen3-TTS with GPU acceleration on Ubuntu Server. Complete with FastAPI wrapper, voice cloning via x-vector embeddings, and all the troubleshooting tips I wish I had when I started.
TTS
Self-Hosted Qwen3-TTS: Voice Without the Cloud
Deploy a 1.7B parameter TTS model with interesing featyures. No cloud APIs, no surprise bills, total data privacy.
authentik
Ditching Azure AD: Self-Hosted Identity with Authentik
How I replaced Microsoft Entra ID with Authentik for self-hosted SSO across Grafana, GitLab, and Portainer. Full control, better latency, zero cloud dependencies.
haproxy
HAProxy Odyssey: From NetScaler to BGP Anycast
From cutting cost-heavy Citrix nets to mastering BGP Anycast, my HAProxy journey shows how resilience, security, and automation evolve.
openclaw
Model Delegation: How OpenClaw Cuts AI Costs 10x
Claude is expensive. Local models are free. OpenClaw delegates text generation to Ollama while keeping Claude for coordination. Here's how.
DIY
Building a Smart Home Roller Shutter Controller with ESP32 and Somfy RTS
Build a DIY Somfy RTS controller for ~€15 using ESP32 and CC1101. Covers wiring, firmware, Home Assistant integration.
TTS
Running Fish Speech TTS on TrueNAS with GPU Acceleration
Setting up Fish Speech text-to-speech on TrueNAS SCALE with Docker and NVIDIA GPU support. Covers the gotchas I hit along the way.
haproxy
From Keepalived to BGP Anycast: HAProxy HA on UniFi UDM Pro Max
How I got true active-active HAProxy load balancing using BGP anycast on a UniFi UDM Pro Max. No expensive routers needed.
citrix
Automating Citrix XML SSL Certificate Binding
A production-ready PowerShell script that automates SSL certificate binding for Citrix Broker Services. Includes validation, safety checks, and WhatIf mode.
ai
Why My Mac Mini M4 Outperforms Dual RTX 3090s for LLM Inference
I built a dual RTX 3090 server for local LLM inference. A Mac Mini M4 turned out to be 27% faster and 22× more efficient. Here's why memory bandwidth beats raw GPU power.
homelab
The State of Homelabs in 2026: Smaller, Smarter, AI-Powered
RAM prices hurt, power bills hurt more, and everyone wants to run local AI. The homelab of 2026 is smaller, smarter, and completely different from what we had five years ago.