Introduction: The Pivot to Agile Self-Hosting

The Backstory

I used to be a very active member in the NAS community. If you were around back then, you likely saw my guides on optimizing TrueNAS for home use. I loved that platform. I still respect it. ZFS is an incredible file system, and the stability of "Enterprise" gear is undeniable.

But eventually, I hit a roadblock.

The "Enterprise" approach comes with a cost that is hard to justify for an agile home lab: Overhead.

• RAM Hunger: ZFS loves RAM. Running a stable ZFS pool often means dedicating gigabytes of memory just to the file system before you even launch a single container.
• Complexity: Managing Docker on some of these traditional NAS OSs can feel like fighting the system. I found myself spending more time debugging permission errors and middleware conflicts than actually building cool things.

I realized I didn't want a "Storage Appliance" anymore. I wanted an Application Platform.

The Philosophy: The "Hybrid AI" Lab

I am an Open Source advocate at heart, but I am also a Pragmatist.

My goal for this guide is not just "self-hosting" for the sake of it. It is to build a High-Leverage AI Laboratory.

1. The Hardware Strategy (Small is Beautiful)

We are deploying this on a Beelink ME Mini NAS (N100).

• It sips power.
• It's quiet.
• It fits on a bookshelf.
• But it is not a supercomputer.
• It is perfect for a person starting out who can't afford a large overhead cost for services. (me)
• And let's be honest I already Own it and it needs put to good use.

2. The AI Strategy (Orchestration vs. Compute)

I faced a choice: buy a $2,000 GPU to run a local LLM that makes my lights dim when it thinks, or find a smarter way. I chose the latter.

• The Orchestrator: Our Beelink server is the "Brain." It runs the logic—n8n workflows, Databases, Vector Stores, and custom Python scripts. These are lightweight.
• The Muscle: For heavy cognitive tasks (LLM inference), we offload to Google Gemini, OpenAI, or Anthropic.

This "Hybrid" approach prevents vendor lock-in where it matters (your data, your code, your infrastructure) while leveraging the massive compute of cloud AI where it makes sense.

The Solution: Coolify on Ubuntu

This is why we are here. Coolify is the missing link. It gives us a Vercel-like, Heroku-like PaaS (Platform as a Service) experience on our own hardware.

• No more manual docker-compose.yml hell for every update.
• No more "what port was that service on?"
• No more 502 Bad Gateway errors because you forgot to restart a proxy.

It just works. And it works beautifully on standard Ubuntu.

Project Roadmap & Status

This guide is a living document. Here is where we stand:

✅ Verified & Live

• Core Infrastructure: Ubuntu 24.04 (eMMC) + RAID 1 (NVMe/SATA).
• Access: Tailscale (MagicDNS) + Cloudflare Tunnel (HTTPS).
• IT Tools: Deployed and secured with Cloudflare Access (OTP/PIN).
• Automation: n8n deployed on a custom subdomain to avoid browser warnings.

🚧 In Progress / Coming Soon

• Supabase: The open-source Firebase alternative.
• VaultWarden: Self-hosted Bitwarden (Phase 8).

🆘 Help Wanted: The GitLab Challenge

We want to self-host GitLab. It is the gold standard for DevOps.

• The Problem: GitLab is a resource hog (eats RAM for breakfast) and the configured deployment on Coolify currently has significant bugs.
• The Request: If you are a GitLab wizard or a Coolify expert, we need you. I am looking for community contributions to debug the GitLab deployment template so we can run it stably on this hardware class.

Call to Action

I built the foundation, but I want us to build the ecosystem.

1. Comment & Request: What software do you want to see deployed next? (Next.js? Drizzle? specifics?)
2. Contribute: If you successfully deployed a complex stack on Coolify, share your configuration!
3. Debug: Help us solve the GitLab puzzle.

Let's build something awesome.