Skip to content

Onboarding

You've been invited to use the homelab LLM gateway. This page gets you from zero to a working request.

What you get

  • An API key (sk-...) issued by Chris.
  • Access to chat / embeddings / transcription via https://api.chris.hellotopia.io/v1.
  • Optional: a WireGuard VPN profile if you need to hit the DGX Spark directly for something non-API (e.g. SSH, hardware diagnostics). Most users never need this.

Day-1 checklist

  1. Save your API key in a password manager. Do not commit it to git.

  2. Set environment variables on your machine:

    # add to ~/.zshrc or ~/.bashrc
export OPENAI_API_KEY="sk-..."
export OPENAI_BASE_URL="https://api.chris.hellotopia.io/v1"

  3. Verify with a curl from Getting Started.

  4. Configure your IDE per IDE Integration.
  5. Pick a default model — start with llama3.1:8b for general chat, coder/qwen2.5-coder:14b for code.

Optional: WireGuard VPN

Only needed if Chris explicitly sets you up for direct Spark access.

  1. Install WireGuard:
  2. macOS / iOS — App Store.
  3. Linux — sudo dnf install wireguard-tools / sudo apt install wireguard.

  4. Windows — wireguard.com/install/.

  5. Import the .conf Chris sent you.

  6. Toggle on. Verify with:

    ping 192.168.5.55

The VPN is scoped — you can only reach the Spark, not the rest of the LAN, not the linux box's SSH.

Etiquette

  • Don't share your key. If you need a second key (e.g. for a separate project), ask Chris.
  • Don't batch huge workloads without asking. A 100k-document embedding job will saturate the 5080 for everyone else. Coordinate first.
  • Let Chris know if it's down. No monitoring / pager system yet.
  • Report weird behavior — bad tokens, hangs, 5xx spikes. Small userbase; signals are real.

Getting help