Prerequisites¶

Do this at home, before the workshop.

This is the one that ruins workshops

A measured cold bootstrap took 2m04s for a single laptop on good network. Multiply the image pull by a full room on conference Wi-Fi and the bootstrap block is gone. See P1.

Background¶

You're comfortable with kubectl and can read Kubernetes manifests.
Basic Git and terminal use.
No ML or AI background required.

Machine¶

Laptop with at least 8 GB RAM free and ~10 GB of disk.
Docker (or Podman) able to run containers locally.

Tools¶

Model access¶

At least one of:

Anthropic API key (ANTHROPIC_API_KEY) — what we've tested most, and what works best.
OpenAI / Codex access — works acceptably.
Ollama locally — untested, and probably not viable alongside a cluster on an 8 GB machine. Don't rely on it.

Corporate laptop?

Check before you travel that your machine can reach the API and run Docker. Proxies and MDM policies are a common blocker, and the room isn't the place to discover it.

Verify¶

git --version
docker version
kubectl version --client
k3d version
python3 --version
make --version

Or, once you've cloned the repo:

make preflight

Every line must say OK — and preflight also pre-pulls the container images, which is the part that actually saves the day. If anything fails, see Setup.