Skip to content

1. Setup

In one glance

  • You will: Prepare the laptop environment for Part I.
  • You need: Working Python knowledge, a terminal, and internet access for installation.
  • Time: about 5 minutes, orientation.

Which pages do you need now?

Start with Python runtime installation and model configuration; the workshop introduces everything else when needed.

  1. Follow 1.0. System and run mise run install:learner.
  2. Follow 1.4. Providers for a Gemini key, optional Ollama, or offline practice.
  3. Continue to 2.1. First Agent.

Python functions, typing, imports, exceptions, virtual environments, and package installation are prerequisites. Prepare with the Python tutorial if needed. This course teaches agent development, not Python fundamentals.

Which pages can wait?

Use the other pages as references when the corresponding boundary appears.

  • 1.0. System (hands-on): install the small learner runtime; add contributor tools later.
  • 1.1. Python (reference): understand the locked Python project and preparation resources.
  • 1.2. Containers (hands-on): verify the container engine before Chapter 5.
  • 1.3. Kubernetes (reference): verify platform prerequisites before Chapter 6.
  • 1.4. Providers (hands-on): configure Gemini, or choose the explicit Ollama alternative.
  • 1.5. Workspace (hands-on): inspect the completed reference and contributor gates after the workshop.

install:learner installs one locked Python runtime. It does not install the documentation environment, development tools, MLflow evaluation extras, container images, or Kubernetes tools. Chapter 4 introduces cd agents/python && mise run install:eval only when recording evaluations. The first offline exercise checks need no provider key. An interactive Gemini conversation uses a proprietary service and consumes its quota; a free allocation is not guaranteed.

When does the harder platform part begin?

Part II begins in Chapter 5 and assumes container and Kubernetes foundations.

The gateway first runs on the laptop. Chapter 6 moves it onto local Kubernetes with kagent. Prepare with Kubernetes Basics before that transition if you cannot inspect Deployments, Services, Secrets, PVCs, probes, and NetworkPolicies.

4.8. Developer Handoff defines an independent completion point for Part I and the reference checkpoint for platform learners. You can finish the developer course without Kubernetes.

What proves this chapter worked?

Verify the worked checkpoints before making any model request.

mise run check:labs

You are done when:

  • The offline worked checkpoints pass in the locked Python runtime.
  • You know whether your interactive path uses Gemini quota or optional local inference.
  • You know which preparation pages to revisit before Part II.

Continue to 2.1. First Agent when the offline checkpoints pass; configure a provider only for interactive work.