Fundamentals track
What this track is
The fundamentals track is a short, ordered path through the ideas ArcFlow assumes you already know before you read longer guides or certification material. Each lesson is one concept, one minimal Python example, and one verification step. You declare agents and steps in Python; the Rust runtime in arcflow-core executes them. Nothing in this track requires an LLM API key.
The track takes about 30 minutes if you install the SDK once and run each example as you read. If you have not built the Python package yet, start with Install and build and return here when from arcflow import Agent, Workflow succeeds.
How to use it
Read the lessons in order. Each file builds on the previous one. Copy the minimal example into a local script, run it, and complete the Verify section before moving on. Examples run locally with no API keys; output is deterministic placeholder text until you add a live provider in lesson 4.
| Order | Lesson | What you learn |
|---|---|---|
| 1 | How ArcFlow thinks | Declaration vs execution, workflow specification, default runtime |
| 2 | Anatomy of an agent | name, role, instructions, validation |
| 3 | Anatomy of a workflow | Workflow(), step(), run(), WorkflowResult |
| 4 | Default runtime vs live LLM | No API key path, switching to OpenAI |
| 5 | When something fails | Configuration vs execution errors, reading messages |
Before you start
You need Python 3.9 or newer, the Rust toolchain, and a working install of the arcflow package from this repository. The Install and build page walks through maturin develop in sdk-python/ on macOS, Linux, and Windows.
Quick sanity check:
If that prints ready, open lesson 1.
After this track
| Goal | Document |
|---|---|
| Fastest end-to-end run | First workflow in five minutes |
| OpenAI and trace reading | Python quickstart |
| Guided verification with trace events | Track A: First workflow |