<aside> 📘

Tutorial 5 of the Kalix tutorial series. You'll drive a Kalix simulation directly from Python using the kalix package, read the outputs as pandas DataFrames, and plot them. Expected time: about 20 minutes.

</aside>

What you'll build

A Jupyter notebook that sits next to your model file and:

• Runs the simulation with one line: kalix.simulate(...)
• Reads the output into a pandas DataFrame
• Plots simulated vs observed flow
• (Bonus) Switches the output to Kalix's native Pixie format for faster, smaller files

This is the foundation for any analysis, scripting, or notebook-driven workflow built on top of Kalix.

Prerequisites

• Tutorial 3 complete (or its concepts) — see Tutorial 3 — Relative paths and trailhead paths. We start from a project laid out the same way: shared data/ with a model in models/baseline/.
• Python 3.9 or newer with Jupyter installed.
• Tutorial files — download the 005/ folder from the KalixTutorials repository. The repo also ships a fully-worked analysis.ipynb if you'd rather skim than type along.

Project layout

005/
├── data/
│   ├── climate_data.csv
│   ├── observed.csv
│   ├── rain_north.csv
│   ├── rain_central.csv
│   └── rain_south.csv
└── models/
    └── baseline/
        ├── stringybark.ini     ← the model (uses trailhead paths)
        └── analysis.ipynb      ← the notebook we'll write

The notebook sits next to the model file. That's deliberate — keep your scripts and analysis alongside the artefact they operate on so relative paths stay simple.

Step 1 — Install the kalix package

pip install kalix

This installs the kalix Python package, which bundles the Kalix simulation engine as a native extension. You don't need to install the Kalix CLI separately — the engine runs in-process from Python.

Verify it works: