Everything you write in a pipeline — scripts, model trainers, plot generators — runs inside a managed Python environment. You do not build or maintain it, but you do choose two things about it: which Python version a pipeline runs on, and which packages your code can import.

Python versions

Several Python versions are available at once, so a new one can arrive without breaking the pipelines built on the previous one. The versions your tenant offers are listed on the AI Lab overview; a pipeline is pinned to one, and a run can be sent to a specific version.

Moving an existing pipeline forward is deliberate work rather than a switch: scripts are re-registered against the new version, the pipeline is tried as an experiment, and only then redeployed — and an edge deployment needs a new pipeline image built on that version. Upgrade your solution Python version walks through the whole sequence.

📘

A version upgrade is the moment library behaviour changes under your scripts. A pandas or scikit-learn call that was merely deprecated on the old version may be gone on the new one, which is exactly what trying the pipeline as an experiment is for.

Custom packages

Beyond the libraries Wizata ships, an administrator can make your own packages importable from PyPI or from GitHub — a shared set of in-house helpers, a vendor SDK, anything your scripts need. Once installed, they are imported in a script step like any other dependency.

Import custom packages to the platform covers installing from both sources and verifying the import works before you build a pipeline on it.