Python-powered AI agents are here

Python gets in on the agentic AI revolution, and you can, too, with the new Google Agent Development Kit. Also, get started with editable installs to make local Python development easier, use the Python client API to tap into Google’s Data Commons libraries, and watch lazy annotations at work in Python 3.14. That and more—all in this week’s edition of The Python Report.

Top picks for Python readers on InfoWorld

Get started with Google Agent Development KitGoogle Agent Toolkit empowers Python and Java users to build AI agents with minimal hassle. This tutorial gets you started building Python-based agents with the LLM of your choice, including your own model.

Python 3.14 changes type hints forever: Lazy annotations explainedAs of Python 3.14, type annotations are lazy by default, so you no longer need to define a type before annotating something with it. Watch lazy annotations at work in this live example.

How to use editable installs for Python packagesIf you create and install your own Python packages on local systems, you know the pain of reinstalling them every time you make a change. Why not let editable installs solve that problem for you?

Google touts Python client library for Data CommonsThe new Python client library makes it easier than ever to access over 200 publicly available datasets hosted on Google’s Data Commons Platform.

More good reads and Python updates elsewhere

Hold the presses! Free-threaded Python is officially supportedUntil recently, the free-threaded Python builds were considered experimental. As of Python 3.14 beta 3, they’re an official but optional part of Python.

‘Automate the boring stuff with Python’: The lost chapterOne chapter of Al Sweigart’s best-selling guide to learning Python by building things didn’t make it to the final cut of the current edition. That lost chapter, all about working with multimedia in Python, is now online.

PhotoshopAPI: An open source C++ library for manipulating Photoshop files with PythonFinally, there’s an open source, liberally licensed toolkit for manipulating Photoshop images programmatically from Python! Some features are coming later (adjustment layers and text layers), and some won’t ever be added (multichannel/Lab color), but the PhotoshopAPI claims far faster performance than Photoshop’s native API.

Reflections on two years of CPython’s JIT CompilerCPython’s JIT Compiler is one of the key components to making Python faster. Ken Jin, a lead developer on the project, reflects on the development process, what’s ahead, and why the compiler hasn’t delivered on its performance promises so far.