Ran is not passed
A station finishing is one fact. Its output passing a gate is another. The design records both, and never lets a check that did not run count as a pass.
Alpha, v0.1 · built in public, one small pull request at a time
Pydantic AI Factory is a Python library for building AI factories. A factory is a repeatable production system. It takes a mission, moves the work through stations, and proves each result with evidence before a person approves it. Your agents stay ordinary Pydantic AI agents; the factory adds the record of what they produced, how it was judged and who approved it.
Install it with pip install pydantic-ai-factory. The public API may change before 1.0. These pages grow with the code: each pull request adds the page for what it builds, and every Python sample is run by the test suite.
Ran is not passed
A station finishing is one fact. Its output passing a gate is another. The design records both, and never lets a check that did not run count as a pass.
Agents are optional
A station is a typed contract. It may use one agent, several, or none. Plain Python is as welcome as a model.
A methodology is not a prompt
A methodology is a capability whose work can be checked: it records evidence as the agent works, and a gate reads it.
One meaning per word
Pydantic AI already has a “capability”. Here the word means that and nothing else. Every word the factory adds has its own page under Concepts.
The first factory is the software factory. Its first station, Design, turns a short mission into a reviewed design that a person approves. Getting started builds a small factory of your own in a few lines. Build, Quality and Release follow, one station at a time; the roadmap says what comes next.