pycaret 4.0.0a3

PyCaret 4.0 — the modern, minimalistic ML control pane on top of scikit-learn.

PyCaret 4.0 — engine

The PyCaret 4.0 engine. Config-driven, stateless, notebook-friendly, built on scikit-learn 1.7+ / NumPy 2 / pandas 2 / Python 3.11+.

Paired with services/api (the FastAPI backend) and apps/web (the React UI) to form PyCaret Control Plane — a self-hosted ML platform. But the engine works standalone in a notebook too:

from pycaret.datasets import get_data
from pycaret.tasks import ClassificationExperiment

df = get_data("juice")
exp = ClassificationExperiment(target="Purchase", session_id=42).fit(df)
best = exp.compare_models().best
exp.save_model(best, "baseline")

Status

4.0.0a1 on PyPI. Published wheel name: pycaret. See the repo root docs/revamp/STATUS.md for the programme-wide status and docs/revamp/ROADMAP.md for what's next.

Layout

packages/engine/
├── pyproject.toml       # builds the `pycaret` wheel
├── pycaret/             # the package
│   ├── api/             # list_models, describe_model, describe_setup_params
│   ├── core/            # Experiment base class + results + errors
│   ├── tasks/           # 5 task subclasses (classification/regression/...)
│   ├── logging/         # BaseLogger + events + MemoryLogger
│   ├── containers/      # model-registry containers (legacy, being drained)
│   └── internal/        # god-class (being drained verb-by-verb — Phase 5)
├── tests/               # pytest suite (32 tests)
└── README.md            # this file

Dev

From the repo root:

uv sync --all-packages --all-extras          # all workspace members + extras
uv run pytest packages/engine/tests/ -q      # 32 engine tests