Multi-AutoML Interface v5.2.1

Multi-AutoML Interface 5.2.1

Fixed

• Three of PyCaret's five catalog rows could not start. Verified by installing
  pycaret==3.3.2 in an isolated Python 3.11 interpreter and running
  run_pycaret_experiment for each task type:
  • Anomaly Detection and Clustering raised TypeError: setup() got an unexpected keyword argument 'fold' - the unsupervised setups have no cross-validation folds. fold is now
    passed only to the supervised ones, and those rows produce IForest and KMeans.
  • Forecast raised ValueError: Estimator naive Not Available as soon as the frame carried
    any column besides the target, because PyCaret's time series module is univariate and
    keeps only the pmdarima family available. The estimator list now follows the frame
    (_ts_include_models), and the date column the UI selects is moved into the index
    instead of being read as an exogenous feature. Both shapes - raw ordering under
    Sequential, lag features under Tabular - train to an EnsembleForecaster.
• Two PyCaret trainings in one process never finished. The functional API keeps a single
  process-global experiment, and this module also ended whichever MLflow run was active; with
  two sessions training at once both threads were stuck for minutes, while each run alone
  takes seconds. Concurrent MLflow runs were tested and are fine, so the engine is serialized:
  run_pycaret_experiment now queues behind a lock and a queued run can still be cancelled.
• The availability guard broke tests that stub the engine module. Two dispatch tests built a
  FLAML orchestrator with a fake module and hit the new "install it with: pip install flaml"
  check on interpreters without FLAML; they now declare the engine present, and a test covers
  the guard itself.

What is inside

Each installer bundles a standalone CPython 3.12 with everything in
requirements.txt already installed, so no Python setup is needed on
the target machine. Runs, models and the data lake are written to the app's
per-user workspace (the Electron userData directory):

Operating system	Location
Windows	%APPDATA%\multi-automl-desktop\workspace\
macOS	~/Library/Application Support/multi-automl-desktop/workspace/
Linux	~/.config/multi-automl-desktop/workspace/

The heavy AutoML backends (AutoGluon, PyCaret, TPOT, Lale, H2O, AutoKeras)
stay optional and are lazy-imported; the bundled runtime
contains the core stack (Streamlit, MLflow, FLAML, scikit-learn, XGBoost,
LightGBM), so the desktop installers offer FLAML until you install whichever
engines you need into it. H2O additionally requires Java 11+.

Signing

These builds are not code-signed or notarized, so SmartScreen and Gatekeeper
will warn on first launch.