pyproject.toml

[project]
name = "pdm-project"
version = "0.1.0"
requires-python = ">=3.11"
# The base install is the slim inference core: everything needed to load a
# trained model and run predict() for anomaly detection, plus the feature
# utilities runtime.py uses. Each deployment target adds only the extras it needs:
#   inference container / batch:  pdm-project[serve]
#   edge (anomaly detection):     pdm-project[edge]
#   edge (classification / RUL):  pdm-project[edge,tabular]
#   edge (survival):              pdm-project[edge,survival]
#   training / development:       `uv sync` (default "dev" group -> [test] -> [train])
# Deployment targets install the exact versions from uv.lock (`uv export`).
dependencies = [
    "numpy>=1.24",
    "pandas>=2.0",
    "scikit-learn>=1.3",
    "scipy>=1.11",
    "joblib>=1.3",
]

[project.optional-dependencies]
# AutoGluon models: classification and RUL. Exact pin: AutoGluon only loads
# predictors saved by the same version, so every target must match training.
tabular = ["autogluon.tabular[lightgbm,catboost,xgboost]==1.6.3"]
# Survival models
survival = ["lifelines>=0.29.0", "scikit-survival>=0.23.0"]
# S3 / Parquet I/O for data loading and batch inference
aws = ["boto3>=1.28", "pyarrow>=14.0", "tqdm>=4.66"]
# SageMaker real-time endpoint and batch Processing Job container
serve = ["pdm-project[tabular,aws]", "flask>=3.0", "gunicorn>=22.0", "typing_extensions>=4.0"]
# Greengrass edge component (add tabular / survival for those formulations)
edge = ["awsiotsdk>=1.21"]
# Full training environment
train = [
    "pdm-project[tabular,survival,aws]",
    "matplotlib>=3.7",
    "shap>=0.43",
    "imbalanced-learn>=0.12",
    "optuna>=3.0",
]
test = [
    "pdm-project[train]",
    "pytest>=7.4",
    "pytest-timeout>=2.2",
]
# Future feature dependencies (not required for core)
monitoring = ["alibi-detect>=0.12"]
tsfresh = ["tsfresh>=0.20"]
deep = ["tsai>=0.3.9", "torch>=2.0", "fastai>=2.7"]
foundation = ["chronos-forecasting>=1.4", "torch>=2.0"]
transfer = ["rul-adapt>=0.4", "pytorch-lightning>=2.0"]
anomaly = ["pyod>=2.0"]

[dependency-groups]
# Synced by default by `uv sync` / `uv run`, so development and skill projects
# always get the full training environment; never installed by pip/uv export.
dev = ["pdm-project[test]"]

[build-system]
requires = ["setuptools>=69"]
build-backend = "setuptools.build_meta"

[tool.setuptools.packages.find]
include = ["pdm*"]

[tool.pytest.ini_options]
testpaths = ["tests"]
timeout = 180
addopts = "-v --tb=short"
filterwarnings = [
    "ignore::DeprecationWarning",
    "ignore::UserWarning:lifelines.*",
    "ignore::FutureWarning:autogluon.*",
]