pyproject.toml

[project]
name = "demokratis-ml"
version = "0.1.0"
description = "Machine learning for Swiss consultation procedures"
readme = "README.md"
requires-python = ">=3.12"
dependencies = [
    "httpx>=0.27.2",
    "huggingface-hub>=0.26.3",
    "iterative-stratification>=0.1.9",
    "lingua-language-detector>=2.0.2",
    "matplotlib>=3.9.2",
    "mlflow>=2.18.0",
    "numpy>=1.26.4",
    "openai>=1.55.1",
    "pandera>=0.20.4",
    "prefect[slack]>=3.1.1",
    "pyarrow>=18.1.0",
    "pymupdf>=1.24.14",
    "python-dotenv>=1.0.1",
    "scikit-learn>=1.7.0",
    "seaborn>=0.13.2",
    "s3fs[boto3]>=2024.10.0",
    "tiktoken>=0.8.0",
    "tqdm>=4.67.1",
    "pdfplumber>=0.11.6",
    "openpyxl>=3.1.5",
    "cleanlab[datalab]>=2.7.0",
    "pandas>=2.2.3",
    "datasets>=3.3.2",
    "hyperopt>=0.2.7",
    # Setuptool added because of https://github.com/hyperopt/hyperopt/issues/939
    "setuptools>=80.9.0",
    "nltk>=3.9.1",
    "duckdb>=1.3.2",
    "xgboost>=3.1.1",
]

[dependency-groups]
dev = [
    "pre-commit>=3.8.0",
    "pytest-mock>=3.14.0",
    "pytest>=8.3.4",
    "ruff>=0.6.7",
    "ipykernel>=6.29.5",
    "ipdb>=0.13.13",
    "memray>=1.18.0",
    "docling>=2.62.0",
]

[tool.prefect]
home = "./.prefect"
server.analytics_enabled = false
api.url = "http://127.0.0.1:4200/api"

[tool.ruff]
line-length = 120

[tool.ruff.lint]
select = ["ALL"]
ignore = [
    # These errors are silenced intentionally.
    "N803",  # invalid-argument-name: to allow uppercase 'X' variables
    "PD901",  # pandas-df-variable-name
    "T201",  # print
    "S101",  # assert
    "FA102",  # future-required-type-annotation
    "COM812",  # missing-trailing-comma
    "ARG001",  # unused-function-argument: very common in sklearn-compatible functions
    "RET504",  # unnecessary-assign
    "FBT",  # boolean trap
    "EM101",  # raw-string-in-exception
    "TRY003",  # raise-vanilla-args
    "FIX",  # flake8-fixme
    "TD",  # to do
    "ERA001",  # `commented-out-code`; too strict since we're doing research?
    "D203",  # one-blank-line-before-class; we prefer `no-blank-line-before-class` (D211)
    "D212",  # `multi-line-summary-first-line`; do not enforce this little detail
    "D213",  # `multi-line-summary-second-line`; do not enforce this little detail
    "ISC001",  # `single-line-implicit-string-concatenation`; conflicts with the formatter
    "INP001",  # `implicit-namespace-package`; nonsense for some directories
    "ANN401",  # `any-type`; sometimes it's better than trying to describe the type accurately
    "TC006",  # runtime-cast-value; let's see if we really need to add the noisy quotes that often...
]

[tool.ruff.lint.per-file-ignores]
"research/*" = [
    "D",  # pydocstyle
    "PLR2004",  # magic value comparisons
]
"test_*.py" = [
    "ANN",  # annotations
    "D",  # pydocstyle
    "PLR2004",  # magic value comparisons
]
"*.ipynb" = [
    "LOG015",  # call on root logger
]