pyproject.toml

[build-system]
requires = ["setuptools>=68", "wheel"]
build-backend = "setuptools.build_meta"

[project]
name = "leadforge"
version = "1.0.0"
description = "Opinionated framework for generating synthetic CRM and GTM datasets from simulated commercial worlds"
readme = "README.md"
license = { text = "MIT" }
requires-python = ">=3.11"
authors = [{ name = "leadforge contributors" }]
keywords = ["synthetic data", "CRM", "lead scoring", "machine learning", "simulation"]
classifiers = [
    "Development Status :: 5 - Production/Stable",
    "Intended Audience :: Developers",
    "Intended Audience :: Education",
    "Intended Audience :: Science/Research",
    "License :: OSI Approved :: MIT License",
    "Programming Language :: Python :: 3",
    "Programming Language :: Python :: 3.11",
    "Programming Language :: Python :: 3.12",
    "Topic :: Scientific/Engineering :: Artificial Intelligence",
]
dependencies = [
    "typer[all]>=0.12",
    "pyyaml>=6.0",
    "pandas>=2.0",
    "pyarrow>=14.0",
    "networkx>=3.2",
    "numpy>=1.26",
]

[project.optional-dependencies]
dev = [
    "pytest>=8.0",
    "ruff>=0.4",
    "mypy>=1.10",
    "pre-commit>=3.7",
    "types-pyyaml>=6.0",
    # scikit-learn 1.9.0 changed HistGradientBoostingClassifier defaults in a
    # way that improves the flat GBM more than the engineered one, making the
    # headline GBM(eng)−GBM(flat) lift in NB02 go non-positive and breaking
    # the G13.1 CI gate.  Pin <1.9 while the notebook targets are recalibrated
    # or the engineered features are updated to restore positive lift on 1.9.
    # See: https://github.com/leadforge-dev/leadforge/issues/114
    "scikit-learn>=1.3,<1.9",
    "matplotlib>=3.7",
    # PR 7.2: the preview-page renderers (scripts/preview_{kaggle,hf}_page.py)
    # call into markdown-it-py at test time via render_*_html().  Keeping
    # the dep here as well as in [publish] means CI's "test" job (which
    # installs only [dev]) does not ImportError mid-test.  pytest.importorskip
    # would also work, but the rendering tests are the primary coverage of
    # this PR — gating them off would defeat the purpose.
    "markdown-it-py>=3.0",
]
scripts = [
    "scikit-learn>=1.3,<1.9",  # keep in sync with [dev] pin above
    "matplotlib>=3.7",
]
# Optional dependencies for the platform release packagers.  Installing
# this extra (``pip install -e ".[publish]"``) enables the gated
# ``load_dataset()`` / Kaggle-CLI smoke tests that verify G11.3 (Kaggle
# package) and G12.3 / G12.4 (HF load_dataset round-trip) without
# pulling the heavy SDKs into the default dev install.  PR 7.2 adds
# ``markdown-it-py`` for the local Kaggle / HF preview pages
# (``scripts/preview_{kaggle,hf}_page.py``) — same publish-extra
# posture, missing import raises a clean error pointing at this extra.
publish = [
    "datasets>=2.14",
    "kaggle>=1.6",
    "markdown-it-py>=3.0",
]
# Optional dependencies for executing the public release notebooks.
# Installing this extra (``pip install -e ".[notebooks]"``) enables the
# G13.1 acceptance gate: the CI ``notebooks`` job nbclient-executes
# ``release/notebooks/*.ipynb`` end-to-end against a freshly built
# public bundle, asserting the notebooks reproduce validation_report.md
# metrics within ±0.05 (G13.2) and never load instructor artefacts
# (G13.3).
notebooks = [
    "nbclient>=0.10",
    "nbformat>=5.10",
    # ``ipykernel`` provides the ``python3`` kernelspec that
    # ``nbclient.NotebookClient(..., kernel_name="python3")`` looks up.
    # Without it CI fails with ``NoSuchKernel: No such kernel named
    # python3`` because the GitHub-hosted runner has no kernelspecs
    # registered out of the box (local dev environments usually do).
    "ipykernel>=6.0",
    "scikit-learn>=1.3,<1.9",  # keep in sync with [dev] pin above
    "matplotlib>=3.7",
]

[project.scripts]
leadforge = "leadforge.cli.main:app"

[tool.setuptools.packages.find]
where = ["."]
include = ["leadforge*"]

[tool.setuptools.package-data]
"leadforge" = ["**/*.yaml", "**/*.yml"]

[tool.ruff]
target-version = "py311"
line-length = 100

[tool.ruff.lint]
select = ["E", "F", "I", "N", "W", "UP", "B", "C4", "PT", "S"]

[tool.ruff.lint.per-file-ignores]
"tests/**/*" = ["S101", "S108"]
# Release notebooks deliberately use ``assert`` for contract checks
# (path discipline, snapshot-safe joins, G13.2 tolerance gate).  ``assert``
# is the conventional notebook idiom and these are exactly the cells we
# want to fail loud on regression.
"release/notebooks/**/*.ipynb" = ["S101"]
# Notebook builders emit dedented heredocs whose lines render as
# markdown tables and print-statement output inside the notebook.
# Line length is a property of the rendered cell, not the .py source,
# so 100c is the wrong yardstick here.
"scripts/build_release_notebook_*.py" = ["E501"]
# Preview-page scripts (PR 7.2) carry inlined CSS + multi-attribute
# HTML strings inside f-string templates; the rendered HTML is the
# product, so wrapping the source CSS at 100c is line noise.
"scripts/preview_kaggle_page.py" = ["E501"]
"scripts/preview_hf_page.py" = ["E501"]
# _preview_common is plain Python (no inline HTML / CSS); leaving
# E501 enabled.

[tool.mypy]
python_version = "3.11"
warn_return_any = true
warn_unused_configs = true
disallow_untyped_defs = true
disallow_incomplete_defs = true
check_untyped_defs = true
no_implicit_optional = true

[[tool.mypy.overrides]]
module = ["pandas", "pandas.*", "pyarrow", "pyarrow.*"]
ignore_missing_imports = true

[[tool.mypy.overrides]]
module = ["networkx", "networkx.*"]
ignore_missing_imports = true

[[tool.mypy.overrides]]
module = ["sklearn", "sklearn.*"]
ignore_missing_imports = true

[[tool.mypy.overrides]]
module = ["matplotlib", "matplotlib.*"]
ignore_missing_imports = true

# Anthropic SDK is loaded lazily inside ``build_anthropic_client`` (PR
# 7.1) so the LLM critique module imports cleanly without the SDK.  CI's
# type-check job doesn't install ``anthropic``; the override stops mypy
# from failing on the missing import stub.  The runtime contract is
# enforced by tests via the ``LLMCritiqueClient`` protocol.
[[tool.mypy.overrides]]
module = ["anthropic", "anthropic.*"]
ignore_missing_imports = true

[tool.pytest.ini_options]
testpaths = ["tests"]