pyproject.toml

[build-system]
requires = ["setuptools>=77", "wheel"]
build-backend = "setuptools.build_meta"

[project]
name = "hmm-studio"
version = "1.1.0"
description = "Constrained Baum-Welch HMM engine with a visual topology editor and web UI."
readme = "README.md"
requires-python = ">=3.11"
authors = [{name = "Robin Denis", email = "robin.denis1207@gmail.com"}]
license = "MIT"
keywords = [
    "hmm", "hidden-markov-model", "baum-welch", "viterbi",
    "constrained-em", "topology", "nhmm", "sequence-modeling",
    "bayesian", "dirichlet-prior", "fastapi", "react",
]
classifiers = [
    "Development Status :: 5 - Production/Stable",
    "Environment :: Console",
    "Environment :: Web Environment",
    "Intended Audience :: Science/Research",
    "Intended Audience :: Developers",
    "Operating System :: OS Independent",
    "Programming Language :: Python :: 3",
    "Programming Language :: Python :: 3.11",
    "Programming Language :: Python :: 3.12",
    "Programming Language :: Python :: 3.13",
    "Topic :: Scientific/Engineering",
    "Topic :: Scientific/Engineering :: Artificial Intelligence",
    "Topic :: Scientific/Engineering :: Information Analysis",
    "Topic :: Scientific/Engineering :: Mathematics",
    "Typing :: Typed",
]
dependencies = [
    "numpy>=1.26",
    "pandas>=2.0",
    "hmmlearn>=0.3,<0.4",
    "scikit-learn>=1.3",
    "pyyaml>=6.0",
    "typer[all]>=0.12",
    "pyarrow>=15.0",
]

[project.optional-dependencies]
web = [
    "fastapi>=0.115",
    "uvicorn[standard]>=0.30",
    "sqlmodel>=0.0.16",
    "python-multipart>=0.0.9",
    "httpx>=0.27",
    "openpyxl>=3.1",   # required for pd.read_excel on .xlsx files (warehouse formats)
]
dcor = ["dcor>=0.6"]
dev = [
    "pytest>=8.0",
    "pytest-cov>=5.0",
    "pytest-timeout>=2.3",  # per-test timeout; turns CI hangs into clean fails
    "hypothesis>=6.100",
    "ruff>=0.5",
    "black>=24.0",
]
docs = [
    "mkdocs>=1.6",
    "mkdocs-material>=9.5",
]
bayesian = [
    # Phase A.6 — BayesianHMMBackend via NUTS sampling.
    # Heavy ~150 MB install ; opt-in. Required only if you import
    # hmm_core.backends.bayesian_backend.BayesianHMMBackend.
    "pymc>=6.0",
    "arviz>=1.0",
]

[project.urls]
Homepage = "https://github.com/RoJLD/HMMstudio"
Repository = "https://github.com/RoJLD/HMMstudio"
Issues = "https://github.com/RoJLD/HMMstudio/issues"
Documentation = "https://rojld.github.io/HMMstudio/"
Changelog = "https://github.com/RoJLD/HMMstudio/blob/main/CHANGELOG.md"

[project.scripts]
hmm-fit = "hmm_core.cli:app"
hmm-studio = "hmm_studio.cli:app"

[tool.setuptools.packages.find]
where = ["src"]

[tool.setuptools.package-data]
hmm_studio = ["server/static/**/*"]

[tool.pytest.ini_options]
testpaths = ["tests"]
python_files = ["test_*.py"]
# `--timeout=180` is a defensive ceiling: the slowest legitimate test is the
# Bayesian NUTS suite at ~30-60s. Setting 180s turns CI hangs (e.g. fork-on-Linux
# coverage deadlocks with ProcessPoolExecutor) into clean test failures instead
# of 25-min runner waste. Tests can override with @pytest.mark.timeout(N).
addopts = "-ra --strict-markers --timeout=180"
markers = [
    "slow: tests that take >10s (Bayesian NUTS sampling, recovery tests, etc.). Skip with -m 'not slow'.",
]

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

[tool.ruff.lint.per-file-ignores]
# `with pm.Model() as model:` is the canonical PyMC idiom — the variable
# is bound for the `with` block and downstream `pm.*` calls read it
# implicitly via the model-stack, even though no name appears in the inner
# scope. F841 (assigned-but-never-used) is a false positive here.
"src/hmm_core/backends/bayesian_backend.py" = ["F841"]
# Local imports colocated with their single consumer (auto_grid + test
# helpers). Hoisting them to the top would obscure the pairing.
"src/hmm_core/selection.py" = ["E402"]
"tests/test_selection.py" = ["E402"]

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