pyproject.toml

[project]
name = "polyglot-tutor"
version = "0.1.0"
description = "Adaptive language tutor: CEFR-aware exercises across the four skills, driven by a predictive spaced-repetition learner model."
readme = "README.md"
requires-python = ">=3.12"
license = "MIT"
authors = [{ name = "Arthur" }]
dependencies = [
    # Runtime deps only — keep the Docker image (HF Space, cpu-basic) lean.
    "gradio>=6.17,<7",
    "huggingface-hub>=1.18.0",
    "jiwer>=4.0.0",
    "numpy>=2.4.6",
    "onnxruntime>=1.26.0",
    "openai>=2.41,<3",          # single OpenAI-compatible client: Gemini / Mistral / Ollama / OpenAI (ADR 0002)
    "pydantic>=2.8",
    "pydantic-settings>=2.14,<3",
    "tokenizers>=0.22.2",
]

[dependency-groups]
dev = [
    "pytest>=9.0",
    "pytest-asyncio>=1.4",
    "ruff==0.15.16",            # exact pin, mirrored in .pre-commit-config.yaml
    "pre-commit>=4.6",
]
data = [
    # Dataset download / EDA (UniversalCEFR & friends). Not in the runtime image.
    "datasets>=5.0",
    "huggingface-hub>=1.18.0",
]
train = [
    "accelerate>=1.14.0",
    # M1: transformers, torch (CUDA, local GPU), optimum/onnxruntime for CPU export.
    "datasets>=5.0.0",
    "mlflow>=3.13.0",
    "onnx>=1.21.0",
    "torch>=2.12.0",
    "transformers>=5.11.0",
]
asr = [
    "faster-whisper>=1.0",
    "soundfile>=0.12",
]

[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"

[tool.hatch.build.targets.wheel]
packages = ["src/tutor"]

[tool.ruff]
target-version = "py312"
line-length = 100
src = ["src", "tests", "scripts"]

[tool.ruff.lint]
select = [
    "E", "W",    # pycodestyle
    "F",         # pyflakes
    "I",         # isort
    "UP",        # pyupgrade
    "B",         # bugbear
    "SIM",       # simplify
    "C4",        # comprehensions
    "PTH",       # pathlib
    "RET",       # return
    "ASYNC",     # async pitfalls
    "RUF",
]

[tool.pytest.ini_options]
testpaths = ["tests"]
asyncio_mode = "auto"
addopts = "-q"