pyproject.toml

[build-system]
requires = ["setuptools>=78"]
build-backend = "setuptools.build_meta"

[project]
name = "prompt-builder-pipeline"
version = "0.2.0"
description = "Pipeline NLP Bahasa Indonesia — PII, NER, Word Quality, Profanity, Konten Berisiko, Filler, Special Char, Syntax, Field-Fit"
requires-python = ">=3.10"
dependencies = []   # semua paket bersifat opsional, pilih grup sesuai kebutuhan

[project.optional-dependencies]

# Model transformer untuk NER, Syntax, dan Field-Fit (ML layer)
# Model diunduh otomatis dari HuggingFace Hub saat pertama dijalankan (~278–600 MB)
ml = [
    "transformers>=4.40",
    "torch>=2.2",             # CPU sudah cukup, GPU opsional
    "sentencepiece>=0.2",     # tokenizer XLM-RoBERTa
    "accelerate>=0.29",       # dibutuhkan transformers pipeline
    "protobuf>=4.25",         # dibutuhkan sentencepiece
    "huggingface-hub>=0.22",  # download & cache model
]

# Deteksi kata slang, alay, dan typo (Word Quality Detector)
nlp = [
    "wordfreq>=3.1",     # kosakata Indonesia dari multi-corpus
    "symspellpy>=6.9",   # spell checker O(1)
]

# Semua dependensi sekaligus (rekomendasi untuk menjalankan pipeline penuh)
full = ["prompt-builder-pipeline[ml,nlp]"]

# Tools pengembangan
dev = [
    "pytest>=8",
    "pytest-cov>=5",
    "ruff>=0.4",
]

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

[tool.ruff]
line-length = 100
target-version = "py310"