pyproject.toml

[project]
name = "writing-agent"
dynamic = ["version"]   # single source: src/writingagent/__init__.py __version__
description = "A self-correcting, autonomous writing system for books and articles (CLI/TUI)."
readme = "README.md"
requires-python = ">=3.10"
license = "MIT"
license-files = ["LICENSE"]
authors = [{ name = "WRITING AGENT contributors" }]
keywords = ["llm", "writing", "agent", "books", "articles", "openrouter", "deepseek", "cli", "tui"]
classifiers = [
  "Development Status :: 4 - Beta",
  "Environment :: Console",
  "Intended Audience :: Developers",
  "Intended Audience :: End Users/Desktop",
  "Operating System :: OS Independent",
  "Programming Language :: Python :: 3",
  "Programming Language :: Python :: 3.10",
  "Programming Language :: Python :: 3.11",
  "Programming Language :: Python :: 3.12",
  "Programming Language :: Python :: 3.13",
  "Topic :: Text Processing",
  "Topic :: Scientific/Engineering :: Artificial Intelligence",
]
dependencies = [
  "openai>=1.0",
  "pydantic>=2.0",
  "pyyaml>=6.0",
  "python-dotenv>=1.0",
  "markdown>=3.5",
  "xhtml2pdf>=0.2.11",
  # Transitive via xhtml2pdf. svglib 1.6.0 hard-requires rlpycairo -> pycairo, and
  # pycairo ships no Linux wheels - a bare `pip install` on Linux then fails trying to
  # build it from source (needs pkg-config + libcairo2-dev). 1.5.x has no cairo chain;
  # xhtml2pdf only needs svglib's svg2rlg, not the cairo rasterizer.
  "svglib<1.6",
  "rich>=13.0",
  "pyfiglet>=1.0",
  "prompt_toolkit>=3.0",
  "ddgs>=0.1",
  "ebooklib>=0.18",
]

[project.optional-dependencies]
dev = [
  "pytest>=7.0",
  "pytest-cov>=4.0",
  "ruff>=0.4",
]
# Deep-researcher fetch backend. Optional: deep_research.py imports `scrapo` lazily and falls back to a
# stdlib urllib + html.parser fetch when absent, so the base package needs nothing here. Scrapo ships
# on PyPI as `scrapo-ai` (import name `scrapo`); it + its Playwright browser tier need Python 3.11+, so
# both are marker-gated (on 3.10 the stdlib path is always used). After install run
# `python -m playwright install chromium` for the browser binary.
deep = [
  "scrapo-ai>=0.10 ; python_version >= '3.11'",
  "playwright>=1.40 ; python_version >= '3.11'",
]
# Zero-install web demo (demo/app.py). A small Gradio front-end over the public
# Agent/Project facade: paste a topic, watch the pipeline run live, read the piece
# plus its evidence report - try it free in fake mode with no API key. Deployable
# as a Hugging Face Space (see demo/README.md) so non-developers can try it in a
# browser. The package itself never imports gradio; only the demo does.
web = [
  "gradio>=4.0",
]

[project.urls]
Homepage = "https://github.com/vikast908/WritingAgent"
Repository = "https://github.com/vikast908/WritingAgent"
Issues = "https://github.com/vikast908/WritingAgent/issues"

[project.scripts]
writing-agent = "writingagent.cli:main"

[tool.setuptools.dynamic]
version = { attr = "writingagent.__version__" }

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

[tool.setuptools.package-data]
# resources/ is the single home for all shipped data: the gold corpus
# (registers.gold_exemplars), persona exemplars (personas.block), the default
# model routing (copied out to the user config on first run), and the 13 seed
# skills (read in place). The webui SPA ships alongside its server. Data, not
# modules - packages.find alone would drop them.
writingagent = ["resources/gold/*.md", "resources/personas/*.md",
                "resources/models.yaml", "resources/seeds/skills/*.md",
                "webui/static/*.html"]

[tool.ruff]
line-length = 110
target-version = "py310"
src = ["src", "tests"]

[tool.ruff.lint]
select = ["E", "F", "W", "I", "UP", "B"]
ignore = [
  "E501",   # line length is handled by the formatter, not enforced hard
  "B008",   # function calls in argument defaults (argparse/Field patterns)
]

[tool.ruff.lint.per-file-ignores]
"tests/*" = ["E402", "B011", "E731"]
# The orchestrator package __init__ is a pure facade: it re-exports the seam
# submodules (from .common import * etc.) so `orchestrator.X` keeps resolving for
# every caller and test. Star re-exports are intentional here.
# The Colab quickstart notebook uses one-line/late imports for brevity in a demo cell.
"examples/*.ipynb" = ["E401", "E402", "I001"]
"src/writingagent/orchestrator/__init__.py" = ["F401", "F403", "F405"]
"src/writingagent/shell/__init__.py" = ["F401", "F403", "F405"]
"src/writingagent/cli/__init__.py" = ["F401", "F403", "F405"]

[tool.pytest.ini_options]
pythonpath = ["src"]   # tests import writingagent from src without an install
testpaths = ["tests"]
addopts = "-q"

[tool.coverage.run]
# Resolve coverage by package name (src layout).
source_pkgs = ["writingagent"]
branch = true

[tool.coverage.report]
# The dev journal / demo entrypoints aren't part of the library's covered surface.
omit = ["*/webui/*", "*/demo/*"]
exclude_also = ["if TYPE_CHECKING:", "raise NotImplementedError", "if __name__ == .__main__.:"]

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