pyproject.toml

[build-system]
requires = ["setuptools>=68.0.0", "wheel"]
build-backend = "setuptools.build_meta"

[project]
name = "codewikibench"
version = "0.1.0"
description = "Benchmark for evaluating AI-generated repository documentation with hierarchical rubrics"
readme = "README.md"
requires-python = ">=3.10"
license = {text = "MIT"}
authors = [
    {name = "CodeWikiBench Contributors"}
]
keywords = ["documentation", "benchmark", "llm", "evaluation", "rubrics"]
classifiers = [
    "Programming Language :: Python :: 3",
    "Topic :: Software Development :: Documentation",
]

dependencies = [
    "pydantic>=2.7",
    "pydantic-ai>=1.0.0",
    "openai>=1.100.0",
    "python-dotenv>=1.0.0",
    "tqdm>=4.60",
    "logfire>=4.0.0",
    "mcp>=1.0.0",
    "markdown-to-json>=2.1.0",
    "PyYAML>=6.0",
    "tiktoken>=0.7.0",
    "networkx>=3.0",
]

[project.optional-dependencies]
# CLI coding agents (claude code / codex) as rubric generators and judges.
# NOTE: the dist name on this branch is "coding-agent-wrapper"; the import name is "caw".
caw = [
    "coding-agent-wrapper @ git+https://github.com/anhnh2002/caw@fix/codex-exec-robustness",
]
# Rubric quality assessment (rubrics_generator/assess_rubrics.py)
assess = [
    "numpy>=1.24",
    "scikit-learn>=1.3",
]

[project.scripts]
codewikibench = "codewikibench.cli:main"

[project.urls]
Repository = "https://github.com/FSoft-AI4Code/CodeWikiBench"

[tool.setuptools]
package-dir = {"codewikibench" = "src"}
packages = [
    "codewikibench",
    "codewikibench.docs_parser",
    "codewikibench.rubrics_generator",
    "codewikibench.judge",
    "codewikibench.tools",
]

[tool.setuptools.package-data]
codewikibench = ["*.sh"]