pyproject.toml

[project]
name = "apiflow-bench"
version = "1.0.0"
description = "APIFlow-Bench — LLM benchmark for real API development work. Industry-contract action space evaluated under Inspect AI."
readme = "README.md"
license = "Apache-2.0"
requires-python = ">=3.13"
authors = [
    { name = "Zelin Wan", email = "zelin.wan@postman.com" },
]

dependencies = [
    # Eval harness
    "inspect-ai>=0.3.207",

    # Closed-source model SDKs (native, for feature parity: extended thinking, reasoning_effort, thinking budget)
    "anthropic>=0.45.0",
    "openai>=1.59.0",
    "google-genai>=0.5.0",
    "truststore>=0.9.0",  # use the OS trust store for TLS

    # Mock backends
    "fastapi>=0.115.0",
    "uvicorn>=0.32.0",
    "strawberry-graphql[fastapi]>=0.250.0",
    "mcp>=1.7.0",

    # Typed contract
    "pydantic>=2.10.0",

    # Stats
    "numpy>=2.0.0",

    # CLI
    "typer>=0.16.0",

    # YAML / templates
    "ruamel.yaml>=0.18.0",
    "jinja2>=3.1.0",
    "pyyaml>=6.0.0",

    # Misc
    "python-dotenv>=1.0.0",
    "requests>=2.32.0",
    "httpx>=0.28.0",
    "tenacity>=9.0.0",
]

[project.urls]
Repository = "https://github.com/postmanlabs/APIFlow-Bench"
Leaderboard = "https://www.postman.com/ai/apiflow-leaderboard/"

[project.optional-dependencies]
dev = ["pytest", "pytest-asyncio"]

[project.scripts]
apiflow = "apiflow_bench.cli:app"

[dependency-groups]
dev = [
    "pytest>=8.3.0",
    "pytest-asyncio>=0.24.0",
    "ruff>=0.7.0",
    "mypy>=1.13.0",
    "ipykernel>=6.29.0",
]

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

[tool.hatch.build.targets.wheel]
packages = ["src/apiflow_bench", "src/mocks"]

[tool.ruff]
line-length = 100
target-version = "py313"

[tool.ruff.lint]
select = ["E", "F", "I", "B", "UP", "RUF"]
ignore = ["E501"]  # line-length handled by formatter

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