pyproject.toml

[build-system]
requires = ["setuptools>=61.0,<75.0", "wheel"]
build-backend = "setuptools.build_meta"

[project]
name = "cascadeflow"
version = "1.2.0"
description = "Agent runtime intelligence layer — optimize cost, latency, quality, budget, compliance, and energy across AI agent workflows."
readme = "README.md"
requires-python = ">=3.9"
license = {text = "MIT"}
authors = [
    {name = "Lemony Inc.", email = "hello@lemony.ai"}
]
maintainers = [
    {name = "Lemony Inc.", email = "hello@lemony.ai"}
]
keywords = [
    "ai",
    "llm",
    "cost-optimization",
    "model-routing",
    "cascade",
    "inference",
    "openai",
    "anthropic",
    "gpt",
    "claude",
    "machine-learning",
    "groq",
    "typescript",
    "javascript",
    "browser",
    "edge-functions",
    "agent-intelligence",
    "runtime-optimization",
    "budget-enforcement",
    "compliance",
    "harness",
    "agent-runtime",
    "kpi",
    "energy-tracking",
]
classifiers = [
    "Development Status :: 5 - Production/Stable",
    "Intended Audience :: Developers",
    "Programming Language :: Python :: 3",
    "Programming Language :: Python :: 3.9",
    "Programming Language :: Python :: 3.10",
    "Programming Language :: Python :: 3.11",
    "Programming Language :: Python :: 3.12",
    "Programming Language :: Python :: 3.13",
    "Topic :: Software Development :: Libraries :: Python Modules",
    "Topic :: Scientific/Engineering :: Artificial Intelligence",
    "Operating System :: OS Independent",
]

dependencies = [
    "pydantic>=2.0.0",
    "httpx>=0.25.0",
    "tiktoken>=0.5.0",
]

[project.scripts]
cascadeflow-gateway = "cascadeflow.server:main"

[project.optional-dependencies]
# Individual providers (API-based - require API keys)
openai = ["openai>=1.0.0"]
anthropic = ["anthropic>=0.8.0"]
groq = ["groq>=0.4.0"]
huggingface = ["huggingface-hub>=0.19.0"]
together = ["together>=0.2.0"]

# Local inference providers (run your own models)
# Note: Ollama doesn't need a Python package - just HTTP requests to localhost:11434
# Install Ollama separately from: https://ollama.ai
vllm = [
    "vllm>=0.14.1; python_version >= '3.10' and python_version < '3.14'"
]  # For running vLLM in-process (or use vLLM server via HTTP)

# Common API providers bundle (most popular)
providers = [
    "openai>=1.0.0",
    "anthropic>=0.8.0",
    "groq>=0.4.0",
]

# Local inference bundle
local = [
    "vllm>=0.14.1; python_version >= '3.10' and python_version < '3.14'",
]

# Semantic routing (optional feature)
semantic = [
    "sentence-transformers>=2.2.0",
    "fastembed>=0.7.0",
]

# OpenClaw integration (auto-enables FastEmbed for semantic routing)
openclaw = ["fastembed>=0.7.0"]

# Rich terminal output (optional)
rich = ["rich>=13.0.0"]

# CrewAI harness integration (opt-in)
crewai = ["crewai>=1.5.0; python_version >= '3.10'"]

# LangChain integration (opt-in)
langchain = ["langchain-core>=0.2.0"]

# OpenAI Agents SDK integration (opt-in)
openai-agents = [
    "openai-agents>=0.8.4; python_version < '3.10'",
    "openai-agents>=0.9.0; python_version >= '3.10'",
]

# Google ADK integration (opt-in, requires Python 3.10+)
google-adk = ["google-adk>=1.0.0; python_version >= '3.10'"]

# PydanticAI integration (opt-in)
pydantic-ai = ["pydantic-ai>=0.1.0; python_version >= '3.10'"]

# Development tools (includes rich for terminal output)
dev = [
    "pytest>=7.4.0",
    "pytest-asyncio>=0.21.0",
    "pytest-cov>=4.1.0",
    "pytest-mock>=3.12.0",
    "python-dotenv>=1.0.0",
    "fastembed>=0.7.0",
    "black>=23.0.0",
    "ruff>=0.1.0",
    "mypy>=1.5.0",
    "isort>=5.12.0",
    "pre-commit>=3.5.0",
    "rich>=13.0.0",
]

# Documentation tools
docs = [
    "mkdocs>=1.5.0",
    "mkdocs-material>=9.4.0",
    "mkdocstrings[python]>=0.23.0",
]

# Everything (all providers + semantic routing)
all = [
    "openai>=1.0.0",
    "anthropic>=0.8.0",
    "groq>=0.4.0",
    "huggingface-hub>=0.19.0",
    "together>=0.2.0",
    "vllm>=0.14.1; python_version >= '3.10' and python_version < '3.14'",
    "sentence-transformers>=2.2.0",
    "fastembed>=0.7.0",
    "langchain-core>=0.2.0",
]

[project.urls]
Homepage = "https://lemony.ai"
Documentation = "https://docs.cascadeflow.ai"
Repository = "https://github.com/lemony-ai/cascadeflow"
"Bug Tracker" = "https://github.com/lemony-ai/cascadeflow/issues"
Changelog = "https://github.com/lemony-ai/cascadeflow/releases"

[tool.setuptools]
packages = {find = {where = ["."], exclude = ["tests*", "examples*", "docs*", "packages*", "node_modules*", ".venv*", "build*", "dist*"]}}

[tool.setuptools.package-data]
cascadeflow = ["py.typed"]

[tool.black]
line-length = 100
target-version = ['py39']
include = '\.pyi?$'
exclude = '''
/(
    \.git
  | \.venv
  | \.idea
  | node_modules
  | \.next
  | \.turbo
  | build
  | dist
)/
'''

[tool.ruff]
line-length = 100
target-version = "py39"

[tool.ruff.lint]
select = [
    "E",   # pycodestyle errors
    "F",   # pyflakes
    "W",   # pycodestyle warnings
    "I",   # isort
    "N",   # pep8-naming
    "UP",  # pyupgrade
    "B",   # flake8-bugbear
    "A",   # flake8-builtins
    "C4",  # flake8-comprehensions
    "PT",  # flake8-pytest-style
]
ignore = [
    "E501",  # Line too long - will fix gradually
    "N801",  # Class name - cascadeflowLiteLLMCallback is external convention
    "N806",  # Variable naming - external APIs use their conventions
    "N816",  # MixedCase variable - external integrations use their conventions
    "N818",  # Exception naming - GuardrailViolation is intentional
    "B904",  # Exception chaining - will add gradually
    "E722",  # Bare except - will fix gradually
    "B007",  # Loop variable not used
    "B008",  # Function call in defaults
    "B017",  # assert raises
    "B018",  # Useless expression
    "B027",  # Empty method
    "UP007",  # Use X | Y - not available in Python 3.9 (repo supports >=3.9)
    "UP045", # Use X | None - requires Python 3.10+ without __future__ annotations
    "UP035",  # typing.Dict -> dict (will migrate gradually)
    "PT011",  # pytest.raises too broad
    "PT017",  # Found assertion on exception
    "PT021",  # Use yield instead of request.addfinalizer
]

[tool.ruff.lint.per-file-ignores]
"cascadeflow/integrations/langchain/types.py" = ["A005"]  # Filename is part of LangChain integration surface
"cascadeflow/integrations/openclaw/openai_server.py" = ["N802"]  # BaseHTTPRequestHandler uses do_GET/do_POST
"cascadeflow/proxy/server.py" = ["N802"]  # BaseHTTPRequestHandler uses do_GET/do_POST
"__init__.py" = ["F401", "E402", "I001"]  # F401: unused imports, E402: imports not at top (needed for backward compat), I001: unsorted imports
"cascadeflow/core/batch.py" = ["F401"]  # Availability check imports
"cascadeflow/core/cascade.py" = ["F401"]  # Availability check imports
"cascadeflow/integrations/litellm.py" = ["F401"]  # Availability check imports
"cascadeflow/quality/quality.py" = ["F401"]  # Availability check imports
"cascadeflow/dynamic_config/watcher.py" = ["F821"]  # ConfigManager forward reference
"cascadeflow/routing/tool_risk.py" = ["E741"]  # Ambiguous variable name in risk scoring
"cascadeflow/schema/__init__.py" = ["A004"]  # TimeoutError re-export shadows builtin
"cascadeflow/schema/domain_config.py" = ["F821"]  # ModelRegistry forward reference
"cascadeflow/schema/exceptions.py" = ["A001"]  # TimeoutError class shadows builtin (intentional)
"tests/**" = ["E402", "F401", "E741", "B023", "B006", "F841", "I001", "PT012"]  # Tests: allow flexible import grouping + complex raises blocks
"tests/_archive_development_tests/**" = ["F"]  # Archived tests: ignore all F (pyflakes) errors
"examples/**" = ["F401"]  # Examples: unused imports OK

[tool.mypy]
python_version = "3.9"
warn_return_any = false
warn_unused_configs = true
disallow_untyped_defs = false
check_untyped_defs = false
ignore_errors = false
exclude = [
    "build",
    "dist",
    ".venv",
]
# Temporarily relaxed settings - will gradually improve type coverage
disable_error_code = [
    "no-untyped-def",
    "no-any-return",
    "assignment",
    "arg-type",
    "var-annotated",
    "attr-defined",
    "operator",
    "index",
    "call-overload",
    "dict-item",
    "return-value",
    "valid-type",
    "union-attr",
    "misc",
    "override",
    "no-redef",
    "has-type",
    "annotation-unchecked",
    "name-defined",
    "import-not-found",
    "call-arg",
    "import",
]

[[tool.mypy.overrides]]
module = [
    "openai.*",
    "anthropic.*",
    "groq.*",
    "huggingface_hub.*",
    "together.*",
    "vllm.*",
]
ignore_missing_imports = true

[tool.pytest.ini_options]
testpaths = ["tests"]
python_files = ["test_*.py"]
python_classes = ["Test*"]
python_functions = ["test_*"]
addopts = [
    "-v",
    "--cov=cascadeflow",
    "--cov-report=term-missing",
    "--cov-report=html",
    "--asyncio-mode=auto",
    "--ignore=tests/_archive_development_tests",
]
markers = [
    "integration: marks tests that require external services or API keys (deselect with '-m \"not integration\"')",
    "requires_api: marks tests that require API keys to run",
    "requires_ollama: marks tests that require Ollama to be running",
    "requires_vllm: marks tests that require vLLM to be running",
    "slow: marks tests as slow (deselect with '-m \"not slow\"')",
]

[tool.coverage.run]
source = ["cascadeflow"]
omit = [
    "*/tests/*",
    "*/test_*.py",
]

[tool.coverage.report]
exclude_lines = [
    "pragma: no cover",
    "def __repr__",
    "raise AssertionError",
    "raise NotImplementedError",
    "if __name__ == .__main__.:",
    "if TYPE_CHECKING:",
]