haystack/pyproject.toml

[build-system]
requires = [
  "hatchling>=1.8.0",
]
build-backend = "hatchling.build"

[project]
name = "farm-haystack"
dynamic = [
  "version",
]
description = "LLM framework to build customizable, production-ready LLM applications. Connect components (models, vector DBs, file converters) to pipelines or agents that can interact with your data."
readme = "README.md"
license = "Apache-2.0"
requires-python = ">=3.8"
authors = [
  { name = "deepset.ai", email = "malte.pietsch@deepset.ai" },
]
keywords = [
  "BERT",
  "QA",
  "Question-Answering",
  "Reader",
  "Retriever",
  "albert",
  "language-model",
  "mrc",
  "roberta",
  "search",
  "semantic-search",
  "squad",
  "transfer-learning",
  "transformer",
]
classifiers = [
  "Development Status :: 5 - Production/Stable",
  "Intended Audience :: Science/Research",
  "License :: Freely Distributable",
  "License :: OSI Approved :: Apache Software License",
  "Operating System :: OS Independent",
  "Programming Language :: Python",
  "Programming Language :: Python :: 3",
  "Programming Language :: Python :: 3.8",
  "Programming Language :: Python :: 3.9",
  "Programming Language :: Python :: 3.10",
  "Topic :: Scientific/Engineering :: Artificial Intelligence",
]
dependencies = [
  "requests",
  "pydantic<2",
  "transformers==4.32.1",
  "pandas",
  "rank_bm25",
  "scikit-learn>=1.3.0", # TF-IDF and metrics
  "lazy-imports==0.3.1", # Optional imports
  "prompthub-py==4.0.0",
  "platformdirs",

  # Utils
  "tqdm",  # progress bars in model download and training scripts
  "networkx",  # graphs library
  "quantulum3",  # quantities extraction from text
  "posthog",  # telemetry
  # audio's espnet-model-zoo requires huggingface-hub version <0.8 while we need >=0.5 to be able to use create_repo in FARMReader
  "tenacity",  # retry decorator
  "sseclient-py",  # server side events for OpenAI streaming
  "more_itertools",  # utilities

  # Web Retriever
  "boilerpy3",

  # Multimodal Embedder haystack/nodes/retriever/multimodal/embedder.py
  "Pillow",

  # OpenAI tokenizer
  "tiktoken>=0.5.1",

  # Schema validation
  "jsonschema",

  # Preview
  "canals==0.8.0",
  "openai",
  "Jinja2",
  "openai-whisper",  # FIXME https://github.com/deepset-ai/haystack/issues/5731

  # Agent events
  "events",

  "requests-cache<1.0.0",
]

[project.optional-dependencies]
inference = [
  "transformers[torch,sentencepiece]==4.32.1",
  "sentence-transformers>=2.2.0",  # See haystack/nodes/retriever/_embedding_encoder.py, _SentenceTransformersEmbeddingEncoder
  "huggingface-hub>=0.5.0",
]
elasticsearch = [
  "farm-haystack[elasticsearch7]",
]
elasticsearch7 = [
  "elasticsearch>=7.17,<8",
  "elastic_transport<8"
]
elasticsearch8 = [
  "elasticsearch>=8,<9",
  "elastic_transport>=8,<9"
]
sql = [
  "sqlalchemy>=1.4.2,<2",
  "sqlalchemy_utils",
  "psycopg2-binary; platform_system != 'Windows'",
]
only-faiss = [
  "faiss-cpu>=1.6.3,<=1.7.2",
]
faiss = [
  "farm-haystack[sql,only-faiss]",
]
only-faiss-gpu = [
  "faiss-gpu>=1.6.3,<2",
]
faiss-gpu = [
  "farm-haystack[sql,only-faiss-gpu]",
]
weaviate = [
  "weaviate-client>2",
]
only-pinecone = [
  "pinecone-client>=2.0.11,<3",
]
pinecone = [
  "farm-haystack[sql,only-pinecone]",
]
opensearch = [
  "opensearch-py>=2",
]
docstores = [
  "farm-haystack[elasticsearch,faiss,weaviate,pinecone,opensearch]",
]
docstores-gpu = [
  "farm-haystack[elasticsearch,faiss-gpu,weaviate,pinecone,opensearch]",
]
audio = [
  "openai-whisper"
]
beir = [
  "beir; platform_system != 'Windows'",
]
aws = [
  "boto3",
  # Costraint botocore to avoid taking to much time to resolve the dependency tree.
  # boto3 used to constraint it at this version more than a year ago. To avoid breaking
  # people using old versions we use a similar constraint without upper bound.
  # https://github.com/boto/boto3/blob/dae73bef223abbedfa7317a783070831febc0c90/setup.py#L16
  "botocore>=1.27",
]
crawler = [
  "selenium>=4.11.0"
]
preprocessing = [
  "nltk",
  "langdetect",  # for language classification
]
file-conversion = [
  "azure-ai-formrecognizer>=3.2.0b2",  # Microsoft Azure's Form Recognizer service (text and table exctrator)
  "python-docx",
  "tika",  # Apache Tika (text & metadata extractor)
  "beautifulsoup4",
  "markdown",
  "python-frontmatter",
  "python-magic; platform_system != 'Windows'",  # Depends on libmagic: https://pypi.org/project/python-magic/
  "python-magic-bin; platform_system == 'Windows'",  # Needs to be installed without python-magic, otherwise Windows CI gets stuck.
]
pdf = [
  "PyMuPDF>=1.18.16" ,  # PDF text extraction alternative to xpdf; please check AGPLv3 license
]
ocr = [
  "pytesseract>0.3.7",
  "pdf2image>1.14",
]
onnx = [
  "onnxruntime",
  "onnxruntime_tools",
]
onnx-gpu = [
  "onnxruntime-gpu",
  "onnxruntime_tools",
]
metrics = [  # for metrics
  "scipy>=1.3.2",
  "rapidfuzz>=2.0.15,<2.8.0",   # FIXME https://github.com/deepset-ai/haystack/pull/3199
  "seqeval",
  "mlflow",
]
ray = [
  "ray[serve]>=1.9.1,<2; platform_system != 'Windows'",
  "ray[serve]>=1.9.1,<2,!=1.12.0; platform_system == 'Windows'",  # Avoid 1.12.0 due to https://github.com/ray-project/ray/issues/24169 (fails on windows)
  "aiorwlock>=1.3.0,<2",
]
colab = [
  "pillow<=9.0.0",
]
dev = [
  "pre-commit",
  # Type check
  "mypy",
  # Test
  "pytest",
  "pytest-cov",
  "pytest-custom_exit_code",  # used in the CI
  "pytest-asyncio",
  "responses",
  "tox",
  "coverage",
  "python-multipart",
  "psutil",
  # Linting
  "pylint",
  "farm-haystack[formatting]",
  # Documentation
  "pydoc-markdown",
  "mkdocs",
  "jupytercontrib",
  "watchdog",
  "toml",
  "reno",
  # dulwich is a reno dependency, they pin it at >=0.15.0 so pip takes ton of time to resolve the dependency tree.
  # We pin it here to avoid taking too much time.
  # https://opendev.org/openstack/reno/src/branch/master/requirements.txt#L7
  "dulwich>=0.21.0,<1.0.0",
]

formatting = [
  # Version specified following Black stability policy:
  # https://black.readthedocs.io/en/stable/the_black_code_style/index.html#stability-policy
  "black[jupyter]~=23.0",
]

all = [
  "farm-haystack[inference,docstores,audio,crawler,preprocessing,file-conversion,pdf,ocr,ray,onnx,beir,metrics,aws]",
]
all-gpu = [
  # beir is incompatible with faiss-gpu: https://github.com/beir-cellar/beir/issues/71
  "farm-haystack[inference,docstores-gpu,audio,crawler,preprocessing,file-conversion,pdf,ocr,ray,onnx-gpu,metrics,aws]",
]

[project.scripts]
haystack = "haystack.cli.entry_point:main"

[project.urls]
"CI: GitHub" = "https://github.com/deepset-ai/haystack/actions"
"Docs: RTD" = "https://haystack.deepset.ai/overview/intro"
"GitHub: issues" = "https://github.com/deepset-ai/haystack/issues"
"GitHub: repo" = "https://github.com/deepset-ai/haystack"
Homepage = "https://github.com/deepset-ai/haystack"

[tool.hatch.version]
path = "VERSION.txt"
pattern = "(?P<version>.+)"

[tool.hatch.metadata]
allow-direct-references = true

[tool.hatch.build.targets.sdist]
include = [
  "/haystack",
  "/VERSION.txt",
]

[tool.hatch.build.targets.wheel]
packages = [
  "haystack",
]

[tool.black]
line-length = 120
skip_magic_trailing_comma = true  # For compatibility with pydoc>=4.6, check if still needed.

[tool.codespell]
ignore-words-list = "ans,astroid,nd,ned,nin,ue"
quiet-level = 3
skip = "test/nodes/*,test/others/*,test/samples/*"

[tool.pylint.'MESSAGES CONTROL']
max-line-length=120
load-plugins = "haystack_linter"
disable = [

  # To keep
  "fixme",
  "c-extension-no-member",

  # To review:
  "missing-docstring",
  "unused-argument",
  "no-member",
  "line-too-long",
  "protected-access",
  "too-few-public-methods",
  "raise-missing-from",
  "invalid-name",
  "duplicate-code",
  "arguments-differ",
  "consider-using-f-string",
  "no-else-return",
  "attribute-defined-outside-init",
  "super-with-arguments",
  "redefined-builtin",
  "abstract-method",
  "unspecified-encoding",
  "unidiomatic-typecheck",
  "no-name-in-module",
  "consider-using-with",
  "redefined-outer-name",
  "arguments-renamed",
  "unnecessary-pass",
  "broad-except",
  "unnecessary-comprehension",
  "subprocess-run-check",
  "singleton-comparison",
  "consider-iterating-dictionary",
  "undefined-loop-variable",
  "consider-using-in",
  "bare-except",
  "unexpected-keyword-arg",
  "simplifiable-if-expression",
  "use-list-literal",
  "broad-exception-raised",

  # To review later
  "cyclic-import",
  "import-outside-toplevel",
  "deprecated-method",
]
[tool.pylint.'DESIGN']
max-args = 38  # Default is 5
max-attributes = 27  # Default is 7
max-branches = 34  # Default is 12
max-locals = 45  # Default is 15
max-module-lines = 2468  # Default is 1000
max-nested-blocks = 7  # Default is 5
max-statements = 206  # Default is 50
[tool.pylint.'SIMILARITIES']
min-similarity-lines=6

[tool.pytest.ini_options]
minversion = "6.0"
addopts = "--strict-markers"
markers = [
  "unit: unit tests",
  "integration: integration tests",

  "generator: generator tests",
  "summarizer: summarizer tests",
  "embedding_dim: uses a document store with non-default embedding dimension (e.g @pytest.mark.embedding_dim(128))",

  "tika: requires Tika container",
  "parsr: requires Parsr container",
  "ocr: requires Tesseract",

  "elasticsearch: requires Elasticsearch container",
  "weaviate: requires Weaviate container",
  "pinecone: requires Pinecone credentials",
  "faiss: uses FAISS",
  "opensearch",
  "document_store",
]
log_cli = true

[tool.mypy]
warn_return_any = false
warn_unused_configs = true
ignore_missing_imports = true
plugins = [
  "pydantic.mypy",
]

[tool.ruff]
select = [
  "AIR",    # Airflow
  "ASYNC",  # flake8-async
  "C90",    # McCabe cyclomatic complexity
  "CPY",    # flake8-copyright
  "DJ",     # flake8-django
  "E501",   # Long lines
  "EXE",    # flake8-executable
  "FURB",   # refurb
  "INT",    # flake8-gettext
  "PL",     # Pylint
  "Q",      # flake8-quotes
  "SLOT",   # flake8-slots
  "T10",    # flake8-debugger
  "W",      # pycodestyle
  "YTT",    # flake8-2020
  # "E",    # pycodestyle
  # "F",    # Pyflakes
  # "NPY",  # NumPy-specific rules
  # "PD",   # pandas-vet
  # "PERF", # Perflint
  # "PT",   # flake8-pytest-style
  # "UP",   # pyupgrade
]
line-length = 1486
target-version = "py38"
ignore = [
  "PLR1714",  # repeated-equality-comparison
  "PLR5501",  # collapsible-else-if
  "PLW0603",  # global-statement
  "PLW1510",  # subprocess-run-without-check
  "PLW2901",  # redefined-loop-name
]

[tool.ruff.mccabe]
max-complexity = 28

[tool.ruff.pylint]
allow-magic-value-types = ["float", "int", "str"]
max-args = 38  # Default is 5
max-branches = 32  # Default is 12
max-returns = 9  # Default is 6
max-statements = 105  # Default is 50

[tool.coverage.run]
omit = [
    "haystack/testing/*",
]