mirror of
https://github.com/microsoft/data-formulator.git
synced 2026-11-04 12:01:37 +00:00
138 lines
4.5 KiB
TOML
138 lines
4.5 KiB
TOML
[build-system]
|
|
requires = [ "setuptools >= 75.0" ]
|
|
build-backend = "setuptools.build_meta"
|
|
|
|
[project]
|
|
name = "data_formulator"
|
|
version = "0.8.0b1"
|
|
|
|
requires-python = ">=3.11"
|
|
authors = [
|
|
{name = "Chenglong Wang", email = "chenglong.wang@microsoft.com"},
|
|
{name = "Dan Marshall", email = "danmar@microsoft.com"},
|
|
]
|
|
readme = "README.md"
|
|
license = "MIT"
|
|
description = "Data Formulator is a research prototype for data visualization powered by AI."
|
|
keywords = ["data visualization", "LLM", "AI"]
|
|
classifiers = [
|
|
"Development Status :: 4 - Beta",
|
|
"Programming Language :: Python"
|
|
]
|
|
|
|
dependencies = [
|
|
"pandas",
|
|
"flask",
|
|
"flask-limiter",
|
|
"openai",
|
|
"python-dotenv",
|
|
# litellm >=1.92 is a native Rust/maturin build; 1.99+ ships wheels for
|
|
# macOS x64/arm64, Windows x64 and manylinux/musllinux. 1.106 imports its MCP server
|
|
# on every Responses call, which needs `mcp` 2.x (below) to be installed.
|
|
"litellm>=1.106,<1.107",
|
|
# MCP client for user-added MCP servers (local mode); also satisfies litellm's import.
|
|
"mcp>=2.3,<3",
|
|
"aiohttp>=3.14.4",
|
|
"duckdb",
|
|
"numpy",
|
|
"vl-convert-python",
|
|
"beautifulsoup4",
|
|
"scikit-learn",
|
|
"pyyaml",
|
|
"jsonschema>=4.18",
|
|
# 25.0.0 bundles mimalloc 3.3.1, which crashes when pyarrow is first
|
|
# loaded on a thread that exits (apache/arrow#50471); fixed in 25.0.1.
|
|
"pyarrow>=13.0.0,!=25.0.0",
|
|
"xlrd",
|
|
"openpyxl>=3.1.0",
|
|
"lxml", # pandas.read_html web table extraction
|
|
"yfinance",
|
|
# Data-loader deps (all included by default; try/except keeps things safe)
|
|
"pymongo>=4.18.2", # mongodb
|
|
"azure-cosmos", # cosmosdb
|
|
"azure-storage-blob", # azure_blob
|
|
"azure-identity", # azure_blob, kusto
|
|
"azure-kusto-data", # kusto
|
|
"pymysql", # mysql
|
|
"clickhouse-connect", # clickhouse
|
|
"mssql-python>=1.11.0", # mssql
|
|
"psycopg2-binary", # postgresql
|
|
"boto3", # s3, athena
|
|
"google-cloud-bigquery", # bigquery
|
|
"google-auth", # bigquery
|
|
"db-dtypes", # bigquery
|
|
"databricks-sql-connector", # databricks
|
|
# SSO / Auth deps
|
|
"PyJWT[crypto]>=2.15.0", # OIDC JWT verification (includes cryptography)
|
|
"cryptography>=50.0.0", # Security floor for auth and local vault encryption
|
|
"requests", # GitHub OAuth code exchange, Superset API calls
|
|
"flask-session>=0.8.0", # Server-side session (SQLite) for TokenStore
|
|
"filelock>=3.20",
|
|
"apscheduler>=3.11,<4",
|
|
"pypdf>=6.16.2",
|
|
]
|
|
|
|
[project.optional-dependencies]
|
|
# Headless-browser rendering for the data-loading agent's fetch_url(render=true).
|
|
# Needed for JavaScript single-page apps and pages behind a browser-verification
|
|
# challenge that a plain HTTP fetch cannot pass. After installing, also run:
|
|
# python -m playwright install chromium
|
|
browser = ["playwright>=1.40"]
|
|
desktop = [
|
|
"pyinstaller>=6.0",
|
|
"pywebview>=5.0",
|
|
]
|
|
|
|
[project.urls]
|
|
Homepage = "https://github.com/microsoft/data-formulator"
|
|
Repository = "https://github.com/microsoft/data-formulator.git"
|
|
"Bug Tracker" = "https://github.com/microsoft/data-formulator/issues"
|
|
|
|
[tool.setuptools]
|
|
package-dir = {"" = "py-src"}
|
|
include-package-data = true
|
|
|
|
[tool.setuptools.packages.find]
|
|
where = ["py-src"]
|
|
include = ["data_formulator*"]
|
|
|
|
# Non-Python resources that must ship inside the installed package. The Azure
|
|
# build runs `pip install .` from a zip (no VCS), so include-package-data alone
|
|
# does not pick these up — declare them explicitly. Analyst skills load their
|
|
# `SKILL.md` body and `tools.json` schemas at runtime from the package dir.
|
|
# Scoped to the exact skill filenames so unrelated data JSON is not bundled.
|
|
[tool.setuptools.package-data]
|
|
"*" = ["SKILL.md", "tools.json"]
|
|
"data_formulator.data_loader.guides" = ["*.md"]
|
|
"data_formulator.analyst.skills.map" = ["countries.json"]
|
|
"data_formulator.workflows" = ["*.yaml", "*.md"]
|
|
"data_formulator.analyst" = ["chart_catalog.json"]
|
|
# Demos shipped with the app (see data_formulator/demos/README.md): profile, front-page
|
|
# images, story seeds and the build script that installs the rest from the page.
|
|
"data_formulator" = [
|
|
"demos/README.md",
|
|
"demos/*/README.md",
|
|
"demos/*/profile.yaml",
|
|
"demos/*/skill/SKILL.md",
|
|
"demos/*/*.py",
|
|
"demos/*/*.csv",
|
|
"demos/*/assets/**/*.png",
|
|
"demos/*/stories/*/state.json",
|
|
"demos/*/stories/*/*.parquet",
|
|
]
|
|
|
|
[project.scripts]
|
|
data_formulator = "data_formulator:run_app"
|
|
data_formulator_desktop = "data_formulator.desktop:run_desktop"
|
|
|
|
[tool.uv]
|
|
dev-dependencies = [
|
|
"build",
|
|
"pytest",
|
|
]
|
|
|
|
[[tool.uv.index]]
|
|
name = "microsoft-package-feed"
|
|
url = "https://packagefeedproxy.microsoft.io/pypi/simple"
|
|
default = true
|