pyproject.toml

Rama main · réplica de lectura del repositorio original.

[build-system]
requires = ["setuptools>=68"]
build-backend = "setuptools.build_meta"

[project]
name = "datasteward"
version = "0.1.0"
description = "Self-hosted Foundry-style governed data platform, AI-agent native: ontology + PBAC/markings + full audit + human-approved revertible writebacks, with a LangGraph agent that only reaches data through the governance kernel"
readme = "README.md"
license = { text = "Apache-2.0" }
requires-python = ">=3.10"
keywords = ["data-governance", "data-platform", "ai-agent", "langgraph", "ontology", "starrocks", "dbt", "audit"]
classifiers = [
    "Development Status :: 4 - Beta",
    "Intended Audience :: Developers",
    "License :: OSI Approved :: Apache Software License",
    "Programming Language :: Python :: 3",
    "Programming Language :: Python :: 3.10",
    "Programming Language :: Python :: 3.11",
    "Programming Language :: Python :: 3.12",
    "Topic :: Database",
    "Topic :: Scientific/Engineering :: Artificial Intelligence",
]
dependencies = [
    "pymysql>=1.1",
    "pandas>=2.0",
    "faker>=25",
    "streamlit>=1.36",
    "plotly>=5.20",
    "pyyaml>=6",
    "mcp>=1.2,<2",               # FastMCP v1 API; mcp 2.x renamed FastMCP and breaks dm.connector.mcp_server
    "requests>=2.31",
    "psycopg[binary]>=3.1",       # v3：管理平台 app/data.py 用
    "psycopg2-binary>=2.9",       # v2：docs/kg/pipeline/sources 及 pgvector.psycopg2 适配用（合并分支后两版并存）
]

[project.urls]
Homepage = "https://github.com/Leonxu6/datasteward"
Repository = "https://github.com/Leonxu6/datasteward"
Issues = "https://github.com/Leonxu6/datasteward/issues"

[project.optional-dependencies]
dev = ["pytest>=8"]
# RAG（S2）：本地嵌入 fastembed（onnxruntime，CPU 友好，比 torch 轻）+ pgvector 适配。
rag = ["fastembed>=0.3", "pgvector>=0.2", "numpy>=1.24"]
# KG（S3）：Neo4j 官方驱动（纯 Python）。
kg = ["neo4j>=5"]
# 连接器：SQL Server/用友 U8（pymssql 或 pyodbc）+ 文件源 Excel（openpyxl）。
connectors = ["pymssql>=2.2", "openpyxl>=3.1"]
# 智能体（L10）：自研 LangGraph + OpenAI 兼容端点（云 API 或本地 ollama/LiteLLM）；PG 检查点。
agent = ["langgraph>=1,<2", "langchain-openai>=1,<2", "langgraph-checkpoint-postgres>=3,<4"]
# 建模转换（L5）：dbt + StarRocks 适配器（dbt-core 版本由适配器解析，勿双 pin）。
dbt = ["dbt-starrocks>=1.12"]
# 编排（L7）：Dagster + dbt 资产集成 + PG 存储。
# 注意 protobuf：dagster 的 grpc-health gencode 7.35 要求 runtime ≥7.35，而 dbt-core 元数据
# pin <7 —— 二者在同一次 resolve 里无解（ResolutionImpossible）。解法＝装完后覆盖升级
# `pip install "protobuf>=7.35"`（deploy/Dockerfile 已内置；实测 dbt 兼容 7.x，见 DEVLOG 坑20）。
orchestration = ["dagster>=1.13", "dagster-webserver>=1.13", "dagster-dbt>=0.29", "dagster-postgres>=0.29"]
# 钉钉通道（可选）：Stream 双向机器人。不用钉钉可不装，dm-dingtalk 会给出清晰报错。
dingtalk = ["dingtalk-stream>=0.22"]

[project.scripts]
dm-load = "dm.warehouse.load:main"
dm-agent = "dm.agent.core:main"
dm-app = "dm.app.__main__:main"
dm-eval = "dm.eval.run_eval:main"
dm-dingtalk = "dm.channels.dingtalk:main"
dm-docs = "dm.docs.index:main"
dm-kg = "dm.kg.build:main"
dm-connect = "dm.connect.onboard:main"
dm-u8 = "dm.connect.u8_mapping:main"

[tool.setuptools]
package-dir = {"" = "src"}

[tool.setuptools.packages.find]
where = ["src"]

[tool.setuptools.package-data]
"dm.eval" = ["eval_set.yaml"]
"dm.ontology" = ["metrics.yaml"]

[tool.pytest.ini_options]
testpaths = ["tests"]
markers = [
    "integration: 需要真实 claude / 网络 / 较慢的端到端用例（跳过用 -m 'not integration'）",
    "stack: 需要可达的 StarRocks/Postgres 栈（docker compose up -d 后运行；不可达自动跳过）",
]