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

[project]
name = "openenv-stocker"
version = "0.1.0"
description = "Stocker — multi-agent council RL environment for stock trading on OpenEnv"
requires-python = ">=3.10"
dependencies = [
    "fastapi>=0.115.0",
    "uvicorn[standard]>=0.30.0",
    "pydantic>=2.0.0",
    "pydantic-settings>=2.6.0",
    "openai>=1.0.0",
    "openenv-core>=0.2.0",
    "pandas>=2.0.0",
    "pyarrow>=14.0.0",
    "pillow>=10.0.0",
    "markdown-it-py>=3.0.0",
]

[project.scripts]
server = "server.app:main"

[project.optional-dependencies]
dev = [
    "pytest>=8.0.0",
]
data = [
    "yfinance>=0.2.0",
    "mplfinance>=0.12.0",
    "matplotlib>=3.8.0",
]
eval = [
    "matplotlib>=3.8.0",
]
serve = [
    "vllm>=0.6.0",
]
#
# NOTE: the [train] extras are intentionally NOT defined here.
#
# The training Space uses a deliberately decoupled dep stack that lives
# in spaces/train/Dockerfile (heavy ML deps with exact pins). Keeping
# train deps out of pyproject.toml means:
#   * Env Space install never pulls torch/transformers/peft/torchao,
#     so demo deploys can never break from training-side conflicts.
#   * Training Space pins live next to the Dockerfile they're used in.
#   * pip install -e ".[dev,data,eval]" still works for env-Space
#     development without ever touching ML dependencies.
#
# To install training deps locally, see spaces/train/Dockerfile or run:
#   pip install torch==2.10.0 torchvision==0.25.0 \
#       transformers==5.6.2 peft==0.19.1 trl==1.2.0 ...
# (matching the exact pins in spaces/train/Dockerfile).

[tool.setuptools]
# Explicit package list — setuptools' flat-layout autodetect would otherwise
# pick up data/ and training/ (which are not importable Python packages).
packages = [
    "app",
    "app.api",
    "app.core",
    "app.council",
    "app.data",
    "server",
]

[tool.uv]
package = true

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