PyPI release: make climate-api installable via pip install climate-api

[turban]

Goal

pip install climate-api installs a working Climate API instance. The package name is available on PyPI.

Current state

• pyproject.toml already has name = "climate-api", version = "0.1.0", a description, and a climate-api console script entry point.
• The package is not published anywhere yet.

Blockers to resolve

1. Git-sourced dependencies must be on PyPI

Two runtime dependencies are pinned to GitHub branches and cannot ship in a PyPI release:

"dhis2eo @ git+https://github.com/dhis2/dhis2eo.git@v1.2.0",
"dhis2-client @ git+https://github.com/dhis2/dhis2-python-client.git@V0.3.0",

Both packages must be published to PyPI with matching version tags before climate-api can be released. Coordinate with the owners of those repos, or vendor the relevant modules if a release is not feasible.

2. Bundle the built-in dataset YAMLs as package data

The built-in templates in data/datasets/ are currently located outside the installable src/climate_api/ tree, referenced by a __file__-relative path in data_registry/services/datasets.py. An installed wheel will not include data/.

Steps:

• Move data/datasets/ to src/climate_api/datasets/ (or another path inside the package).
• Add it to the wheel via pyproject.toml:

  [tool.uv_build]
  include = ["src/climate_api/datasets/**"]

• Update _BUILTIN_DATASETS_DIR in datasets.py to use importlib.resources instead of __file__:

  from importlib.resources import files
  _BUILTIN_DATASETS_DIR = files("climate_api") / "datasets"

3. Bundle the pygeoapi base config

config/pygeoapi/ is also outside the package. Either move it inside src/climate_api/config/pygeoapi/ and include it as package data, or make the path configurable via an environment variable with a sensible fallback using importlib.resources.

4. Complete package metadata

Add the missing fields to pyproject.toml so the PyPI page is useful:

[project]
license = { text = "BSD-2-Clause" }   # confirm correct license
authors = [{ name = "DHIS2", email = "..." }]
readme = "README.md"
keywords = ["dhis2", "climate", "zarr", "stac", "geospatial"]
classifiers = [
    "Programming Language :: Python :: 3",
    "License :: OSI Approved :: BSD Software License",
    "Intended Audience :: Science/Research",
    "Topic :: Scientific/Engineering :: GIS",
]

[project.urls]
Repository = "https://github.com/dhis2/climate-api"
Documentation = "https://github.com/dhis2/climate-api/tree/main/docs"

5. Publish workflow

Add a GitHub Actions workflow that builds and publishes to PyPI on tagged releases (using Trusted Publishing — no API token needed):

# .github/workflows/publish.yml
on:
  push:
    tags: ["v*"]

jobs:
  publish:
    runs-on: ubuntu-latest
    environment: pypi
    permissions:
      id-token: write
    steps:
      - uses: actions/checkout@v4
      - uses: astral-sh/setup-uv@v5
      - run: uv build
      - uses: pypa/gh-action-pypi-publish@release/v1

Configure the pypi environment in the repo settings and add the project on PyPI under Trusted Publishing.

6. Verify install in a clean environment

Before the first release:

uv venv /tmp/test-install && source /tmp/test-install/bin/activate
pip install .
climate-api   # should start the server

Out of scope for this issue

• Runtime data directories (data/downloads/, data/pygeoapi/) — these are instance-specific and already handled by environment variables (CACHE_OVERRIDE, PYGEOAPI_CONFIG). No change needed.
• The climate-api.yaml instance config — installed users are expected to create their own.