Publish bead to conda-forge for conda/mamba installation

[korenmiklos]

Goal

Make bead installable via conda install -c conda-forge bead or mamba install bead by publishing to conda-forge.

Dependencies

• Depends on: Publish bead to PyPI for pipx installation #1 (PyPI publication) - conda-forge strongly prefers packages to be on PyPI first

Current Status

• Package uses modern pyproject.toml configuration
• Has clear dependencies listed
• Python 3.10+ requirement is conda-friendly

Required Steps

1. Prerequisites

• Complete PyPI publication (Publish bead to PyPI for pipx installation #1) - REQUIRED FIRST
• Wait for PyPI package to be indexed (usually ~15 minutes)

2. Generate Initial Recipe with Grayskull

# Install grayskull
conda create -n grayskull
conda activate grayskull
conda install -c conda-forge grayskull

# Generate recipe from PyPI
grayskull pypi --strict-conda-forge bead

3. Review and Adjust Recipe

The generated meta.yaml will need review for:

• Verify all dependencies are available on conda-forge
• Check if argcomplete needs special activation scripts
• Ensure entry point bead is correctly specified
• Add appropriate test commands including pip check
• Set proper license identifier (The Unlicense → Unlicense)

4. Fork and Submit to staged-recipes

• Fork https://github.com/conda-forge/staged-recipes
• Create branch: add-bead-recipe
• Add recipe to recipes/bead/meta.yaml
• Follow conda-forge linting standards

5. Create Pull Request

• Submit PR to conda-forge/staged-recipes
• Address reviewer feedback
• Wait for CI checks (Azure Pipelines) to pass
• Get approval from conda-forge reviewers

6. Post-Acceptance

Once the PR is merged:

• Bot creates dedicated feedstock repo: bead-feedstock
• Package builds automatically for multiple platforms
• Available on conda-forge channel within hours

7. Maintenance

• Become feedstock maintainer
• Set up automated updates via regro-cf-autotick-bot
• Monitor for dependency updates

Recipe Template (Starting Point)

{% set name = "bead" %}
{% set version = "0.8.1" %}

package:
  name: {{ name|lower }}
  version: {{ version }}

source:
  url: https://pypi.io/packages/source/{{ name[0] }}/{{ name }}/{{ name }}-{{ version }}.tar.gz
  sha256: <TO_BE_FILLED_FROM_PYPI>

build:
  noarch: python
  number: 0
  entry_points:
    - bead = bead_cli.main:main
  script: {{ PYTHON }} -m pip install . -vv

requirements:
  host:
    - python >=3.10
    - pip
    - flit-core
  run:
    - python >=3.10
    - appdirs
    - argcomplete >=3.6.2
    - attrs
    - cached-property
    - pip >=25.1.1

test:
  imports:
    - bead
    - bead_cli
  commands:
    - pip check
    - bead --help

about:
  home: https://github.com/codedthinking/bead
  summary: Linked frozen computations
  license: Unlicense
  license_file: LICENSE
  description: |
    BEAD is a format for freezing and storing computations while bead 
    is a tool that helps capturing and managing computations in BEAD formats.

extra:
  recipe-maintainers:
    - <YOUR_GITHUB_USERNAME>

Benefits of conda-forge

• Cross-platform binary packages (Windows, macOS, Linux)
• Automatic dependency resolution
• Integration with scientific Python ecosystem
• Works with both conda and mamba
• Automated rebuilds when dependencies update
• Community maintenance support

Timeline Estimate

1. After PyPI publication: 1-2 days to prepare recipe
2. PR review: 3-7 days typically
3. Post-merge availability: 2-6 hours

References

• conda-forge documentation
• Grayskull recipe generator
• staged-recipes repository