Publish AgentOps Toolkit to PyPI #21

Objective

Publish agentops-toolkit to PyPI with a clean, reproducible release process so developers can install it via pip / pipx and use it immediately.

Scope

This card covers PyPI packaging readiness, versioning, build validation, and the initial public release, including release documentation and a minimal verification checklist.

Tasks

• Confirm packaging metadata in pyproject.toml:

  • name: agentops-toolkit
  • version strategy (manual SemVer or tool-based)
  • Python requirement: 3.11+
  • license, description, classifiers, keywords

• Ensure included files are correct:

  • README rendered properly on PyPI
  • include src/ package content
  • include required non-code assets (if any)

• Add build/release tooling:

  • python -m build
  • twine check dist/*

• Define release steps (documented):

  • build
  • test install from wheel
  • publish to TestPyPI (optional but recommended)
  • publish to PyPI

• Add a release verification checklist:

  • pip install agentops-toolkit
  • agentops --version
  • agentops eval run starts (or runs against minimal sample config if available)

• Add a “Quick install” section in docs aligned with PyPI installation

• (Optional) Add GitHub Actions release workflow for publishing (manual trigger is fine)

Acceptance Criteria

• agentops-toolkit is published on PyPI under the expected name
• pip install agentops-toolkit installs successfully on Python 3.11
• The installed package exposes the agentops CLI entry point
• agentops --help and agentops --version work after install
• Release documentation exists and can be followed to cut the next release

Out of Scope

• Adding new CLI commands beyond what Sprint 1 already includes
• Advanced release automation (semantic-release, multi-channel, etc.) unless already planned
• Marketplace/Copilot skill distribution (that’s Sprint 2)