release: v0.3.0 — per-OS download breakdown badges

[cmeans-claude-dev]

Summary

Bumps version 0.2.0 → 0.3.0 and promotes the [Unreleased] section to [0.3.0] - 2026-04-29. One feature PR plus three small chore PRs land in this release.

Why minor (not patch)

run_pypinfo()'s return type changed from dict[str, int] to a TypedDict carrying both by_installer and by_system aggregates in #57 — same kind of internal-contract evolution that drove the 0.1.x → 0.2.0 jump. The feature surface is also new (3 new shields.io endpoint badge files per package per window: linux / macos / windows), making 0.3.0 the right semantic line.

What's in the diff

• pyproject.toml — version 0.2.0 → 0.3.0.
• CHANGELOG.md — ## [Unreleased] heading replaced with ## [0.3.0] - 2026-04-29; link refs updated ([Unreleased] now points at v0.3.0...HEAD; new [0.3.0] ref). Section content untouched — already accumulated across the four PRs that landed since v0.2.0.

What's shipping (already on main)

Added

• feat(collector): per-OS download breakdown badges (feat: per-OS download breakdown badges (v3) #57, d7688ce) — three additional badge files per package per window (os-linux-Nd-non-ci.json, os-macos-Nd-non-ci.json, os-windows-Nd-non-ci.json); pypinfo group-by extends from ci installer to ci installer system; _health.json gains a counts_by_system field. Hero count and the seven v0.2.0 installer-* files unchanged in filename, schema, and value. Already verified end-to-end on the live CT 112 deploy with real production data (11 files per package across all 4 dogfood packages).
• ci: add weekly uv lock --upgrade refresh cron (ci: add weekly uv lock --upgrade refresh cron #54, 3ed03b9) — scheduled workflow runs uv lock --upgrade every Thursday 12:00 UTC as a backstop for transitive freshness. Skip-gate + test-gate keep it from creating overlapping or broken PRs.

Changed

• chore(deps): refresh uv.lock transitive pins (chore(deps): refresh uv.lock transitive pins (2026-04-29) #55, 92ec030) — routine uv lock --upgrade resolve. No pyproject.toml range changes.
• chore(.gitignore): ignore .deploy/ for private operator tooling (chore(.gitignore): ignore .deploy/ for private operator tooling #52, 5bae606) — maintainer-shaped deploy scripts stay out of public history.
• chore(.gitignore): ignore .claude/settings.local.json (chore(.gitignore): ignore .claude/settings.local.json (per-machine Claude Code overrides) #53, 2bd764a) — per-machine Claude Code permission overrides.

Backwards-compat verified

• Hero downloads-Nd-non-ci.json filename / schema / value unchanged for any given pypinfo response.
• The seven v0.2.0 installer-* files unchanged.
• _health.json count, counts, window_days keys preserved verbatim.
• Hero count = sum(by_installer.values()) regardless of system_name.

Test plan

• uv build (sdist + wheel) — local validation.
• CI green (lint, typecheck, test 3.11/3.12/3.13, deploy-smoke).
• On merge: publish.yml extracts [0.3.0] notes from CHANGELOG, builds + uploads to PyPI via trusted publisher, creates GitHub release page.
• After release: collector run on CT 112 emits 11 files per package per window (already validated pre-merge from feature branch — no regression expected from version bump alone).
• After release: PyPI pypi-winnow-downloads 0.3.0 page renders correctly.

Release framing

Target tag: v0.3.0 — minor bump per SemVer. Additive feature; no breaking changes to v0.2.0 contracts.

🤖 Generated with Claude Code

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[cmeans]

LGTM

[cmeans]

QA round 1 — PASS (no findings)

Clean release PR — minimal-pair version bump (0.2.0 → 0.3.0 in pyproject.toml) + CHANGELOG section stamp (## [Unreleased] → ## [0.3.0] - 2026-04-29) + new link reference. No content changes — the section was already accumulated across PRs #52, #53, #54, #55, #57.

Verifications

Section accounting — all 5 shipped PRs accounted for in the v0.3.0 section:

• ### Added:
  • .github/workflows/uv-lock-refresh.yml ← PR #54
  • Per-OS badge files (v3 OS distribution feature) ← PR #57
• ### Changed:
  • uv.lock transitive pins ← PR #55
  • .gitignore .deploy/ ← PR #52
  • .gitignore .claude/settings.local.json ← PR #53

KaC ordering preserved (Added → Changed); no orphaned entries.

CHANGELOG link refs — [Unreleased] updated to v0.3.0...HEAD; new [0.3.0] reference added pointing at v0.2.0...v0.3.0; older refs preserved verbatim. After this PR merges, the document has no ## [Unreleased] heading at the top — the first post-release PR introduces it (consistent with the v0.2.0 release pattern and project convention).

Auto-release-page extractor pre-test — pre-running the live awk script from publish.yml against this branch's CHANGELOG with VERSION=0.3.0 produces a 12-line non-empty body (2 ### Added bullets + 3 ### Changed bullets + their headings, leading/trailing blank lines trimmed). Build-job gate passes; release job will get clean release notes when the v0.3.0 tag is pushed.

Build artifacts — rm -rf dist/ && uv build produces both:

• pypi_winnow_downloads-0.3.0.tar.gz (184,899 bytes)
• pypi_winnow_downloads-0.3.0-py3-none-any.whl (19,509 bytes)

Version resolution — uv run python -c "from pypi_winnow_downloads import __version__; print(__version__)" → 0.3.0 ✓

Test suite + style on release-branch HEAD 88ba81b:

• uv sync --frozen --extra dev → succeeded
• uv run pytest --cov → 88 passed at 100% coverage (Required test coverage of 100.0% reached)
• uv run ruff check src/ tests/ → All checks passed
• uv run ruff format --check src/ tests/ → 11 files already formatted
• uv run mypy src/pypi_winnow_downloads/ → Success: no issues found in 5 source files
• CI: 8 pass / 3 expected-skip (changelog, on-label, on-unlabel) / 1 pending QA Gate ✓

Backwards-compat invariants — already verified during PR #57 QA (test_collect_one_v0_2_0_files_unchanged_alongside_os_files, test_health_json_preserves_v0_2_0_fields, etc.); no code change in this release PR, so they carry forward.

Test plan checkboxes 3-5

• On merge: publish.yml extracts [0.3.0] notes from CHANGELOG, builds + uploads to PyPI via trusted publisher, creates GitHub release page — pre-tested via the awk extractor above; will fire on tag push.
• Collector run on CT 112 emits 11 files per package per window — already validated pre-merge from #57's feature branch per PR body. Version bump alone introduces no regression risk.
• PyPI page renders correctly — post-publish verification by maintainer.

Labels: Ready for QA → QA Active → Ready for QA Signoff. Awaiting maintainer QA Approved.

[cmeans]

Audit trail: applying Ready for QA Signoff — minimal-pair version bump 0.2.0→0.3.0, all 5 shipping PRs (#52/#53/#54/#55/#57) accounted for in the [0.3.0] section with KaC ordering preserved, link refs correctly updated, awk extractor produces 12-line non-empty body for v0.3.0, uv build produces 0.3.0 sdist + wheel, version resolves to 0.3.0, 88/88 pytest at 100% coverage, ruff/format/mypy clean, CI fully green. Workflow: Ready for QA → QA Active → Ready for QA Signoff.