chore: release v0.5.0

[cmeans-claude-dev]

Summary

Release PR for v0.5.0 — the first release since v0.4.2 (2026-03-19).

Bumps `pyproject.toml` version `0.4.2 → 0.5.0` and finalizes the `CHANGELOG` by converting the `[Unreleased]` section to `## [0.5.0] - 2026-04-22` with a fresh empty `[Unreleased]` above for future work.

Why minor (not patch)

The `[Unreleased]` section contains the MIT → Apache-2.0 license switch (PR #22). A license change affects downstream compliance and warrants at minimum a minor bump. Published versions at or below 0.4.2 on PyPI remain MIT permanently; v0.5.0 onward is Apache-2.0.

Other user-visible changes that justify a minor:

• Scheduler catch-up spawn storm fix (closes Scheduler catch-up spawn storm: heartbeat fires N consecutive runs when waking from sleep #17)
• QA workflow + PR label state machine (infra)
• README badges, Codecov, PyPI metadata
• `workflow_run` registration fix for label automation (5-PR saga)
• Fresh-clone `pip install -e .` bug (caught during pre-release validation this cycle)

15 merged PRs since v0.4.2 in total — see `CHANGELOG.md` `[0.5.0]` for the full grouped entry.

Pre-release gate (all four required, all four satisfied)

1. ✅ `scripts/smoke-test.sh` — 19/19 passed, version correctly reports `0.5.0`.
2. ✅ `pytest tests/` — 449 passed on `chore/release-v0.5.0`.
3. ✅ `ruff check src/ tests/` — clean.
4. ✅ Live `--dry-run` session signed off by maintainer on 2026-04-22:
   • Fresh-clone `pip install -e .` succeeds (verifies PR fix: remove README-pypi.md generation dance, point pyproject at README.md #37 fix).
   • Browser launch, UA masking, subscription scrape (138 subs), blocklist fetch (both deslop and aislist), list-growth detection, state diff (already-blocked), feed JSON cache (22 entries on initial load), scroll + termination, clean exit all exercised successfully.
   • 0 feed matches across both sources — reflects a well-trained algorithm on a subscription-heavy account, not a regression. Absence of `POSSIBLE SELECTOR FAILURE` warnings confirms selector health.
   • Menu-click code path not exercised (data dependency), but is unchanged in `[Unreleased]` and is well-exercised in production (38 prior successful blocks in account state).

Clickbait FP assessment is N/A — no clickbait-related changes in `[Unreleased]`.

Release mechanics after merge

After this PR is QA-approved and squash-merged:

1. Maintainer only: `git tag v0.5.0` on `main` and `git push origin v0.5.0`.
2. `publish.yml` triggers automatically on the tag push:
   • Builds the wheel + sdist from the tag commit.
   • Publishes to PyPI via trusted publishing (no API token needed).
   • Creates a GitHub Release with auto-generated notes diffing `v0.4.2..v0.5.0`.
3. `sync-labels.yml` runs on the merge to `main` and keeps the label list in sync with `.github/labels.yml`.

I will not tag this release — that's explicitly maintainer-only per project memory.

Test plan

• `scripts/smoke-test.sh` — 19/19.
• `pytest tests/ -q` — 449 passed.
• `ruff check src/ tests/` — clean.
• `yt-dont-recommend --version` → `0.5.0`.
• Live `--dry-run` on both deslop and aislist — clean, no warnings.
• QA: spot-check `CHANGELOG.md` `[0.5.0]` entry covers all 15 merged PRs accurately.
• QA: confirm no regressions surfaced in CI on this branch.

[cmeans]

LGTM

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[cmeans]

QA Review — PR #38 (release v0.5.0)

Verdict: Ready for QA Signoff

All four pre-release gate items reproduce on my machine; CHANGELOG [0.5.0] entry maps cleanly to git history; [Unreleased] is correctly empty above it. One non-blocking note on the body's PR count (won't ship).

Pre-release gate — re-verified locally

• uv build on a clean dist/ → Successfully built dist/yt_dont_recommend-0.5.0.tar.gz and …-0.5.0-py3-none-any.whl. Wheel/sdist both name v0.5.0.
• pip install -e . --quiet → succeeds (verifies PR #37's fix is intact on the bumped branch).
• yt-dont-recommend --version → yt-dont-recommend 0.5.0. Confirms the pyproject.toml bump propagates through _resolve_version() / _get_current_version().
• pytest tests/ -q → 449 passed in 33.94s.
• ruff check src/ tests/ → all checks passed.
• bash scripts/smoke-test.sh → 19/19 PASS, final line: All 19 checks passed — ready to release v0.5.0.
• CI all required checks SUCCESS on chore/release-v0.5.0; mergeable. qa-gate is the only PENDING context — that's the maintainer-applied QA Approved gate by design.

CHANGELOG [0.5.0] ↔ git history

Audited the [0.5.0] section against gh pr list --state merged --base main --search "merged:>2026-03-19":

Section / entry	PR(s) covered
Added — QA workflow + label state machine	#19 ✓
Added — Stricter _get_current_version fallback	#35 (closes #32) ✓
Added — Coverage push + # pragma: no cover sweep	#30 ✓
Added — Codecov	#21 ✓
Added — README badges + FUNDING.yml	#20 ✓
Changed — License MIT → Apache-2.0	#22 (license portion) ✓
Fixed — Fresh-clone pip install -e .	#37 ✓
Fixed — Stale python -m yt_dont_recommend examples + CLAUDE.md cleanup + License docstring	#34 (closes #31) ✓
Fixed — Scheduler catch-up spawn storm	#18 (closes #17) ✓
Fixed — workflow_run registration	#24, #25, #26, #27, #28 (grouped, root cause #28) ✓
Fixed — PyPI badges / Trove classifiers	#22 (PyPI metadata portion) ✓

That accounts for 14 substantive PRs (#18, #19, #20, #21, #22, #24, #25, #26, #27, #28, #30, #34, #35, #37). The other three merged PRs since v0.4.2 are CHANGELOG-meta (#29 backfill of #18/#20/#21 entries; #33 fix-misattributions across pre-v0.4.2 entries; #36 fill-Unreleased gaps for #19 and #24-#28) — by design they don't get self-referential entries because their substance IS the entries they added/fixed elsewhere. So the [0.5.0] section is complete.

[Unreleased] shape

## [Unreleased]

## [0.5.0] - 2026-04-22

### Added
…

awk '/^## \[Unreleased\]/{flag=1; next} /^## \[/{flag=0} flag' CHANGELOG.md returns nothing — empty [Unreleased] above [0.5.0]. Correct Keep a Changelog shape for the post-release state.

Note — non-blocking

The PR body says "15 merged PRs since v0.4.2 in total". By literal gh pr list count, 17 PRs have merged into main since v0.4.2 was tagged on 2026-03-19 (#18 #19 #20 #21 #22 #24-#28 #29 #30 #33 #34 #35 #36 #37). The "15" figure looks like it's 17 − {#33, #36} (the two CHANGELOG-meta PRs whose substance lands as entries Dev wrote on behalf of other PRs) — a defensible interpretation, but the "in total" wording reads as a literal headcount. Doesn't affect anything that ships (CHANGELOG entries are accurate; the count claim is body-only and won't make it into the release notes that publish.yml auto-generates from git log v0.4.2..v0.5.0). Worth tightening the wording on the next release PR (e.g. "15 substantive feature/fix PRs (excluding CHANGELOG-meta PRs #33 and #36); 17 merged PRs in total") but not blocking on this round.

After merge — QA reminder for the maintainer

Per CLAUDE.md and project memory, only the maintainer runs git tag v0.5.0 && git push origin v0.5.0. I will not tag.

QA checkboxes ticked. Promoting to Ready for QA Signoff.