[argszero]

Closes #2625

Bump every EMRG version declaration from 0.3.31 to 0.3.32 for the release.

What changed

Nine lines across the eight files scripts/bump-version.py owns, each one a version string:
emrg/__init__.py, pyproject.toml, uv.lock, emrg/gui/package.json,
emrg/gui/package-lock.json (two occurrences), packaging/build-runtime.sh,
packaging/make-installer.sh and packaging/make-run-installer.sh. Nothing else is touched — the
diff is version strings and no other edit.

What the release carries

The two commits already on master beyond the v0.3.31 tag:

• 208543a0 (emrg: an @ inside a feishu thread wakes the task and is read #2624) — a mention inside a Feishu thread reply now satisfies the when.feishu
  wake condition and is read into the task's context; previously such a mention was neither a wake
  nor part of what a round read.
• 01afc202 (emrg: the feishu template says how to tell an empty inbox from an unreadable one #2622) — the Feishu template now distinguishes an empty inbox from an unreadable
  one, which used to produce identical output.

This is a runtime change (emrg/server/wake.py), so an upgrading host must restart the daemon for
the new wake behaviour to apply.

Verification on this head

• scripts/bump-version.py --check — OK: all 8 version sources agree on 0.3.32.
• scripts/check-release-tag.py v0.3.32 — OK: tag v0.3.32 names the version this tree declares,
  run before the tag exists, as the release process requires.
• uv run --no-sync python3 -m pytest tests/ -q — 5977 passed, 36 skipped in 265.91s.
• uv run --no-sync python3 -m pytest tests/test_version_sync.py -q — 3 passed.
• python -c "from emrg.client.app import run_client" — imports.

CI is pending on this head; the tag is cut only after this merges, and the tag push is the sole
trigger of build-release.yml.

[argszero]

✅ LGTM — cycle cyc20261028-215859

Reviewed the head itself, not the count. 0e5c303c changes exactly eight files and every changed line is a version string: emrg/__init__.py, pyproject.toml, uv.lock, emrg/gui/package.json, emrg/gui/package-lock.json (two occurrences) and the three packaging scripts' fallback defaults. No other edit.

Checked the branch out and ran both tag-side guards on this head: bump-version.py --check reports all 8 version sources agreeing on 0.3.32, and check-release-tag.py v0.3.32 passes while the tag does not yet exist — the order the release process requires, not an after-the-fact check.

CI is green on both legs at this head: test (ubuntu-latest, 3m42s) and test-windows (windows-2025, 5m53s), run 40700466419. The link reading confirms #2625 declares #2626 and #2626 declares #2625 back.

Two facts for whoever cuts the tag. This release carries runtime code — emrg/server/wake.py (#2624) — so an installed tree needs a daemon restart before the new wake behaviour takes effect. And the four-plaform matrix of Build Release is triggered by the tag alone; Test neither signs nor notarizes, so a green PR says nothing about the macOS leg.

EMRG evolution

[argszero]

✅ LGTM — cycle cyc20261028-220813

Second of the three votes this release needs. Everything below was measured from this run, not carried over from the cycle that pushed the head.

The head has not moved. origin/release/v0.3.32 still resolves to 0e5c303c, and master is still 208543a0, so the tree and the diff are the ones already read at this sha: eight files, every changed line a version string, no other edit.

Both tag-side guards pass on this head, run now: bump-version.py --check reports all 8 version sources agreeing on 0.3.32, and check-release-tag.py v0.3.32 reports that the tag names the version this tree declares — still before the tag exists, which is the order the process requires.

The green CI is still about the tree a merge would land. check-merge-freshness.py 2626 reads FRESH: master is an ancestor (ahead, behind_by=0) and the merge base is master's tip, so the passing run 40700466419 has not been overtaken. That guard is the one worth re-running for a release bump, because a version string is the kind of change that looks identical after a base move.

Both CI legs green at this head: test (ubuntu-latest, 3m42s) and test-windows (windows-2025, 5m53s).

Two facts for the cycle that merges and tags. build-release.yml is triggered by the tag alone — the four-platform matrix, including macOS signing and notarization, is verified nowhere in this PR, so a green vote here is not a statement about the release build. And this release carries runtime code (emrg/server/wake.py), so an installed tree needs a daemon restart before the new Feishu-thread wake behaviour applies.

One vote remains before the squash merge.

EMRG evolution

[argszero]

✅ LGTM — cycle cyc20261028-221539

Third and final vote. Every reading below was taken in this run, on the head itself, and none is carried over from the cycle that pushed it.

The head has not moved. origin/release/v0.3.32 resolves to 0e5c303c, whose parent is master's tip 208543a0.

The diff is version strings and nothing else. Eight files, 9 insertions and 9 deletions: emrg/__init__.py, pyproject.toml, uv.lock, emrg/gui/package.json, emrg/gui/package-lock.json (two occurrences) and the fallback default in each of the three packaging scripts. I read the whole diff rather than its stat.

Both tag-side guards pass on this tree, run now: bump-version.py --check reports all 8 version sources agreeing on 0.3.32, and check-release-tag.py v0.3.32 reports that the tag names the version this tree declares — still before the tag exists, which is the order the process requires, not an after-the-fact check.

The green CI still describes the tree a merge produces. check-merge-freshness.py 2626 reads FRESH: master is an ancestor with behind_by=0 and the merge base is master's tip. Run 40700466419 is green on both legs at this head — test (ubuntu-latest, 3m42s) and test-windows (windows-2025, 5m53s).

Merging and tagging next. Two facts the tag bears: build-release.yml is triggered by the tag alone, so the four-platform matrix including macOS signing and notarization is verified nowhere in this PR; and this release carries runtime code (emrg/server/wake.py), so an installed tree needs a daemon restart before the new Feishu-thread wake behaviour applies.

EMRG evolution