Skip to content

Release 1.5.0: NSGA-III's parent selection, one source of reference vectors, AGE-MOEA-II and RVEA's PlatEMO variant, CTP and IDTLZ - #368

Merged
NicolasRodriguezUribe merged 2 commits into
mainfrom
release/final-1.5.0
Nov 19, 2026
Merged

NicolasRodriguezUribe merged 2 commits into
mainfrom
release/final-1.5.0

Conversation

@NicolasRodriguezUribe

@NicolasRodriguezUribe NicolasRodriguezUribe commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor

Summary

Version 1.5.0 of vamos-optimization is a minor release of 1.4.0, dated 2026-11-19, on release/final-1.5.0. It collects the 11 pull requests merged since the v1.4.0 release merge (254209a7, #356), up to main 460f13e1 (#367). The release notes group them for users:

Release changes

  • Version and date. src/vamos/foundation/version.py and CITATION.cff say 1.5.0, released 2026-11-19, and so do the version consistency test, the release workflows (VAMOS_RELEASE_VERSION, the release/final-1.5.0 head ref and the v1.5.0 tag), docs/release_smoke.md (where v1.4.0 joins the immutable tags), AGENTS.md and the tests that pin the version. Every mention that names the current version says 1.5.0: install commands, "targets VAMOS", "Experimental surface in VAMOS", "Known limitations in VAMOS", "Study execution is sequential in", the capability table and the problems catalog (260 registered problems). 1.5.0 appears 320 times in 144 files, the markers and the lessons included.

  • Historical mentions stay. A mention of 1.4.0 stays when it says what 1.4.0 did or introduced: the "VAMOS 1.4.0:" markers, the comparisons that the merged pull requests wrote ("VAMOS 1.4.0 and earlier ran ...", "the resolved spec of VAMOS 1.4.0", "where VAMOS 1.4.0 created it as a directory", the NSGA-III grammars' "The spaces of VAMOS 1.4.0 only sized a tournament"), the VAMOS-CONFIG design, which ships in 1.4.0, and the 1.4.0 release notes.

  • Markers. Every "Development after 1.4.0:" note becomes "VAMOS 1.5.0:", and every comment or docstring "Development after 1.4.0:" becomes "VAMOS 1.5.0:": 136, plus five that wrap across lines (experiments/scripts/generate_reference_fronts.py, src/vamos/experiment/cli/orchestration.py, tests/reference/test_scientific_defaults.py, tests/reference/baselines.md and src/vamos/algorithms.py, whose moead_populations "is new in VAMOS 1.5.0"). The comments of the six NSGA-III grammars change too; the spaces' hashes come from the parsed document, so they do not.

  • CHANGELOG.md. The [1.5.0] section is built from the pull requests' "Changelog entry" sections, with an introduction, the pull request number added to the entries that lacked it, and three corrections:

    chore(deps): Bump mkdocs-material from 9.7.7 to 9.8.1 #358 has no entry (Dependabot). [Unreleased] is empty, and the link references compare from v1.5.0.

  • Release notes. docs/project/release-notes-1.5.0.md is new, with a "Which results change" table (NSGA-III; MOEA/D with four objectives and 100 to 119 individuals; the NSGA-III and AGE-MOEA tuning spaces; the stricter vector options). It is listed under Reference → Support & Versions, in the reference index and in the navigation inventory, and the README intro summarizes 1.5.0 and links to it.

  • Known limitations. No entry changes; only the title names 1.5.0.

  • Primary-documents test. tests/docs/test_primary_release_content.py asserted that the primary documents do not mention 1.5.0, a future version when the test was written for 1.0.0. 1.5.0 is now the current version, so the assertion goes, as the 1.1.0 release dropped its own.

  • Lessons. The fundamental, study and coverage lessons and the algorithm guides with recorded output are regenerated with a 1.5.0 wheel built from this branch at 6147ca24, in a new, non-editable Python 3.12.3 environment with the recorded dependency versions and no PYTHONPATH, and LESSON_RELEASE is now "1.5.0". All three --check runs pass, before and after the prose updates. No lesson prints other results:

    • the fundamental lessons' output is unchanged;
    • the study lessons' hypervolume, IGD+ and front hashes are unchanged in all 120 rows; only study, attempt and run identifiers, manifest hashes and runtimes differ;
    • the coverage lessons print the same results; the benchmark timings differ, as timings do, and AGE-MOEA's recorded configuration also lists "variant": "agemoea", the field that feat(agemoea): AGE-MOEA-II as AGE-MOEA's variant, checked against jMetal #359 adds.

    No lesson runs NSGA-III, and the study lessons run MOEA/D with two objectives, so the corrections of fix(nsgaiii): select parents as configured, at random by default #364 and fix: every packaged weight file loads (W4D_100.dat, W4D_165.dat) and NSGA-III reads them (A14, A8) #361 change none of them, and no prose cites a changed number. The prose that named 1.4.0 as the tested release now names 1.5.0. The generators write CRLF on Windows, so the written files are normalized to LF, as committed, before the checks. The lessons' evidence records 6147ca24, the head the wheel was built from.

Validation

tools/release_check.py --version 1.5.0 --expected-branch release/final-1.5.0 --expected-commit be482f9e --tag-state pre-release --typing-python <typing venv> --output-dir <new evidence directory> passed on Windows with Python 3.12.3, NumPy 2.4.6 and SciPy 1.17.1, main's CI pins, in new environments installed from this branch with constraints/ci.txt and the pinned release tools (release/requirements-build.txt, release/requirements-tools.txt), as docs/release_smoke.md describes: one with the development, documentation, compute, analysis, examples and Studio extras, and a dependency-minimal .[dev] typing environment. Of its 35 checks, 34 passed. The other check, zero full-source diagnostics, is informational, and it finds none (0 diagnostics).

  • Repository identity (branch release/final-1.5.0, commit be482f9e, clean), canonical repository identity, version consistency, release documents, license metadata, tag state (v1.5.0 absent; v1.0.0 to v1.4.0 annotated, equal to the remote and ancestors), path and secret scan, repository hygiene, pre-release remnants and the stable API, CLI and schema fixtures (tests/compatibility, 19 passed) pass.
  • Typing: strict, the stable public API, the full-source ratchet (new=0, increased=0) and the release typing policy pass; health passes. Ruff lint and format and compileall pass.
  • Public examples (tests/docs): 362 passed, 1 skipped. Complete tests: 11223 passed, 256 skipped, 9 xfailed, in 75 minutes, the slow tier included. The suite runs without xdist, so the known Windows failure of tests/experiment/test_cli_ablation.py::test_cli_ablation_runs_and_writes_variants under xdist does not arise, and no MemoryError occurred.
  • The documentation builds with --strict, without warnings.
  • The sdist and the wheel vamos_optimization-1.5.0-py3-none-any.whl build in isolation and pass the distribution hygiene, twine, the wheel content policy and the distribution inspection. A clean install of the wheel with the compute extra runs the full release smoke.
  • The dependency audits find nothing among 26 runtime dependencies, 3 build dependencies and 68 release-tool dependencies. The CycloneDX SBOM, checksums and provenance are written, and the worktree stays clean.

The hosted candidate workflow and CI pass on this head, the "Canonical release checker" job included, and the automated review found nothing.

Before that, on the same head: tools/health.py (typing environment), mkdocs build --strict and the quick tier with architecture and compatibility (tests/test_check_agent_docs.py tests/architecture tests/docs tests/compatibility tests/test_release_automation.py tests/test_environmental_selection.py: 986 passed, 1 skipped) passed separately, and the three lessons' --check and their tests (63 passed) passed.

After merging

  1. Dispatch release.yml on the final main commit and wait for all source and installed-wheel gates, including "Freeze validated evidence and artifacts", as docs/release_smoke.md prescribes.
  2. Confirm the TestPyPI and PyPI Trusted Publisher registrations and add v1.5.0 to the tag rules of the testpypi and pypi environments, keeping the production approval.
  3. Create the annotated tag v1.5.0 on that commit and push it (git tag -a v1.5.0 <final main commit> -m "VAMOS 1.5.0", then git push origin v1.5.0). upload_pypi.yml recovers the frozen artifacts, publishes to TestPyPI and checks the installation and hashes, then publishes to PyPI once the maintainer approves the pypi deployment, and creates the GitHub Release with the notes' absolute links.
  4. Dispatch docs-cloudflare.yml with version=1.5.0, expected_source_sha the final main commit and archive_release=v1.4.0. The maintainer approves the cloudflare-production deployment.

🤖 Generated with Claude Code

…rkers

The version moves to 1.5.0, dated 2026-11-19: the runtime version, the
citation, the release and publication workflows, the version consistency
test, the tests that pin the version, AGENTS.md, release_smoke.md (1.4.0
joins the immutable tags), the install commands and the current-version
statements of the guides (67 scripted edits), the reference index, the
navigation and the README. Historical mentions of 1.4.0 stay: the
"VAMOS 1.4.0" markers, the notes that compare with what 1.4.0 did ("VAMOS
1.4.0 and earlier ran ...", "the resolved spec of VAMOS 1.4.0"), the
VAMOS-CONFIG design, and the 1.4.0 release notes. The primary-documents
test no longer asserts that 1.5.0 is absent: 1.5.0 is now the current
version, as 1.1.0 dropped its own assertion.

Every "Development after 1.4.0" marker becomes "VAMOS 1.5.0" (136, plus
five that wrap across lines), in documentation, docstrings, comments and the
comments of the NSGA-III grammars, whose parsed spaces and hashes do not
change. The problems catalog counts 260 problems in 1.5.0.

CHANGELOG.md gains [1.5.0] from the "Changelog entry" sections of the pull
requests merged since 1.4.0 (#357 to #367; #358 is Dependabot and has
none), with the pull request number added where an entry lacked it, and
three corrections: #361's NSGA-III path rule, which #365 widened to
directories ("a path that does not exist raises"), and the different seeded
tuning trials of the AGE-MOEA spaces (#359) and the NSGA-III spaces (#364),
which their entries left out; its introduction calls 1.5.0 a minor release,
as the stability policy prescribes for new stable functionality.
docs/project/release-notes-1.5.0.md leads with
NSGA-III's parent selection and the reference and weight vectors, and says
which results change. The known limitations do not change.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…OS 1.5.0

The fundamental, study and coverage lessons, and the algorithm guides
with recorded output, are regenerated with a 1.5.0 wheel built from the
release branch at 6147ca2, in a new non-editable environment with the
recorded dependency versions (Python 3.12.3, LESSON_RELEASE = "1.5.0");
all three --check runs pass. The generators write CRLF on Windows, so the
written files are normalized to LF before the check, as committed.

No lesson prints other results. The fundamental lessons' output is
unchanged. The study lessons' hypervolume, IGD+ and front hashes are
unchanged in all 120 rows; only study, attempt and run identifiers,
manifest hashes and runtimes differ. The coverage lessons print the same
results; the benchmark timings differ, as timings do, and AGE-MOEA's
recorded configuration also lists "variant": "agemoea", the field that
#359 adds. No lesson runs NSGA-III, and the study lessons run MOEA/D with
two objectives, so the corrections of #364 and #361 change none of them.
The prose cites no changed number; the prose that named 1.4.0 as the
tested release now names 1.5.0.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 19, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-19T07:41:12.213897Z be482f9 PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@github-actions

Copy link
Copy Markdown
Contributor

VAMOS documentation preview

https://pr-368-vamos-docs-preview.vamos-optimization.workers.dev

Built from be482f9. Canonical metadata points to the production .org domain; the preview itself remains isolated.

@NicolasRodriguezUribe
NicolasRodriguezUribe added this pull request to the merge queue Nov 19, 2026
Merged via the queue into main with commit d40896b Nov 19, 2026
20 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant