[soodoku]

Batch dopt_average, dopt_pc, and marginal_var_sum omitted previously assigned members. This change scores past and newly selected members together while keeping only incoming units eligible. dopt_pc uses a common PCA basis; max_pc and min_pc retain their incoming-batch projection.

The audit also reproduced cancellation in the existing dispersion helper: adding a common offset of 1e10 to the example [20, 0, 10] with prior member 100 changed scores [3200, 5000, 4050] to zeros and selected the wrong unit. Centered differences now preserve the scores and choice. Mixed-type labels are preserved; invalid ridge scales and tie rules are rejected. Documentation distinguishes tested R parity from deliberate corrections.

Prepare bug-fix release 0.1.1, align the citation/changelog, update the publishing action to its latest stable release, and temporarily exclude only the two exact Pepy URLs returning HTTP 503.

Validation:

• Same Python 3.13 environment: baseline 177 tests pass; candidate 227 pass with 95% coverage. All 50 allocation contract tests pass; 34 fail against baseline production code. The eight new offset regressions separately fail on the original PR and pass with the numerical correction.
• Independent GPT-6 Astra review: 227 tests pass, lint/format/types/docs pass, and 450 randomized batch allocations match independent centered-sum and eigenvector oracles. No correctness findings.
• Ruff, Pyright, pydoclint, all pre-commit hooks, duplicate-code scan, Sphinx with warnings as errors, wheel/sdist builds, and Twine metadata checks pass.
• Final commit 4218f20: all hosted checks pass, including Python 3.12/3.14, wheel, lint, docs, dependency review and lock regression. Strict Preen passes locally. Clean-wheel suite: 227 passed. Final metadata delta independently reviewed with no findings. Registry README rendered and checked against wheel metadata. No publication has occurred yet.

[chatgpt-codex-connector]

Codex Review Summary

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

Review	Status	Commit	Review trigger
📝 Code Review	⚠️ Failed 2026-11-02T16:49:33.537309Z We couldn’t complete this review. Comment @codex review to try again.	5f143ed	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.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.