[Pigbibi]

AI assistance

Codex (GPT-6; host-selected configuration).

What changed

Pin UES to c16c81bdf21824615f7280554559adc6c721917c in pyproject.toml and uv.lock. Align V7_APPROVED_UES_REVISION and its paused-process regression test with that exact revision. UES #715 fixes the Russell Top50 cash_reserve_ratio TypeError.

Add a build-time installed-revision check after frozen dependency installation. The check requires exact PEP 610 Git commit evidence and manifest agreement; missing, invalid, unreadable or mismatched evidence fails with fixed diagnostics. Add adverse metadata and controlled-revision regression coverage.

The Russell platform context now reads the independently supplied
market_regime_control_expected binding and market_regime_control signal
from the existing strategy configuration on every evaluation. The signal is
an artifact; the binding is runtime configuration. Missing or invalid evidence
remains fail-closed; no expected digest or healthy signal is manufactured.

How verified

• Full pytest suite: 2934 passed, 1 skipped, 335 subtests passed.

• Focused runtime, regime, UES revision and V7 application suite: 72 passed; 7 subtests passed.

• Before the V7 constant update: 2 regression failures / 3 passes; after repair, both failures pass and unapproved revisions remain rejected.

• Installed-revision verification, frozen dependency sync, lock consistency, Ruff and diff hygiene: PASS.

• Documentation check: zero blocking findings; one pre-existing README advisory.

• Offline bounded workflow shell suite: PASS (92 image-only cases, 9 lifecycle cases); source-build authentication exclusion checks: PASS; Dot review Node suite: 11 passed.

• Docker build not run because Docker is unavailable. Hosted CI and actual broker/runtime adoption are separate and unproven.

• Regime regression: the original context builder fails 5 of 10 final cases;
  the repaired integration passes all 10. Real UES ranking and risk gates run
  against synthetic snapshot/evidence and explicit synthetic risk limits.

• Source-aware scan of the complete PR diff: zero identifier finding lines.

• The new commit has a compliant header/body. The previous commit's long body
  remains unchanged under the explicit no-amend/no-rebase/no-force-push scope.
  CI validates all PR commits, so a new commit cannot remove that old failure.

Why

The previous dependency revision lacks the Russell Top50 cash-reserve repair. Updating only the dependency pin causes the controlled V7 PAPER readback to reject the installed revision; the explicitly authorized constant change keeps that gate strict and consistent.

Risk

High: dependency installation, Docker build verification and controlled PAPER revision paths are affected. No activation, order/risk policy, protected switch, workflow approval, label, deployment or production configuration is changed. Runtime recovery and serving-image overlay applicability are not established by offline tests. Review and any owner approval remain with the established review-and-merge process.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

• X
• Mastodon
• Reddit
• LinkedIn

[Pigbibi]

Superseded by #700 (single-commit recut, identical reviewed tree).