feat: align build and install checks with Python 3.12

[Jammy2211]

Overview

Align PyAutoHands' scheduled Python census and PyAutoHeart's deep install verification with the ecosystem's newly merged Python 3.12 floor. Required evidence must cover 3.12 and 3.13, Python 3.11 must prove an exact floor release is rejected, and Python 3.14 must remain isolated, visible, and non-required while its known factor-graph regression is outstanding.

This is Phase 2 of the coordinated Python-floor campaign. It changes build and health verification only; it does not publish, yank, tag, or release any package.

Plan

• Reduce required scheduled library and workspace coverage to Python 3.12 and 3.13, removing obsolete below-floor conditionals and support claims.
• Add a distinct, non-required Python 3.14 evidence leg that reports failures without weakening the required 3.12/3.13 jobs.
• Rewrite deep install Check B around the new contract: same exact target succeeds on 3.12/3.13 and is rejected by package metadata on 3.11.
• Add focused regression coverage for the matrix and rejection semantics, then update adjacent help/comments/documentation.
• Run both repository suites, shell/YAML validation, the full manually dispatched scheduled matrix, and install verification without performing a release.

Detailed implementation plan

Work Classification

Library workflow across two organism repositories: PyAutoHands (primary) and PyAutoHeart.

Affected Repositories

• PyAutoHands (primary)
• PyAutoHeart

Branch Survey

Repository	Current Branch	Dirty?
./PyAutoHands	main	unrelated untracked run_logs/; tracked tree clean and equal to origin/main
./PyAutoHeart	main	clean and equal to origin/main; clean campaign worktree already exists at merged Phase 4 head

The existing python-312-floor campaign owns PyAutoHeart, so this is a same-task continuation rather than a cross-task worktree conflict. PyAutoHands will be attached in the isolated campaign root; its unrelated main-checkout run_logs/ will not be touched.

Suggested branch: feature/python-312-floor

Worktree root: /home/jammy/Code/PyAutoLabs/.codex-worktrees/python-312-floor/

Implementation Steps

1. In PyAutoHands/.github/workflows/python_matrix.yml, set required unit-test and workspace-smoke matrices to exactly Python 3.12 and 3.13. Remove the obsolete 3.11-only JAX branch and old 3.9–3.11 support/banner summary.
2. Add an explicitly experimental Python 3.14 job in the same workflow. Keep it isolated from required jobs, run from clean setup-python environments, use continue-on-error, retain per-library visibility, and report its result separately so it cannot turn required failures green.
3. Add or extend PyAutoHands tests (under tests/) to pin the required version lists, the isolated 3.14 contract, and the absence of stale below-floor support wording. Parse the workflow rather than relying only on grep where practical.
4. In PyAutoHeart/heart/checks/verify_install.sh, replace Check B's 3.9/3.10/3.11 successful-import and soft-banner behavior. Resolve or accept one exact target version, require that target to install/import on 3.12 and 3.13, and require Python 3.11 pip to reject that exact target specifically because of Requires-Python >=3.12. Treat an unexpected success, a network/index error, or an unrelated resolver failure as FAIL rather than a false expected rejection.
5. Update PyAutoHeart/tests/test_verify_install_script.py with fast contract tests for supported-version wiring, exact-target reuse, expected-rejection classification, help text, and removal of banner assertions. Update adjacent workflow comments and docs/release_validation.md where Check B's meaning is described.
6. Validate with bash -n, YAML parsing/action structure checks, the full PyAutoHands and PyAutoHeart test suites on supported interpreters, and a manually dispatched python_matrix.yml run on the feature branch. Exercise Check B end to end against an exact floor artifact available without publication; the authoritative TestPyPI exact-wheel proof remains the coordinated Phase 3 rehearsal.
7. Build immutable review commits, obtain formal ReviewSurface and local Claude Opus 5 CLEAN judgments, refresh Heart, and ship separate pending-release PRs. Merge only exact reviewed heads after all checks are green under the standing human authorization; do not release or close issues.

Key Files

• PyAutoHands/.github/workflows/python_matrix.yml — scheduled required and experimental Python evidence.
• PyAutoHands/tests/ — workflow contract regression coverage.
• PyAutoHeart/heart/checks/verify_install.sh — deep install-path contract and result reporting.
• PyAutoHeart/tests/test_verify_install_script.py — fast install-check surface tests.
• PyAutoHeart/.github/workflows/workspace-validation.yml — release-rehearsal Check B comments/caller contract if needed.
• PyAutoHeart/docs/release_validation.md — exact-wheel verification description if needed.

Original Prompt

Click to expand starting prompt

Python 3.12 floor — Phase 2: build and health verification

Type: feature
Target: PyAutoHands
Difficulty: large
Autonomy: supervised
Priority: high
Status: formalised

Parent: python_312_ecosystem_floor.md
Depends on: python_312_floor_phase_1_core.md

Scope

Align @PyAutoHands and @PyAutoHeart with the core floor: required 3.12/3.13
coverage, explicit 3.11 install rejection, and a clean isolated non-required
3.14 evidence leg. Remove old 3.9-3.11 success/banner assertions and duplicated
matrix work without weakening release/install gates.

Gates

Full scheduled matrix and end-to-end install verification pass, and a fresh
Heart verdict contains no new campaign-attributable RED reason.

[Jammy2211]

Phase 2 implementation checkpoint

PyAutoHands worktree (feature/python-312-floor, base ef0f5a3):

• Required weekly unit matrix reduced from 3.9–3.13 to exactly 3.12/3.13.
• Workspace smoke matrix reduced from 3.11–3.13 to exactly 3.12/3.13; obsolete conditional JAX enablement removed.
• Separate five-library Python 3.14 evidence matrix added with continue-on-error: true; it is visible but non-required.
• Old supported-with-warning/banner/JAX-3.11 summary removed.
• Full suite: 239 passed on Python 3.12; 239 passed on Python 3.13.
• Clean isolated local 3.14 evidence: Nerves 154p; Array 929p; Fit 1557p/3s; Galaxy 999p/1s; Lens 488p. This does not promote 3.14 because the known factor-graph defect is workspace-level.

PyAutoHeart worktree (feature/python-312-floor, base merged main 8fb1171):

• Check B now installs/imports one exact target on 3.12 and 3.13, then requires pip to reject that exact target on 3.11 specifically for autolens Requires-Python >=3.12; generic resolver/network/dependency failures do not pass.
• Added --find-links for honest pre-publication exact-wheel evidence. Sidecars record index: find-links, and readiness now keeps a passing local artifact STALE until PyPI/TestPyPI evidence replaces it; local failure remains RED.
• Adjacent help, release-validation workflow comments, and docs updated.
• Full suite: 294 passed on Python 3.12; 294 passed on Python 3.13; shell and YAML parse checks pass.
• Real local-wheel Check B: 3.12 PASS, 3.13 PASS, 3.11 rejection PASS, 0 skips; JSON ready=true/version=1.0.dev0/index=find-links.

No upload, tag, release, yank, or issue closure occurred. Next: immutable formal + Claude Opus 5 review, then exact-tree mechanical ship. The full hosted weekly matrix will be manually dispatched on the reviewed PyAutoHands branch before merge.

[Jammy2211]

Phase 2 exact-head review gate is CLEAN.

Reviewed heads:

• PyAutoHands f2e990e0929d929ff84ccd21ad5bf2af92f9181e
• PyAutoHeart 377833355b3e5408fa0afa573181297404752384

Validation on those exact heads:

• PyAutoHands: 4 focused tests; full suite 240 passed on Python 3.12 and 240 passed on 3.13.
• PyAutoHeart: 147 focused tests; full suite 301 passed on Python 3.12 and 301 passed on 3.13.
• YAML parsing, shell syntax, and git diff --check pass.
• Live five-wheel Check B: exact autolens==1.0.dev0 installs/imports on 3.12 and 3.13; the identical pin is rejected on 3.11 specifically for Requires-Python >=3.12; sidecar ready=true, index=find-links.
• Independent Claude Opus 5 max-effort review: CLEAN on these exact hashes after two remediation rounds.

Heart refresh remains the accepted unchanged baseline: YELLOW 65, no RED:

1. workspace validation not passing (13 failed, 2026-07-21T19-05-22Z)
2. 33 stale parked scripts
3. tenant-firewall manifest drift (6 mismatches)
4. release validation stale because the five core sources moved

Next: push the unchanged reviewed heads, open pending-release PRs, run all PR checks plus the full hosted Python matrix, and merge only if the exact heads remain unchanged and required checks are green. No release/tag/upload/yank is authorized in this phase.

[Jammy2211]

Phase 2 merged unchanged after the exact-head gate.

• PyAutoHands PR ci: align Python support matrix with 3.12 floor #207 merged as 1e9ac6d5a0811c1968b545d4c3c5fb4d6c4c4c7b from reviewed head f2e990e0929d929ff84ccd21ad5bf2af92f9181e.
• PyAutoHeart PR Note repo-list identity check against PyAutoMind/repos.yaml #115 merged as eda92a6b93d35f26a7b2a5cf168cd185f7ae9b53 from reviewed head 377833355b3e5408fa0afa573181297404752384.
• Hosted Python matrix run 30453073189 completed 22/22 jobs successfully on the exact Hands head.
• Heart PR checks: all Python 3.12/3.13 jobs passed.
• Canonical main branches are fast-forwarded; feature branches and temporary review/task worktrees are removed.
• Issue remains open as the campaign record; no release, tag, upload, yank, or issue closure was performed.

The campaign now enters Phase 3 coordinated core release readiness. Manual release execution remains human-required and will stop for the version/release decision.