[Jammy2211]

Summary

PyAutoArray#566 (PR #567) adds two optional keyword arguments, solver="pdip" and stats=None, to inversion_util.reconstruction_positive_only_from. The profiling harness's solver injections monkeypatch that function and assert (_assert_signature_covers) that the wrapper's signature covers the library's, so against the new library they failed. This PR makes both injections declare the new kwargs and forward them only to a library that accepts them, so the harness runs against the pre-#566 and the #566 library alike. Route (d)'s PDIP fallback now pins solver="pdip" explicitly, so a library config selecting its own certified solver cannot change what the harness falls back to.

Merge order: this PR may only merge AFTER PyAutoLabs/PyAutoArray#567 (library-first gate). This repo's lint workflow checks out PyAutoArray main, so test_fallback_pins_the_library_pdip_solver (which needs the #566 signature) skips there until #567 merges and asserts once it has; CI stays green on both.

Scripts Changed

• scripts/misc/likelihood_breakdown/library_solver_injection.py — certified injection declares solver/stats, forwards them via _library_kwargs only when the library has them; fallback pins solver="pdip".
• scripts/misc/likelihood_breakdown/fixed_light_numpy_solvers.py — numpy solver injection declares solver/stats and forwards them on the delegated JAX path only to a library that accepts them (inlined, no JAX import); injected kernel renamed injected to free the solver keyword.
• scripts/misc/test/test_fixed_light_library.py — tests for kwarg forwarding (old + new signature) and the pinned PDIP fallback (skips on a pre-#566 library).
• scripts/misc/test/test_fixed_light_numpy_solvers.py — tests for the numpy injection's signature coverage and forwarding.

Upstream PR

PyAutoLabs/PyAutoArray#567

Test Plan

• pytest scripts/misc/test against the NEW library (#567 branch, activate.sh): 795 passed / 5 skipped / 0 failed.
• pytest scripts/misc/test against the OLD library (PyAutoArray main 22e6d608 first on PYTHONPATH): 794 passed / 6 skipped / 0 failed — the extra skip is test_fallback_pins_the_library_pdip_solver, which pytest.skips when reconstruction_positive_only_from has no solver kwarg (commit 2b013f9); every harness path (old-signature forwarding included) passes, so the harness still works pre-release.
• ruff check . clean; ruff format --check . clean (208 files); build_readme.py --check exit 0; wall/check_submits.py --check 0 failing.
• Import smoke: AUTOLENS_PROFILING_SMOKE=1 python scripts/imaging/likelihood_breakdown/fixed_light_trace.py --mesh delaunay exit 0.

Heart RED override

• Heart verdict (re-read immediately before this PR, verbatim): RED release validation FAILED (stage integrate); YELLOW workspace validation not passing (4 failed, cloud#35579888156: autolens notebooks/cluster/modeling.ipynb, autolens notebooks/weak/a2744.ipynb, autolens scripts/cluster/modeling.py, +1 more); YELLOW manifest drift: hub organism blurb (organs present) — 7 mismatch(es) vs PyAutoMind/repos.yaml. Freeze: not frozen.
• Branch gates passed: profiling tests 795 passed vs the new library, 794 passed + 1 extra skip vs the old; ruff/format/readme/check_submits clean; import smoke exit 0; independent review (Brain review faculty) CLEAN at 0473c6a (2b013f9 adds only the test skip).
• Authorization: live user message i authorize, to the named issue-#566 (+ linked autolens_profiling fix) development override, per AUTONOMY.md "Human override for Heart RED (development only)".
• Scope: commit / push / pending-release PRs only. No merge, no release, no CI bypass; this does not fix Heart. Merge needs a separate human /prm, after PyAutoArray#567 merges and with required checks green.

Generated by the PyAutoLabs agent workflow.

🤖 Generated with Claude Code