ci: drop the vestigial jax<0.7 pin from smoke_install.sh

[Jammy2211]

Summary

This repo's install epilogue carried the same stale line as autolens_workspace_test, where it was found and diagnosed — see PyAutoLabs/autolens_workspace_test#266 and its companion PR PyAutoLabs/autolens_workspace_test#268.

.github/scripts/smoke_install.sh pinned jax<0.7 jaxlib<0.7. That pin was added in autolens_workspace_test#82 for exactly one reason: keeping tensorflow-probability==0.25.0 importable, since tfp.substrates.jax referenced jax.interpreters.xla.pytype_aval_mappings, removed in JAX 0.7.0. autolens_workspace_test#184 then deleted that dependency when the stack moved to tfp-nightly (pinned by PyAutoArray[optional]). The pin outlived what it protected.

It was not merely inert. jax is a base dependency of autonerves (jax>=0.7.0,<0.12.0, PyAutoLens#702), so the preceding line already installed a conforming jax; the pin then downgraded it to 0.6.2 — unsupported by the stack — and only the following [optional] re-resolution repaired it. The install was correct by line ordering, not by constraint, and a reordering would have dropped the smoke suite onto jax 0.6.2 silently.

This removes the pin, leaves a comment recording the trail, and asserts the resolved jax version at the end of the script so the outcome is checked rather than inferred from a green run.

Scripts Changed

• .github/scripts/smoke_install.sh — removed pip install "jax<0.7" "jaxlib<0.7"; replaced with a comment recording why no jax pin belongs here (issue references qualified as autolens_workspace_test#82 / #184, since bare numbers would resolve to unrelated issues in this repo); appended a post-install assertion that the resolved jax falls in [0.7.0, 0.12.0), printing the version to the CI log.

No workspace scripts/ or notebooks/ are touched — this is CI install configuration only.

Test Plan

• Smoke tests pass for all affected workspaces
• The install log prints resolved jax <version> with a version in [0.7.0, 0.12.0)
• The JAX likelihood-function smoke scripts pass — they are the scripts that would break had the install genuinely landed on 0.6.2

Validation already performed

• bash -n clean on the modified script.
• Guard boundary cases unit-tested: 0.7.0 / 0.9.0 / 0.10.2 / 0.11.4 pass; 0.6.2 / 0.12.0 / 1.0.0 fail.
• The guard was executed as a real bash heredoc against a stub jax: 0.10.2 prints resolved jax 0.10.2 and exits 0; 0.6.2 and 0.12.0 exit 1 with the explanatory message. With set -e at the top of the script, that aborts the install step.
• Confirmed PyAutoHeart/.github/workflows/smoke-tests.yml uses actions/setup-python@v5 on ubuntu-latest, so python is on PATH.

Implementation notes

• The assertion compares a major/minor tuple rather than using packaging.version, avoiding a dependency this epilogue does not otherwise declare.
• import jax is deliberately unguarded: this workflow runs on ubuntu-latest, where autonerves' platform marker always installs jax, so an absent jax is itself a failure worth catching.

Sibling repos

autocti_workspace_test and autofit_workspace_test were also checked; neither carries a jax pin, so no change is needed there.

Generated by the PyAutoLabs agent workflow.

---

Generated by Claude Code