fix(cron): import run_job inside _run_cron_tracked to fix NameError

[jasonjcwu]

Problem

_run_cron_tracked() runs inside a worker thread (threading.Thread) at api/routes.py. It called run_job(job) but run_job was only imported as a local variable inside _handle_cron_run() — a completely separate function scope. When the thread executed, run_job was not in scope → NameError.

This caused "Run Now" in the cron panel to crash every time.

Fix

Moved from cron.scheduler import run_job into _run_cron_tracked() itself so the import resolves in the worker thread's scope. Removed the now-redundant import from _handle_cron_run().

Files changed:

• api/routes.py — moved import into _run_cron_tracked, removed from _handle_cron_run
• tests/test_cron_run_job_import.py — 3 AST-based regression tests

Testing

[jasonjcwu]

Fixes #1312 and Fixes #1310

[nesquena-hermes]

Thanks for the fix and especially for the regression tests — the AST-based check that run_job is actually imported inside _run_cron_tracked is a nice safety net for this exact category of bug.

Confirmation

Diagnosis is correct, and matches what's in master:

• api/routes.py:95 — _run_cron_tracked() calls run_job(job) but has no import.
• api/routes.py:3503 — _handle_cron_run() imports run_job at function-local scope. That import is invisible to _run_cron_tracked, which executes in a worker thread spawned at line 3514 (threading.Thread(target=_run_cron_tracked, ...)).

Result: NameError: name 'run_job' is not defined whenever a manual cron run is dispatched. The diff resolves both halves cleanly — local import added to _run_cron_tracked, redundant import removed from _handle_cron_run.

Overlap with #1312

This PR overlaps with the existing #1312 (same one-line import fix, same hunk). The substantive differences in your PR:

1. You also remove the now-redundant from cron.scheduler import run_job from _handle_cron_run(). That cleanup is correct (the import is no longer needed there) and is a small improvement fix(cron): import run_job in _run_cron_tracked to fix NameError #1312 doesn't have.
2. You add three regression tests in tests/test_cron_run_job_import.py. The AST-based approach is robust against future refactors and exactly the kind of test that would have caught this defect at PR-review time.

A small test-code note (non-blocking)

In TestRunCronTrackedImport.test_run_job_imported_inside_function there's some dead code:

ImportCollector = type(
    "ImportCollector",
    (ast.NodeVisitor,),
    {
        "imports": set(),
        "visit_ImportFrom": lambda self, node: (
            self.imports.add(a.name for a in node.names),
        ),
    },
)

This class is constructed but never instantiated or visited — the actual import collection happens in the loop below. You can drop the entire ImportCollector block; the test still passes and is easier to read. Also note set.add(<generator>) would put a generator object in the set rather than the names, so even if it were used it wouldn't behave as intended.

Recommendation

Either of #1312 or #1317 fixes the runtime bug. This PR is the more complete one (cleanup + tests). I'd suggest:

• Drop the dead ImportCollector block.
• Either close fix(cron): import run_job in _run_cron_tracked to fix NameError #1312 in favor of this, or cherry-pick the cleanup hunk + tests into fix(cron): import run_job in _run_cron_tracked to fix NameError #1312 (whichever the author prefers).

Maintainer will pick. Thanks for the careful work — the regression tests in particular are valuable.

🤖 Automated triage via nesquena-hermes

[nesquena-hermes]

Shipped in v0.50.245 via batch release PR #1334 (merge commit 52e1567). Thanks @fxd-jason! 🙏

Live now at https://github.com/nesquena/hermes-webui/releases/tag/v0.50.245.

Released alongside 9 other contributor fixes — see the v0.50.245 entry in CHANGELOG.md.