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

[fecolinhares]

Thinking Path

• Hermes WebUI allows users to manually trigger cron jobs from the Tasks panel
• Manual execution calls _run_cron_tracked() in api/routes.py
• This function wraps cron.scheduler.run_job() to track running state
• The function was calling run_job(job) without importing it first
• This caused NameError: name 'run_job' is not defined when clicking "Run Now" on any cron job
• The fix is a single-line import at the start of the function

What Changed

Added one import statement inside _run_cron_tracked():

def _run_cron_tracked(job):
    """Wrapper that tracks running state around cron.scheduler.run_job."""
    from cron.scheduler import run_job  # ← Added this line
    
    try:
        run_job(job)
    finally:
        _mark_cron_done(job.get("id", ""))

Files modified:

• api/routes.py (+2 lines: import + blank line)

Why It Matters

• User impact: Manual cron execution was completely broken — clicking "Run Now" crashed the background worker
• Scope: Affects all WebUI users who schedule cron jobs and want to test/run them manually
• Severity: High — feature advertised in UI but non-functional

Verification

• ✅ Python compilation: python -m py_compile api/routes.py — passes
• ✅ Existing cron tests: pytest tests/ -k cron — 45 passed, 9 skipped (require running agent)
• ✅ No other occurrences of run_job without import in the file (searched entire routes.py)

Risks / Follow-ups

• Risk: None — this is a straightforward missing import with no side effects
• Follow-up: Could add a test that mocks a cron job execution to catch similar issues in CI

Model Used

• Provider: Opencode Go
• Model: qwen3.6-plus

AI Usage Disclosure

• AI-assisted: Yes
• Provider: Opencode Go
• Model: qwen3.6-plus
• Human review: Full code and PR text reviewed before submission

---

Fixes #1310

[nesquena-hermes]

Confirmed the diff and the matching bug surface in master:

• api/routes.py line 95-100: _run_cron_tracked() calls run_job(job) but the only from cron.scheduler import run_job in the file is inside an unrelated function at line ~3503 (the manual-run handler), so the threaded wrapper raises NameError exactly when cron worker hits it.
• This PR's two-line addition (from cron.scheduler import run_job at function entry) is the minimal correct fix — function-local import keeps the existing module-level import structure intact and matches the lazy-import pattern already used at line ~3503.
• The diff is +2 / -0, no logic change, no API surface change, no test change required (existing cron tests still pass per the PR description).
• Reporter of Manual cron run crashes with NameError: run_job is not defined #1310 (ynott) already validated the diff matches their local workaround.

Looks safe to merge. Closes #1310.

🤖 Automated triage via nesquena-hermes

[nesquena-hermes]

Thanks for jumping on #1310 so quickly, @fecolinhares! 🙏

Closing this in favor of #1317 which lands the same one-line fix plus AST-based regression tests in tests/test_cron_run_job_import.py that pin the worker-thread scope so a future refactor can't silently re-break this. The test file is what tipped the scales — both fixes are otherwise identical.

The fix from #1317 is going out as part of release v0.50.245, with credit also pointing back to #1312 and #1329 in the changelog as parallel discoveries.

Really appreciate you investigating the worker-thread scope angle in your write-up — it matched the actual root cause exactly. Looking forward to the next one!