fix(cron): record no_agent runs as sessions so they show in run history

[yingliang-zhang]

Problem

no_agent: true cron jobs short-circuit run_job() (cron/scheduler.py) before any SessionDB work, so they never produce the cron_{job_id}_{timestamp} session row that the run-history endpoint (GET /api/cron/jobs/{id}/runs, backed by SessionDB.list_cron_job_runs) is built from. Manually triggering such a job from the Desktop GUI gives zero feedback: no running indicator, no run record, no output — the script runs fine, the GUI just can't see it.

Reported in #44080 and #42433.

Fix

In the no_agent branch of run_job():

• Before executing the script, create the run session (cron_{job_id}_{ts}, source='cron' — the exact shape the agent path produces) plus a short user message (no_agent script: <path>) that becomes the run's preview. Because the session is open while the script runs, the runs endpoint's existing is_active computation now yields a running indicator for in-flight manual triggers, with no frontend changes needed.
• After execution, persist the outcome doc (status + script output, already redacted by _run_job_script) as an assistant message, title the session using the same scheme as the agent path, and end_session(..., "cron_complete").

This covers all four exit paths: success, script failure, empty-stdout silent run, and wakeAgent=false silent run.

Design notes

• Best-effort recording. A missing/broken state store degrades to the old no-record behaviour and never blocks the script run (guarded the same way the agent path guards its SessionDB use).
• The no_agent cost contract is preserved. run_agent/AIAgent are still never imported on this path (existing test_run_job_no_agent_never_invokes_aiagent still passes); the only addition is a cheap sqlite session write.
• No new surface. The Desktop GUI, /runs endpoint, and list_cron_job_runs already do the right thing once the session rows exist — the fix is purely making the no_agent path produce the records everything else already consumes.

Tests

Added to tests/cron/test_cron_no_agent.py:

• test_run_job_no_agent_success_records_run_session — success run appears in list_cron_job_runs with source='cron', ended state, cron_complete, and the script output persisted in the run's messages
• test_run_job_no_agent_failure_records_run_session — failed run is recorded (not silent) with cron_failed end reason
• test_run_job_no_agent_silent_records_run_session — silent run (empty stdout) still leaves a run record
• test_run_job_no_agent_broken_session_store_does_not_break_run — a broken SessionDB raising doesn't break the run itself

The shared hermes_env fixture now also reloads hermes_state so DEFAULT_DB_PATH (bound at import time) points at the per-test HERMES_HOME — keeping the new tests hermetic.

tests/cron/: 22 passed, 0 failures.

Relationship to existing PR

Supersedes #44087 — same approach, rebased on current main (1479 commits ahead of #44087's base) so it merges cleanly without conflicts.

Fixes #44080.
Related: #42433, #41935.

[alt-glitch]

This was generated by AI during triage.

Duplicate of #44087 — same fix for #44080: both create the cron_{job}_{ts} session row in the no_agent branch of run_job() (cron/scheduler.py) so manually-triggered no_agent jobs show a running indicator + run record in the Desktop GUI, while preserving the no_agent cost contract. #44087 (open since 2026-06-11) is the earlier canonical PR. Flagging for a maintainer to pick the canonical fix.

[yingliang-zhang]

Closing as duplicate of #44087 (open since 2026-06-11, earlier canonical). Both create the cron_{job}_{ts} session row in the no_agent branch of run_job() so manually-triggered no_agent jobs show in run history. Deferring to the earlier PR.