PR #16881: fix(cli): warn about stale dashboard processes after hermes update

• Repository: NousResearch/hermes-agent
• Author: Societus
• State: closed | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/16881

Description (problem / solution / changelog)

Fixes #16872

What

After a successful update, scan the process table for running hermes dashboard processes and print a warning with their PIDs and restart instructions. Mirrors the existing pattern for gateway processes (lines 6809-7135 in main.py).

The dashboard is a long-lived server process users start and forget. Unlike the gateway (auto-restart via systemd/launchd), the dashboard has no service manager. It runs as a bare nohup or background process. When hermes update replaces files on disk, the running process holds the old Python code in memory while the JS bundle on disk gets updated, producing a silent frontend/backend mismatch.

Changes

• hermes_cli/main.py: Add _warn_stale_dashboard_processes() that scans for dashboard processes via pgrep (Linux/macOS) or wmic (Windows), called at the end of both _cmd_update_impl (git path) and _update_via_zip (zip path).
• tests/hermes_cli/test_update_stale_dashboard.py: Tests for the warning: fires when dashboard PIDs exist, silent when none found, self-PID excluded, errors handled without crashing.

How to test

1. Start hermes dashboard --port 9119 --no-open in a terminal
2. In another terminal, run hermes update (or the test suite)
3. Observe the warning message with the PID
4. Verify no warning when no dashboard is running

Changed files

• hermes_cli/main.py (modified, +79/-0)
• tests/hermes_cli/test_update_stale_dashboard.py (added, +106/-0)

PR #16895: fix(cli): warn about stale dashboard processes after hermes update

• Repository: NousResearch/hermes-agent
• Author: teknium1
• State: closed | merged: True
• Link: https://github.com/NousResearch/hermes-agent/pull/16895

Description (problem / solution / changelog)

Salvages #16881 (@Societus) with a follow-up tightening commit.

What this PR does

After hermes update finishes (both git and zip paths), scans the process table for running hermes dashboard processes and prints a warning with PIDs + restart instructions.

Why

v0.11.0 added X-Hermes-Session-Token. A dashboard process started before the update keeps the old Python backend in memory while the JS bundle on disk gets replaced — the new frontend sends headers the stale backend doesn't recognize, so every API call returns 401 with no visible error (HTML loads, all data empty). Fixes #16872.

The dashboard has no service manager (unlike the gateway, which systemd/launchd auto-restart), so we can only warn — not auto-kill.

Changes

• hermes_cli/main.py: new _warn_stale_dashboard_processes() called from _cmd_update_impl (git path) and _update_via_zip (zip path). Cross-platform: ps on Linux/macOS, wmic on Windows. Excludes self-PID. Swallows FileNotFoundError/TimeoutExpired/OSError.
• tests/hermes_cli/test_update_stale_dashboard.py: 10 unit tests — warning fires/doesn't, multi-PID, self excluded, missing binary, timeout, malformed lines, grep-line filter, and a regression guard for the previously greedy pattern.

Follow-up tightening (commit 63b71c52)

The original Linux branch used pgrep -f "hermes.*dashboard" — a greedy regex that matches any cmdline containing both words (e.g. a chat session discussing "dashboard" or an unrelated grafana/dashboard-server). Replaced with ps -A -o pid=,command= + the explicit patterns list already used on the Windows branch and in hermes_cli.gateway._scan_gateway_pids:

• hermes dashboard
• hermes_cli.main dashboard
• hermes_cli/main.py dashboard

Validation

• 10/10 unit tests pass
• E2E: spawned fake python3 -m hermes_cli.main dashboard --port 9119 via exec -a, confirmed detection. Also detected a real pre-existing dashboard on the same machine.

Closes #16881. Fixes #16872.

Changed files

• hermes_cli/main.py (modified, +92/-0)
• tests/hermes_cli/test_update_stale_dashboard.py (added, +177/-0)

PR #17149: test(update-autostash): include stdout in fake_run defaults so stale-dashboard scan doesn't crash

• Repository: NousResearch/hermes-agent
• Author: briandevans
• State: open | merged: False
• Link: https://github.com/NousResearch/hermes-agent/pull/17149

Description (problem / solution / changelog)

Summary

• Lifts a CI baseline introduced by 66b114238 — two cmd_update integration tests in test_update_autostash.py have been crashing on origin/main since the _warn_stale_dashboard_processes() call landed in _cmd_update_impl.

The bug

Commit 66b114238 ("fix(cli): warn about stale dashboard processes after hermes update") added a new

subprocess.run([\"ps\", \"-A\", \"-o\", \"pid=,command=\"], capture_output=True, text=True, timeout=10)

call inside _cmd_update_impl and reads result.stdout.split(\"\\n\"). The two cmd_update integration tests in tests/hermes_cli/test_update_autostash.py mock subprocess.run with a catch-all SimpleNamespace(returncode=0) that omits stdout. After 66b114238, both tests crash inside _warn_stale_dashboard_processes with:

hermes_cli/main.py:5309: AttributeError
'types.SimpleNamespace' object has no attribute 'stdout'

Affected tests:

• tests/hermes_cli/test_update_autostash.py::test_cmd_update_succeeds_with_extras
• tests/hermes_cli/test_update_autostash.py::test_cmd_update_retries_optional_extras_individually_when_all_fails

Reproduces deterministically on a clean clone of origin/main (verified at a830f25f7).

The fix

Add stdout=\"\", stderr=\"\" to the catch-all SimpleNamespace(returncode=0) returns in both tests' fake_run so the new ps-scan finds zero dashboard processes and short-circuits cleanly. The two explicit pip-install branches that returned bare SimpleNamespace(returncode=0) are extended the same way for symmetry — no production behavior changes, only test fixture hygiene.

This is intentionally test-only. The production path is already covered by the _warn_stale_dashboard_processes regression tests added in #17123 (and the wmic-encoding fix in #17074).

Test plan

• uv run --with pytest --with pytest-xdist python3 -m pytest tests/hermes_cli/test_update_autostash.py -v — 23 passed, was 21 passed + 2 failed on origin/main.
• uv run --with pytest --with pytest-xdist python3 -m pytest tests/hermes_cli/test_update_stale_dashboard.py -v — 10 passed (no regression).
• Regression guard: removed the stdout=\"\" additions on a clean origin/main clone, reproduced the original AttributeError: 'types.SimpleNamespace' object has no attribute 'stdout' failure with identical traceback at hermes_cli/main.py:5309. Restored the fix, both tests pass.

Related

• Source of the regression: 66b114238 ("fix(cli): warn about stale dashboard processes after hermes update"), fixes #16872.
• Production path is exercised by the regression suite added in #17123 and #17074.

Changed files

• tests/hermes_cli/test_update_autostash.py (modified, +4/-4)