fix(update): detect profile dashboard processes in stale-dashboard sweep

[liuhao1024]

What does this PR do?

Fixes the stale-dashboard sweep after hermes update so that dashboard processes launched with --profile <name> (non-default profiles) are correctly detected and killed.

Root cause: _find_stale_dashboard_pids() uses substring matching against the process command line (e.g. "hermes dashboard"). When a non-default profile launches a dashboard with --profile <name> between the binary and the subcommand — hermes --profile bruce dashboard --isolated — the contiguous pattern never matches. The process is invisible to the sweep and continues running stale Python code after the update, causing ImportError when new symbols (like is_output_cap_error) are added.

Fix: Strip --profile <name> / -p <name> tokens from the command line (and collapse resulting double-spaces) before pattern matching. Applied to both the POSIX (ps) and Windows (wmic) code paths.

Related Issue

Fixes #56717

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)

Changes Made

• hermes_cli/main.py: Normalise process command lines in _find_stale_dashboard_pids() by stripping --profile/-p flags before pattern matching (both POSIX and Windows code paths). +21/-2 lines.
• tests/hermes_cli/test_stale_dashboard_profile_detection.py: 3 regression tests covering --profile, -p, and default (no profile) cases.

How to Test

1. Run python -m pytest tests/hermes_cli/test_stale_dashboard_profile_detection.py -v — all 3 tests should pass.
2. Run python -m pytest tests/hermes_cli/test_update_stale_dashboard.py tests/hermes_cli/test_dashboard_lifecycle_flags.py -v — all 22+10 existing tests should pass (no regressions).
3. Manual verification: start a profile dashboard (hermes --profile test dashboard), then run hermes update — the profile dashboard should be detected and killed (previously it was invisible).

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS 26.4.1

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — or N/A
• I've updated cli-config.yaml.example if I added/changed config keys — or N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide
• I've updated tool descriptions/schemas if I changed tool behavior — or N/A

[xiawiie]

Hi @liuhao1024, thanks for putting together #56723. I checked it against #56717 and the root cause you identified looks right: a profile selector between hermes and dashboard makes the stale-dashboard sweep miss non-default profile dashboard processes.

While testing that path, I found two nearby edge cases in the same stale-runtime class:

• Hermes also accepts equals-form profile selectors (--profile=bruce, -p=bruce), which the current regex normalization does not strip.
• The systemd gateway discovery still used the narrower hermes-gateway* glob in update-related service-PID discovery, which can miss profile-shaped units such as hermes-secondbrain-gateway.service.

I opened #56745 as a draft follow-up built on top of your branch/commit rather than reimplementing the fix from scratch, so your original work is preserved. The extra changes are intentionally small: token-aware profile-selector stripping for dashboard/serve process scans on POSIX and Windows WMIC, a shared GATEWAY_SYSTEMD_UNIT_GLOB = "hermes*gateway*", and regression tests for the added cases.

Validated locally with:

scripts/run_tests.sh tests/hermes_cli/test_stale_dashboard_profile_detection.py tests/hermes_cli/test_gateway.py tests/hermes_cli/test_update_stale_dashboard.py tests/hermes_cli/test_dashboard_lifecycle_flags.py tests/test_windows_subprocess_no_window_flags.py -q
# 106 tests passed, 0 failed

uv run --frozen --extra dev ruff check hermes_cli/main.py hermes_cli/gateway.py tests/hermes_cli/test_stale_dashboard_profile_detection.py tests/hermes_cli/test_gateway.py tests/test_windows_subprocess_no_window_flags.py
# All checks passed

Happy to adjust or close #56745 if maintainers would rather fold these follow-ups directly into this PR.

[liuhao1024]

Thanks for the thorough review and the follow-up, @xiawiie! Glad the root cause analysis held up.

The two additional edge cases you found — equals-form profile selectors and the narrower systemd glob — are good catches. I see #56745 extends the fix with token-aware profile-selector stripping and a shared GATEWAY_SYSTEMD_UNIT_GLOB. Building on top of the existing commit rather than reimplementing is the right approach.

Happy to rebase or adjust this PR if maintainers prefer folding the follow-ups directly here, or if #56745 lands first I can close this one.