Bug: admin/restart is self-destructive on Windows (pkill/bash/tsx unavailable, daemon always killed)

[asorry75]

Bug Description

POST /api/v1/admin/restart is self-destructive on Windows: the daemon always dies and is never replaced, leaving the Memory Viewer unreachable.

Flow in server/routes/admin.ts:

1. terminateHermesChat() runs pkill -TERM -f "/bin/hermes" / "hermes chat" — pkill does not exist on Windows, so this returns false and does nothing (silently).
2. spawnReplacementDaemon() runs bash -c "sleep 3 && <node> <tsx> bridge.cts --daemon" — bash/tsx are not reliably available on Windows (and tsx lives in devDependencies, absent from the published package).
3. scheduleHermesShutdown(200) then sends SIGTERM to its own PID after 200 ms — this always succeeds.

Net result: the original daemon is killed 200 ms after the restart request and the replacement never comes up, so the viewer port goes silent. This also happens if the user only saves settings, because SettingsView.tsx triggers the restart overlay after save.

Steps to Reproduce

1. Windows host, memos-local-plugin running under Hermes.
2. Open viewer → Settings → save any change (or click the restart prompt).
3. Viewer becomes unreachable; the daemon process exits and nothing replaces it.

Expected Behavior

On Windows, restart should either (a) work — spawn a replacement via cmd.exe/PowerShell or a Windows-friendly mechanism — or (b) degrade gracefully: tell the user to restart Hermes instead of killing the daemon.

Suggested Fix

• Detect process.platform === "win32" and skip the pkill/bash replacement path; either restart in-process or instruct the user to restart the host agent.
• Only self-terminate after a replacement is confirmed listening (or skip the 200 ms suicide entirely).

Related

• fix: align Hermes restart UX #1586 align Hermes restart UX
• fix(bridge): preserve stdio connection on restart + increase import limit #1564 preserve stdio connection on restart

Environment

• OS: Windows
• Plugin: memos-local-plugin 2.0.12 (Hermes adapter)

[Memtensor-AI]

🤖 AutoDev has picked up this issue and started working on it.

Task ID: dbbd8654162c0789
Working branch: bugfix/autodev-2213-20260805050240965
Target branch: latest dev* branch
Workflow: opsp (analysis → coding → testing → PR)

I will post the PR link here once done. If I need more information, I will ask in the comments.

[Memtensor-AI]

✅ AutoDev task dbbd8654162c0789 completed.

Summary: Fixed #2213: POST /api/v1/admin/restart (and the same code path in /api/v1/admin/clear-data) no longer self-destructs the Memory Viewer daemon on Windows. On Windows portable installs the historical pkill + bash -c "... tsx bridge.cts --daemon" sequence silently failed while the follow-up SIGTERM-self-kill always succeeded, so the daemon died with no replacement — a fault that also triggered on every "Save settings" click because SettingsView.tsx posts to /admin/restart after each save.

The fix adds an isWindowsPlatform() guard in apps/memos-local-plugin/server/routes/admin.ts (with a new ServerOptions.platform seam so tests inject the platform cleanly). When the runtime is Windows AND no supervisor is detected (INVOCATION_ID / XPC_SERVICE_NAME absent), we now skip the pkill+bash+shutdown chain, keep the daemon alive, and return { ok: false, restarting: false, manualRestartRequired: true, message: ... } for /admin/restart; /admin/clear-data still tears down MemoryCore and unlinks the DB / WAL / SHM / bridge-status files but returns restarting: false, manualRestartRequired: true and lets the user restart Hermes themselves. Supervised Windows hosts keep the original self-shutdown path so a supervisor can respawn the daemon.

Viewer client updates: viewer/src/stores/restart.ts::triggerRestart() recognises the new payload and enters a new manualRestartRequired phase (instead of endlessly polling a healthy server), and RestartOverlay.tsx renders the phase with the same terminal treatment as restartFailed. Bilingual copy added under restart.manualRequired / restart.manualRequiredHint in viewer/src/stores/i18n.ts.

Verified: tests/unit/server/admin.test.ts — 7/7 green (3 new Windows-specific tests + 4 preserved). Broader tests/unit/server/ tests/unit/config/ — 137 passing, 1 pre-existing failure in http.test.ts::POST /migrate/openclaw/run that reproduces on the untouched base commit (confirmed via git-stash sandwich). Both tsc -p tsconfig.json and tsc -p tsconfig.viewer.json complete with no errors. cc @whipser030 @hijzy

Base branch: main
Branch: bugfix/autodev-2213-20260805050240965
Commit: 2256ff2
PR: #2214
Assigned to: @hijzy
Reviewers: @whipser030, @hijzy