dsh-web-watchdog

Developers running dsh web on Windows have likely hit the same wall: low-level issues such as Node 24 libuv exit assertions or native TS type stripping can crash the host process without warning, leaving a dead terminal window behind. Worse, the exit code, the tail of stderr/stdout, and the Windows event log are scattered across different sources, making post-mortem debugging painful.

dsh-web-watchdog solves this by spawning a fully independent PowerShell supervisor process from within the host. When the host exits abnormally, the supervisor appends the exit code, the tail of the host stdout/stderr, and the relevant Windows event log entries to crash.log. It then restarts the process using an exponential-backoff policy (default: 10 s initial delay, 600 s cap, up to 10 consecutive restarts). A clean exit (code 0) will not trigger a restart. A panel button or the REST endpoint POST /api/watchdog/restart can initiate a scheduled restart; the supervisor completes the stop-old / start-new cycle within two seconds. After the new host takes over, it spawns its own supervisor, and the old one exits automatically once it detects the pid file has been claimed, preventing double supervision.

This plugin is aimed at Windows developers who rely on dsh web daily and want automatic recovery plus complete crash logging without installing system-level services. It works out of the box after installation, and endpoints like /api/watchdog/status expose host PID, supervisor health, configuration, and the tail of the crash log at any time.