dsh-admin

Day-to-day upkeep of a DeepSeek Harness Web GUI deployment—restarting the service, checking for updates, upgrading the Harness itself—normally means hopping into a terminal and chaining together a fragile sequence of commands. A restart, in particular, has to respect systemd cgroup isolation and sweep orphan processes off port :3080; get it wrong and zombie daemons linger. dsh-admin folds all of that into the Web GUI: a small status button sits in the session header, opening a floating management panel, and a full DSH Admin page lives under Settings, so every routine action is one or two clicks away without leaving the browser.

Three capabilities do the heavy lifting. Manual restart fires a backend priority ladder: first a systemd-run transient unit that executes the complete stop → clear :3080 orphans → clean start → health-check script inside an isolated cgroup; if unavailable it falls back through dsh-plugin-op, plain systemctl, and finally a socket-pid kill-and-relaunch, guaranteeing the service is never inadvertently torn down by its own unit. Automatic version checking compares the locally installed @deepseek-ai/dsh against the npm latest tag on page load and every six hours thereafter, respecting semver precedence (stable over release candidate), and surfaces the result as a green, yellow, or red dot so you always know at a glance whether you are current. The one-click update button appears only when a newer version exists, running the global npm install and auto-restarting the service in a single step so the new build is live immediately.

Built for operators running DeepSeek Harness globally via npm on a Linux + systemd host—people who would rather not memorize terminal incantations for every routine maintenance task and want restarts to be safe, health-verified, and reachable entirely from the browser.