dsh-keepalive

The dsh-keepalive plugin addresses the critical issue where the DeepSeek Harness (DSH) Web process cannot restart itself after exiting. Existing mechanisms like stream idle watchdogs and browser reconnect backoff do not supervise the host process, leading to service downtime. This plugin introduces a detached Node watchdog process that monitors health and automatically relaunches the same DSH binary with its arguments, ensuring continuous operation. It also provides manual restart controls and compatibility patches for seamless integration into DSH settings.

Beyond basic keepalive, dsh-keepalive's core capability is its optional repair agent. When repeated boot failures occur, the agent inspects the plugin workspace, takes snapshots, and rolls back changes to automatically fix issues like missing variable declarations or syntax errors without damaging source code. The repair process features real-time status pages and global progress banners for transparency. Additionally, the plugin optimizes front-end plugin failure reporting for dynamic-import errors, enhancing reliability.

This plugin is ideal for developers and operators using DeepSeek Harness 0.1.0-rc.6 or compatible versions, especially those relying on DSH Web environments for development, testing, or deployment. It suits scenarios where enhancing DSH process stability and automating failure recovery are priorities, but it is not designed as a high-availability cluster, security sandbox, or production-grade service manager.