dsh-withSsh

When you are debugging a server issue, you typically end up bouncing between a chat with an AI assistant and a separate terminal window, pasting logs back and forth with no unified record of who ran what. dsh-withSsh places AI collaboration, a real SSH PTY terminal, and a full audit trail on a single page inside a DeepSeek Harness conversation, so every command, its remote output, exit code, and the operator behind it stay visible and reviewable.\n\nKey capabilities include a left-side server overview (CPU, load, memory, swap, uptime); a scrollable persistent PTY in the top-right where human commands appear in blue and AI commands in yellow with a mandatory Chinese description; a read-only file tree rooted at / in the bottom-right; and an ssh_exec tool that lets the AI run commands whose stdout, stderr, and exit code are written to both the terminal stream and the conversation log. Passwords never enter the model context or are persisted, and high-risk commands trigger a Harness approval gate.\n\nIt is aimed at ops and SRE engineers performing routine inspections and incident triage, backend developers who want to verify remote command results in the same thread as their AI discussion, and team leads or on-call engineers who need to clearly distinguish human-executed actions from AI-suggested ones.