Operate processes and shared live consoles.
Long-running applications have managed lifecycle state, the latest process identity or exit code, resource metrics, and a shared browser console when the workload supports a PTY.
Authorized operators can start, stop, restart, or force-kill a process. Prefer a normal stop or restart so the application can handle its configured signal and graceful-drain timeout. Force-kill is for a process that no longer responds.
The process view shows PID, start time, current state, and the most recent exit code.
The console is shared per application, similar to a game-server panel. Every viewer sees the same output and process-state stream, while input remains tied to the viewer's signed session.
When a viewer joins, ServerChirp loads scrollback and then splices in live output by byte offset to avoid a gap or duplicate lines. The PTY uses the smallest terminal size requested by active viewers.
Console input carries a per-session sequence number and an agent-minted challenge nonce that binds it to the active agent process. Signed input cannot be replayed against a restarted agent.
While the process view is open, the agent reports CPU, memory, disk-read, and disk-write samples on a short watch lease. The lease must be renewed and expires agent-side, so a closed browser does not leave permanent high-frequency sampling behind.
The process state retains the latest exit code rather than a separate exit-history ledger. Use it with surrounding console output, deployment activity, and server metrics before changing the restart policy.
