dsh-intermi

See and use the exact terminal your DSH Agent is using. dsh-intermi puts the Agent's existing persistent shell in DSH NEXT's right sidebar, so you can follow its work and step in when needed.

简体中文 · English

Screenshots

View an Agent's live persistent terminal beside its conversation:

The terminal selector uses concise names. Open it to switch shells or delete one from its row:

Why dsh-intermi?

Most terminal views show a separate shell. dsh-intermi attaches to the PTY already owned by the Agent: one session, one live stream, shared between you and the Agent.

• Watch live work — follow terminal input and output as the Agent works.
• Switch between Agents — choose a session by its DSH conversation title without closing other shells.
• Take over when needed — focus the terminal and type into the same shell. Press Ctrl+C to interrupt the foreground command; the Agent can continue in that shell.
• Keep the full output — PTY output is saved locally without a rolling disk limit. The visible scrollback has a memory-conscious default of 50,000 lines, adjustable from 5,000 to 100,000 in Terminal display settings.
• Avoid blind command repeats — when DSH's send result has no output, dsh-intermi recovers the recent PTY output or reports that the command state is uncertain; it never resends the command.
• Make it readable — choose a light or dark theme, or set custom foreground and background colors.

Install

In DSH NEXT's Plugin Market, install dsh-intermi from GitHub:

Lysander-9/dsh-intermi

If the market asks for a repository URL, use:

https://github.com/Lysander-9/dsh-intermi.git

Restart DSH NEXT if the market prompts you to do so, then open dsh-intermi in the right sidebar.

Get started

1. Start or select an Agent conversation.
2. Let the Agent run a terminal task. dsh-intermi directs its shell work to that Agent's persistent DSH terminal.
3. Select the conversation from the terminal selector. The terminal displays that Agent's live PTY.
4. Click inside the terminal to type. Use Ctrl+C to stop a foreground command without closing the shell.

A terminal appears after the Agent opens its persistent shell. Switching the selector only changes which shell you are viewing; other shells keep running.

Shells and saved output

• Each Agent conversation has its own persistent shell. dsh-intermi does not create a second, look-alike terminal.
• Closing or switching the sidebar view disconnects the view; it does not close the shell.
• Deleting a shell closes its PTY and running processes, and removes that shell's saved terminal output. Audit events remain stored locally.
• dsh-intermi appends observed PTY output to a local per-shell log and replays it when you reopen the view. Logs have no rolling disk cap and use local disk space until the shell is deleted. The in-memory scrollback defaults to 50,000 lines; change it from Terminal display settings (5,000–100,000 lines). Exiting DSH NEXT ends running PTY processes; the saved output remains available for read-only review, but the process itself does not resume after restart.
• Terminal history and command events are stored locally in DSH home, under itatmn/ (events.jsonl contains command events).
• To keep shell work attached to the audited PTY, standalone pwsh, bash, and ops_bash shell tools are blocked while dsh-intermi is active.

Requirements

• DSH NEXT >= 0.2.1-rc.2 and < 0.3.0
• A DSH Agent with access to the native persistent terminal tools

Development and acknowledgements

The Host entry is lib/index.js; the right-sidebar client is lib/client.js. The browser terminal uses vendored xterm.js assets. See THIRD_PARTY_NOTICES.md for attribution and license details.

License

The package currently declares UNLICENSED in its manifest.