lk251066/dsh-tui-pro

English | 简体中文

A multi-session terminal workbench for DeepSeek Harness. Keep a durable Assistant and the project sessions you actively maintain in one full-screen TUI.

Install · Latest release · dshfind · Report an issue

Quick start

You need Node.js ^22.19.0 || >=24.0.0 and DEEPSEEK_API_KEY configured outside this repository.

Already have the dsh CLI? Install the plugin and its profile layer once:

dsh plugin --profile tui add @lk251066/dsh-tui

Then launch it from any project directory:

dsh --profile tui

Without a global dsh installation, use the same profile through npx:

npx -y @deepseek-ai/dsh@latest plugin --profile tui add @lk251066/dsh-tui
npx -y @deepseek-ai/dsh@latest --profile tui

The package owns both the TUI plugin and its cordis.patch.yml profile layer. No second bundle package is required.

Why this TUI

One terminal, multiple projects

The Assistant is a durable session outside every project workspace. Active project sessions stay grouped by workspace in the fixed sidebar, so long-running work remains available without repeatedly leaving and resuming sessions.

A stable full-screen canvas

Only the transcript scrolls. The editor, current plan, live status, model, context usage, and active sessions remain in place while user messages, Markdown, thinking, tools, diffs, plans, todos, and subagents use distinct compact treatments.

Native DeepSeek Harness composition

The package is an out-of-tree dsh plugin bundle, not a fork of the harness. Sessions, tools, skills, models, permissions, compaction, and subagents continue to use the services provided by the selected dsh profile.

Session workflow

Launching from a directory opens its first manually ordered active project session. If the directory has none, dsh creates one and adds it to that workspace.

• /new creates another session in the current project.
• /new <path> creates a session for another project.
• /assistant returns to the fixed Assistant.
• /switch, Alt+Left, Alt+Right, or a sidebar click changes the active project session.
• /quit closes the current project session, preserves its history, and returns to the Assistant.
• /sessions browses complete history; /exit closes the TUI.

The Assistant can manage active project sessions and read their completed user/assistant dialogue. Hidden reasoning, tool traffic, and unfinished output are not exposed across sessions.

Interaction

Action	Control
Send immediate steering	Enter while a turn is running
Queue the next turn	Tab
Recover the latest queued message	Esc before cancellation
Recall the latest submission	Empty-input Up
Browse current-conversation checkpoints	Double Esc
Expand tool and context detail	Ctrl+O
Expand settled reasoning	Ctrl+R
Select and copy transcript text	Drag with the left mouse button
Attach a clipboard image	Alt+V
Control session memory	/memory, /memory on, /memory off

Run /help for the complete command list. Models, reasoning effort, skills, themes, settings, permissions, questions, and approvals use the fixed bottom area instead of centered popup windows.

Images and terminals

@ completes workspace files. Alt+V attaches a clipboard image when the selected model advertises image input. Windows uses PowerShell to read clipboard images, macOS requires pngpaste, and Linux requires wl-paste or xclip.

The TUI requires an ANSI-capable terminal. The sidebar hides below 65 inner columns, and a terminal height of at least 24 rows is recommended.

Configuration

- id: tui
  name: '@lk251066/dsh-tui'
  config:
    sidebarWidth: 32
    showReasoning: false
    maxToolOutputLines: 6
    maxMessageLines: 30

See the package reference for the supported entry path and exported composition plugins.

Project links

• npm package
• GitHub releases
• Community discussions
• Development setup
• Verification guide
• Change history
• Contributing

This plugin follows the developer-preview dsh release line. Each release is checked from source, from its packed npm artifact, through a clean public dsh profile, and through a real Linux PTY before publication.

License

MIT. Based on the original TUI implementation from DeepSeek Harness.