dsh-TUI - An open-source terminal interface plugin for DeepSeek Harness.
dsh-TUI is a Claude Code-style terminal interface plugin for DeepSeek Harness, developed by the community and featured on the official DSH WeChat account. The tool offers a pixelated whale top bar, real-time status bar, flow-based thinking expansion, and dual...
What is dsh-TUI?
dsh-TUI is a Claude Code-style terminal interface plugin for DeepSeek Harness, developed by the community and featured on the official DSH WeChat account. The tool offers features such as a pixelated whale top bar, real-time status line, streaming thought expansion, and double-clicking Esc to roll back the session. It fully supports DSH capabilities including Agent presets, Skills, and MCP. With zero core modifications, it can be installed and uninstalled seamlessly via npm with a single click. Optimized for long sessions, it's ideal for developers who prioritize keyboard efficiency.
Main functions of dsh-TUI
-
Claude Code style interfaceThe pixelated whale-shaped top bar, combined with the dual-flowing large-font title and real-time status bar, makes it look great on the device.
-
Streaming terminal interactionSupports Markdown rendering and tab completion.
@File referencing, history search, and dual rendering modes. -
Agent status is visible in real timeThe bottom status bar synchronously displays TPS, cache hit rate, context progress, and inference level.
-
Complete Session WorkflowSupports creating, restoring, compressing, exporting, and querying; double-clicking...
EscYou can then roll back or fork the session. -
Full DSH capability integrationSeamlessly invoke official Agent Preset, Skills, MCP, Goals, Todos, and sub-agents.
-
One-click environment diagnosis:
/doctorSelf-check operating environment/costCheck the dosage./mcpThe connection supports custom API endpoints. -
Long session performance optimization: Prevent the terminal from becoming sluggish due to the session by using event-driven projection, differential output and message virtualization.
How to use dsh-TUI
- Environmental preparationEnsure that Node.js (recommended ≥24) and pnpm (≥10) are installed.
- Install pluginsGlobally install the official DSH and dsh-TUI:
npm install -g @deepseek-ai/dsh @deepseek-harness-tui/dsh-tui。 - Configuration keyExport in terminal
DEEPSEEK_API_KEYEnvironment variables (optional)DEEPSEEK_BASE_URL)。 - Startup screenEnter the project directory and execute.
dsh-tuiLaunch the terminal interface. - Environmental self-inspection:enter
/doctorCheck if the environment, credentials, and session status are normal. - Send taskType the first task instruction in the input box and press Enter to send.
- Common commands:use
/helpView commands,/modelSwitch models,/compactCompress the context. - shortcut key operation:according to
Ctrl+CInterrupt turn, double-clickEscRollback sessionCtrl+OExpand on the details of your thinking. - Non-interactive modeUse in script scenarios
dsh --profile headless "指令"Perform a one-time, non-interactive task.
The core advantages of dsh-TUI
- Claude Code style interfaceThe pixelated whale-shaped top bar, dual-flowing large-font title, and real-time status bar make the terminal interface both beautiful and practical.
- Zero-intrusion designPure plugin mounting, zero core modifications; it is activated upon installation and leaves no trace after uninstallation.
- Officially recognized inclusionSelected as a "Beta User Choice Plugin" by the official DeepSeek Harness WeChat account, its quality and compatibility are guaranteed.
- Full DSH capability integrationIt fully supports all official core capabilities, including Agent Preset, Skills, MCP, Goals, Todos, and sub-agents.
- Real-time status visualizationThe bottom status bar synchronously displays TPS, cache hit rate, context progress, inference level, and Git branch information, making the Agent status clear at a glance.
- Long session performance optimizationIt employs event-driven projection, differential terminal output, message virtualization, and bounded caching technologies, ensuring that sessions will not experience lag or memory bloat, no matter how long they are.
- Powerful session management:double click
EscIt can rewind/fork (time rewind), and also supports advanced operations such as compressing context, restoring history, exporting sessions, and querying. - Native terminal interactionSupports streaming Markdown rendering, structured tool cards, and tab auto-completion.
@File referencing, history search, and dual rendering modes.
dsh-TUI project address
- Project official website:https://dshtui.com/
- GitHub repository:https://github.com/ccch1mneyyy/dsh-TUI
Comparison of dsh-TUI with similar competing products
| Dimension | dsh-TUI | Claude Code |
|---|---|---|
| Essential positioning | DSH ecosystemInterface Plugin Layer | independentAI Programming Agent |
| Model support | Supports multiple models (DeepSeek/Claude/OpenAI, etc.) through DSH. | Only Claude models are supported. |
| Autonomy | Agent capabilities that rely on the DSH kernel | Native support for self-planning, test-and-repair loops, and sub-proxies |
| cost | Free and open source (pay for API) | Subscription fee ($20~$200/month) |
| Git Workflow | Managed through DSH Agent | Native auto-commit/branch management |
Application scenarios of dsh-TUI
- Native terminal developmentIt can be run directly in iTerm2, the built-in terminal of VS Code, or tmux, without leaving the command-line environment, making it suitable for developers who are used to keyboard-driven development.
- SSH remote server operation: Starts after connecting to a remote server via SSH, providing full DSH Agent capabilities in server environments where a desktop GUI cannot be run.
- Long session code review and refactoring:use
/compactCompress context, double-clickEscrewind/fork session backtracking, combined with differential output and message virtualization, is used to handle large code tasks that require multiple iterations. - Real-time performance monitoring and debuggingThe bottom status bar displays TPS, cache hit rate, context segmentation progress, and token consumption in real time, making it suitable for debugging scenarios that require precise observation of the Agent's running status.
- Full Keyboard Efficient WorkflowReject mouse operation; use Tab for auto-completion.
@File references,Ctrl+OExpand your thinking, switch models using shortcut keys, and achieve highly efficient interaction driven entirely by the keyboard.