Who dsh-TUI Is For
Use this page when a developer has already chosen DeepSeek Harness and wants a full-screen terminal with streaming output, tool cards, session controls, Git context, and status information. A nontechnical founder deciding how to shape a customer workflow does not need this layer.
The repository now describes different extension seams separately: some are stable candidates, some remain experimental, and some follow upstream DSH behavior. That is more useful than applying one blanket stability label to the whole project. Review the current compatibility guide before depending on a specific extension surface.
Install the Named Repository and Package
The current repository documents this global setup:
npm install -g @deepseek-ai/dsh @deepseek-harness-tui/dsh-tui
Start it with either command:
dsh-tui
dst
The project also documents adding the package to a separate dsh-tui profile. A different dsh-tui/dsh-tui repository uses the @dsh-tui/dsh-tui package and different installation rules, so do not mix its commands with this page.
After startup, check that the header identifies the expected session, the working directory matches the test repository, and one harmless prompt produces streaming output. If the command is unavailable, check the package identity and global executable path before changing the DSH profile.
What the Full-Screen Interface Shows
The interface can expose current activity, context usage, token information, cache behavior, throughput, reasoning-effort setting, Git state, and session metadata. These details help a developer understand whether the agent is waiting, using tools, approaching a context limit, or working from the expected repository state.
Streaming Markdown keeps the response readable while it arrives. Tool calls are displayed as structured cards, and the interface can expand the reasoning display and tool details provided by the current model and DSH event stream.
This should not be described as revealing a model's private chain of thought. The /thinking command is a display toggle, and its state is not persistent. Some commands or features also depend on the selected preset, launcher, platform clipboard tools, or an idle session.
For image evidence inside a related technical flow, modlens explains how a screenshot can become structured input.
If the product decision comes before repository work, the Adalo alternative compares a visual mobile editor with a managed web-app workflow.
Test a Separate Profile and Return Safely
dsh-TUI does not patch the DSH core. A separate profile gives the interface its own configuration and makes the return path straightforward: leave the TUI and start the original DSH surface or profile again. The project does not publish a dedicated rollback command, so do not treat session rewind, a double-Escape action, or an update command as an uninstall mechanism.
Permissions come from the DSH profile. The TUI does not add another sandbox. Test the interface with a disposable task before using it for a repository that contains sensitive or production work.
How HappySeeds Fits This Workflow
dsh-TUI is a developer interface; HappySeeds is an app-building product for shaping and launching a customer experience. They do not require a direct integration. Stay in HappySeeds when the goal is to define the customer journey, generate the web app, inspect Preview, and publish. Move to repository tooling only when the team has explicitly chosen a code-owned handoff.
Follow the HappySeeds quick start when the goal is to build and publish without choosing a developer terminal.
FAQ
No. dsh-TUI is for repository teams already using DeepSeek Harness. HappySeeds planning, generation, preview, and publishing do not require a developer terminal.
No. It shows the reasoning display and events made available by the current model and DeepSeek Harness.
Exit dsh-TUI and start the original DSH surface or profile. The project is mounted as a plugin rather than a core patch.
Launch with HappySeeds
Built the app but not the business? Start with HappySeeds to launch a live, monetizable AI app, or browse AI app templates to see what you can sell.