openma-ai/deepseek-harness-tui
A Rust/ratatui terminal client that speaks the DSH SDK JSON-RPC protocol directly and runs standalone or as a profile bundle.
Install
via GitHub · GitHub source
Installing from GitHub runs the project's build script, which pnpm blocks until you allowlist it — run the command once and pnpm prints the exact key to add under `allowBuilds` in ~/.dsh/profiles/web/pnpm-workspace.yaml.
What happened when we ran it
Installed cleanly when we ran it
Every command here is run in a throwaway container against a clean profile, and the result is whatever the harness recorded — not a guess from the source. Last run 3d ago.
Due diligence
What it is — It is a Rust-based terminal client that communicates directly with the DeepSeek Harness SDK JSON-RPC protocol and can run standalone or as a profile bundle.
Who it is for — Suitable for users who need to monitor streaming inference, tool calls, and agent timelines in the terminal while using DeepSeek Harness for multi-turn conversations.
It is not suitable for users who prefer browser or GUI interfaces because it focuses on terminal-native interface and keyboard shortcuts.
Watch out — Installation requires compiling from GitHub source code, needing Rust stable and Node.js 18+. Sandbox testing passed with successful registration in a fresh profile, though it can execute shell commands. No other obvious issues found.
The verdict — I would install it because it offers complete agent timelines and tool call visibility for monitoring long sessions, but only if you have DeepSeek Harness already installed and use a terminal that supports kitty graphics protocol.
Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.
deepseek-harness-tui is a DeepSeek Harness plugin that provides a Rust/ratatui terminal client for streaming agent conversations, tool calls, Skills, image prompts and persistent sessions. DeepSeek Harness can run it as a profile surface, or dsh-tui can attach to an ACP agent in standalone mode.
The plugin uses a separate Node Cordis client tree and Rust painter process. The client communicates with the Host through standard stdin/stdout using ACP, while the painter owns the TTY, handles input and renders declarative state. It reads the agent’s advertised models, composition, permissions, authentication methods and commands. Conversation state includes streaming updates, tool results, subagent lifecycles and token/cache metrics; persistent sessions use JSONL storage and are managed with /new, /resume and --session-id.
deepseek-harness-tui is intended for developers who want a terminal-native DeepSeek Harness surface or a client for another ACP agent. It is a poor fit for environments without an interactive terminal, or where a separate terminal client is unnecessary. Node.js 18+ is required for the packaged profile path. The terminal-surface risk is direct: the Rust process takes control of the TTY and processes keyboard and mouse input, clipboard operations and terminal rendering protocols. As a client, it can display and submit prompts to the connected agent, so the agent retains its own configured tools and permissions.
DeepSeek Harness can load deepseek-harness-tui in a tui profile. The profile path attaches the ACP plugin to the Base Cordis Host tree and starts a separate TUI Client process. The TUI bundle treats ACP as a runtime dependency and uses the ACP version resolved from its own dependency graph. A profile that already contains another ACP bundle may retain both packages, but the TUI surface disables the older transport/provider entries.
The standalone client accepts an agent command and repeated argument pairs:
dsh-tui --agent dsh-acpdsh-tui --agent dsh --agent-arg --profile --agent-arg acpdsh-tui is the main command; dsb remains a compatibility alias. --demo starts a gallery without a runtime or API key. --demo-skin mounts the ember gallery package. A locally built binary can be selected with DSH_TUI_BIN; the README places packaged binaries under npm/vendor/<platform>/.
The client provides /new, /resume and /theme. --session-id selects a persistent session. During a turn, follow-up messages can be queued or the current turn can be steered immediately. Skills and built-in commands appear in a searchable, scrollable slash menu.
DeepSeek Harness requires Node.js 18+ for the packaged profile workflow. The client is designed for an interactive terminal and supports native terminals and tmux. The Rust painter owns the TTY and processes keyboard input, mouse interaction, clipboard integration, and supported terminal rendering protocols, including kitty graphics where available.
DeepSeek Harness carries session, prompt, authentication, configuration and session/update traffic over standard ACP. The client displays agent advertisements, Markdown, tool states, subagent lifecycle events and token/cache indicators. Persistent sessions use JSONL; workspace-mode information is cached. Up to eight images can be staged from files, the clipboard or paste operations. The client does not spawn a second ACP agent in profile mode; standalone mode can spawn or attach to an ACP agent.
Written from the project's own documentation and kept in sync with it. Where the two disagree, the source is authoritative — read the README on GitHub
Same category
zhu1090093659
Plugin and skin collection for the DSH Web UI: task board, Git graph, right-side panel, remote mobile UI, pet, live token stats, and a skin center.
No one-line install — this plugin lives inside a larger repository and publishes no npm package.
omdsh-dev
Full sidebar workbench with file rendering and editing, terminal, Git, and subagents; third-party plugins can register new tabs.
Installed cleanly when we ran it
ccch1mneyyy
Claude Code-style full-screen terminal UI: pixel-whale header, live status line, and streaming thought expansion.
Installed cleanly when we ran it
omdsh-dev
Codex-style `@file` mentions: search workspace files in the composer and attach their contents to prompts.
Installed cleanly when we ran it
omdsh-dev
Interactive UI components rendered inline in replies: layout, charts, forms, quizzes, mermaid, 3D scenes, and an action event loop back to the model.
Installed cleanly when we ran it
huiliyi37
A terminal UI (TUI) for DeepSeek Harness.
Installed cleanly when we ran it