Aisland-SJL/dsh-worktable
🖥️ Agent-project workbench for DeepSeek Harness — sidebar app drawer + dockable split workspace + a live control room watching every project.
What it solves
dsh-worktable is a workbench for DeepSeek Harness (DSH) that transforms the standard chat interface into a multi-window, project-based workspace. It solves the problem of managing multiple AI-driven projects simultaneously by providing a dockable interface where each project can have its own dedicated windows, file explorers, and terminals, rather than relying on a only a single chat stream.
How it works
It functions as a Cordis plugin for DSH, injecting a sidebar app drawer and a split-pane workspace engine into the host shell. The system allows users to create "projects" that are bound to specific AI conversations.
Key mechanisms include:
- Project Binding: Each project is linked to a specific conversation, so switching projects automatically switches the chat pane.
- Custom Windows: Users can request a feature from an agent; the agent writes a
widget-result.jsonfile to the project folder, which the workbench then auto-mounts as a locked window. - Control Room: A built-in project that mirrors the host session runtime snapshots to monitor the status of all bound projects (working, needs attention, or done) in real-time without calling a model.
- Annotation Protocol: A "point-and-speak" feature where users can click an element in a window and type a request, which packages the coordinates and element data into a message for the agent to process.
Who it’s for
Developers and power users of DeepSeek Harness who manage complex, multi-step AI agent projects and need a professional IDE-like environment to monitor and interact with with their agents' outputs.
Highlights
- Dockable Split Workspace: Supports declarative layout presets with draggable dividers and per-pane tabs.
- Integrated Tooling: Built-in panes for file explorers, terminals (PowerShell on Windows), and browsers.
- Real-time Monitoring: A glassmorphism-style "Control Room" for tracking project statuses.
- Visual Annotation: Ability to drop input boxes directly onto UI elements to request changes from the AI agent.
Related
- Project
- Project
- Project
- Project
- Project