dsh-deepseek-flow: A Visual Markdown Workflow Editor for DeepSeek Harness

kanghelyu/dsh-deepseek-flow

Workflow & AutomationVerified
Listed on 2026-08-20
Page last updated 2026-08-20

A visual workflow editor for DeepSeek Harness, Markdown-first with two-way sync between canvas and docs.

Turn DeepSeek Harness's WORKFLOW.md and STEP.md files into an editable diagram with two-way canvas-Markdown sync, making workflow design more intuitive and files always portable.

install
dsh plugin --profile web add github:kanghelyu/dsh-deepseek-flow
Category
Workflow & Automation
Platform
DSH-Plugin
Author
kanghelyu
Distribution
Plugin

dsh-deepseek-flow Key Features

Markdown sourceVisual editingTwo-way syncTheme supportBackground AI jobs

dsh-deepseek-flow Repository

kanghelyu
Publisher
kanghelyu
Repository
kanghelyu/dsh-deepseek-flow
47
Stars
4
Forks
1
Watchers
1
Open issues
LanguageJavaScript
LicenseMIT License
Website
Repo last push2026-08-19 23:49:27
Repo created2026-08-15 02:15:15
Topics
ai-workflowdark-modedeepseek-harnessdeveloper-toolsdsh-pluginflow-editormarkdownvisual-workflowvisualizationworkflowworkflow-builderworkflow-management
Default branchmain

dsh-deepseek-flow Repo Summary

DeepSeek Flow is a visual workflow editor plugin (DSH plugin) for DeepSeek Harness, built for the Web UI, with Markdown as the single source of truth. It solves the readability and portability problems in workflow design and maintenance: it renders WORKFLOW.md and per-step STEP.md files into an editable diagram, keeping the canvas and Markdown in two-way sync while keeping files fully portable. Core capabilities include: a real visual editor (create, move, connect, reconnect, label, and delete nodes and arrows), source-aware topology transactions (human canvas edits receive a full current-Session review, while topology produced by direct Session file edits can use an invisible deterministic finalize path), executable gate semantics (the exported contract includes formulas, operands, predicates, and deterministic Boolean results), and per-session isolation (each session keeps its own workflows, with optional shared templates; deletion is guarded by confirm, and managed workspaces move to a trash area and can be recovered). It also offers comfortable large-flow navigation (collapsible and resizable side panels, pan and zoom, fit-to-view, animated node focus, independent scrolling regions), native theme support (follows Harness light and dark themes and the active WebUI language), manual AI assistance (logic validation, optimize one document with review, or optimize the complete workflow), background AI jobs (switching documents, views, or sessions does not interrupt accepted jobs; document proposals are restored when you return), persistent results and drafts (logic-validation findings, AI proposals, and unapplied canvas drafts are persisted to disk; view switches, session switches, and dsh web restarts lose nothing; cleared only by explicit discard or successful commit), and I/O and memory safeguards (unchanged documents are never rewritten; Markdown edits inside the 650 ms autosave window are flushed immediately when you leave the view). The project is maintained by kanghelyu, licensed under MIT, last updated 2026-08, with the official site at deepseekflow.kanghelyu.org.

What core features does the DeepSeek Flow DSH plugin support?

DeepSeek Flow's core features revolve around Markdown-first visual editing: it turns a WORKFLOW.md and step-level STEP.md files into an editable diagram inside DeepSeek Harness, with two-way synchronization between canvas and Markdown, ensuring files remain portable. It provides a real visual editor—create, move, connect, reconnect, label, and delete nodes and arrows. Source-aware topology transactions give human canvas edits a full current-Session review, while topology produced by direct Session file edits can use an invisible deterministic finalize path. Executable gate semantics export formulas, operands, predicates, and deterministic Boolean results without running Agent steps. It also supports per-session isolation (each session keeps its own workflows, with optional shared templates; deletion is guarded by confirm, and managed workspaces move to a trash area and can be recovered), comfortable large-flow navigation (collapsible and resizable side panels, pan and zoom, fit-to-view, animated node focus, independent scrolling regions), native theme support (follows Harness light and dark themes and the active WebUI language), manual AI assistance (logic validation, optimize one document with review, or optimize the complete workflow), background AI jobs (switching documents, views, or sessions does not interrupt accepted jobs; document proposals are restored when you return), persistent results and drafts (logic-validation findings, AI proposals, and unapplied canvas drafts are persisted to disk; view switches, session switches, and dsh web restarts lose nothing; cleared only by explicit discard or successful commit), and I/O and memory safeguards (unchanged documents are never rewritten; Markdown edits inside the 650 ms autosave window are flushed immediately when you leave the view).

How do I install DeepSeek Flow?

To install DeepSeek Flow as a DSH plugin, run the following command in your Harness environment:

bash
dsh plugin --profile web add github:kanghelyu/dsh-deepseek-flow

After installation, the plugin appears as a visual workflow editor in the Web UI, integrated with Harness sessions.

Which DeepSeek Harness versions and platforms are compatible with DeepSeek Flow?

DeepSeek Flow is designed specifically for the DeepSeek Harness Web UI, running as a web plugin. It depends on multiple official DSH packages (such as @deepseek-ai/dsh-tools, @deepseek-ai/dsh-session, @deepseek-ai/dsh-llm, etc.), ensuring deep integration with the Harness core. The project is written in JavaScript, licensed under MIT, maintained by kanghelyu, and last updated in 2026-08. The official site is deepseekflow.kanghelyu.org, offering more documentation and examples.

How are Markdown workflow files organized in DeepSeek Flow?

DeepSeek Flow uses Markdown as the single source of truth: one master WORKFLOW.md file, plus one STEP.md workspace for each step. Edits made on the canvas and in the Markdown editor are written back to the workflow files, keeping them in two-way sync. This design ensures workflow portability—you can view and modify workflows with any Markdown tool, and DeepSeek Flow will render them as visual diagrams.

How does DeepSeek Flow ensure data safety and performance?

DeepSeek Flow includes multiple safeguards: unchanged documents are never rewritten, avoiding unnecessary I/O; Markdown edits inside the 650 ms autosave window are flushed immediately when you leave the view, preventing loss; background AI jobs are not interrupted when switching documents, views, or sessions, and proposals are restored on return; logic-validation findings, AI proposals, and unapplied canvas drafts are persisted to disk, surviving even dsh web restarts, and are cleared only by explicit discard or successful commit. These designs keep large-flow operations smooth while ensuring data safety.

See more DSH plugins at dsh-plugin.org

DSH-Plugin FAQ