For AI agents: the complete documentation index is available at https://docs.flezise.com/llms.txt, the full documentation bundle is available at https://docs.flezise.com/llms-full.txt, and this page is available as Markdown at https://docs.flezise.com/work-items/pipeline-editor.md.

Pipeline editor

The pipeline editor is a DAG canvas at /projects/<project>/dag-editor where you define which artifact types generate and in what order. Each node on the canvas is an artifact type; each edge declares a dependency. The editor is not in the project sidebar. Navigate there directly from the URL or follow the link inside the workflow-design artifact.

The DAG editor: a canvas of artifact nodes connected by Requires edges, an artifact-type palette on the left, the toolbar at the top showing Auto Layout and Save, and a Node Config panel open on the right

How generation works

A node dispatches (starts a generation run) once every artifact listed under its requires edges has reached approved or locked. The run produces a draft; submitting that draft for review and getting it approved then unlocks the nodes that depend on it.

Some artifact types are configured as auto_approve: the generated artifact skips review and lands directly at approved, immediately unblocking any nodes that depend on it.

Artifact types with cardinality: 1:N fan out into a batch of sibling artifacts. Each sibling does not wait for a human to approve its upstream sibling; it dispatches as soon as the upstream sibling is settled, meaning in_review, approved, or locked. This lets a set of related artifacts generate in parallel.

The entire downstream pipeline is gated on the workflow-design artifact. Nothing generates until the workflow-design is locked.

Toolbar

ControlShown whenWhat it does
Back to ProjectAlwaysLeaves the editor. Opened from the workflow-design artifact, the link reads Back to Workflow Design
Requires / BenefitsAlwaysChooses which kind of edge you draw next: a hard dependency, or an optional one
Auto LayoutAlwaysRearranges the nodes automatically
Apply BlueprintAlwaysReplaces the canvas from a template
DiscardCanvas is dirtyReverts all unsaved changes
Save / Saving...Canvas is dirtyPersists the DAG to the workflow-design artifact

Nothing on the canvas is persisted until you press Save. Validation errors surface at that point.

Canvas and palette

The palette on the left groups artifact types by phase: Analysis, Planning, Solutioning, Implementation, and Custom Node. Drag a type onto the canvas. An empty canvas displays the prompt: Drag artifact types from the palette / or use Apply Blueprint to start from a template.

Draw an edge between two nodes. With Requires selected in the toolbar the edge is a hard dependency; with Benefits it is an optional benefits_from dependency that does not block dispatch.

Apply Blueprint

Apply Blueprint opens a picker where you choose a template. Before replacing the canvas the modal warns: This will replace the current DAG. Unsaved changes will be lost. Save first if you want to keep what is there.

Node Config

Select any node to open the Node Config panel.

The Node Config panel: Label, Type (slug), and Phase fields at the top; read-only Scope and Spawns scope below; and a Requires section with an Add requires dependency button
FieldEditableNotes
LabelYesDisplay name shown on the canvas node
Type (slug)YesArtifact type identifier
PhaseYesLifecycle phase the artifact belongs to
ScopeNoProject, Epic, Story, or Change (from the artifact type)
Spawns scopeNoScope created when this node dispatches
RequiresYesHard dependencies; shows No required dependencies until you add one with Add requires dependency
Optional dependenciesYesbenefits_from edges; shows No optional dependencies when empty

Scope and Spawns scope are read-only.

Next steps