Workflows

💡TL;DR

Notemd workflows chain multiple tasks into a single one-click action. Define sequences like add-links > extract-concepts > research > diagram using a simple DSL. Workflows appear as sidebar buttons that run the full chain on the current note or folder. Ships with predefined workflows; create custom ones in settings. Each step uses its own per-task model configuration.

This is part of the Obsidian AI Knowledge Management Guide.

Overview

A workflow eliminates the friction of running tasks one at a time. Instead of right-clicking four times to add links, extract concepts, research unfamiliar terms, and generate a diagram, you press one sidebar button and the entire chain executes. Notemd handles the sequencing, error propagation, and progress reporting.

Workflows are defined in a lightweight DSL (domain-specific language). They live in settings, appear as clickable buttons in the Obsidian sidebar, and can be applied to either the current note or an entire folder.

How It Works

Workflow Execution Pipeline

1. Parse -- The DSL string is split on > (or >) into an ordered list of task identifiers.
2. Resolve -- Each identifier maps to an internal command (add-links, extract-concepts, research, translate, diagram, etc.).
3. Execute -- Steps run sequentially. Each step uses its configured per-task provider and model.
4. Error handling -- If a step fails, the workflow either aborts or continues to the next step, depending on your error policy.
5. Done -- A toast notification reports success or lists any failed steps.

DSL Format

Workflows are defined as a >-separated sequence of task identifiers:

process-current-add-links>extract-concepts-current>research-and-summarize

Available task identifiers:

Identifier	Action
process-current-add-links	Add wiki-links to the active note
extract-concepts-current	Extract concepts from the active note
research-and-summarize	Research the selected text or note title
process-current-translate	Translate the active note
summarize-to-mermaid	Generate a diagram from the active note
generate-from-title	Generate content from the note title
extract-original-text	Extract original text (for OCR / scanned content)

Folder-level variants replace current with folder in the identifier name.

Predefined vs. Custom Workflows

Notemd ships with ready-made workflows for common patterns:

Workflow	Chain	Use Case
One-Click Extract	add-links > extract-concepts > research	Process a research paper in one pass
Full Pipeline	add-links > extract-concepts > research > diagram	Complete knowledge extraction with visualization
Translate + Link	translate > add-links	Translate then link concepts in the target language

Custom workflows are created in settings:

1. Open Settings --> Notemd --> Workflows
2. Click "Add Workflow"
3. Enter the DSL chain (e.g., process-current-add-links>extract-concepts-current)
4. Give it a display name (e.g., "Quick Link + Extract")
5. The new button appears in the sidebar immediately

Configuration

Setting	Default	Effect
workflows	Predefined set	Array of workflow definitions (name + DSL)
workflowContinueOnError	true	Continue to next step if the current step fails
workflowShowProgress	true	Show a progress toast after each step completes

Per-Task Models in Workflows

Each step in a workflow uses its own per-task model configuration. You do not need to specify models in the DSL itself. The resolution order is:

1. Per-task provider/model if useMultiModelSettings is on
2. Global activeProvider otherwise

This means add-links can run on DeepSeek while research runs on GPT-4o -- all within the same workflow click.

Example

You just imported a PDF of a machine learning paper into your vault and want full knowledge extraction:

1. Open the imported note
2. Click the "Full Pipeline" sidebar button
3. Notemd executes:
   • Step 1: Add wiki-links -- [[attention mechanism]], [[transformer]], etc.
   • Step 2: Extract concepts -- creates concept notes in your concept folder
   • Step 3: Research -- summarizes web sources for key terms
   • Step 4: Diagram -- generates a Mermaid mindmap of the paper's structure
4. After ~30 seconds, your note has links, concept notes exist, research is appended, and a diagram file is saved

All from a single click.

Tips

• Start with predefined workflows -- they cover the most common patterns. Customize only when you need a different sequence.
• Enable workflowContinueOnError -- a failed diagram step should not abort the entire pipeline.
• Use folder workflows for bulk processing -- right-click a folder, pick a workflow, and every note gets processed.
• Name workflows clearly -- sidebar space is limited. Use short, action-oriented names like "Quick Extract" or "Translate + Link".

---

Next Steps

• Research -- Understand what the research step does before adding it to workflows
• Wiki-Links -- Core linking feature used in most workflows
• Concept Notes -- Concept extraction as a workflow step
• Batch Processing -- Concurrency and progress reporting for folder workflows