Creating Workflows

Thallus provides two ways to create workflows: describe what you want in natural language and let AI generate the workflow, or build one manually on the visual DAG canvas. Most users start with generation and refine with the editor.

Two creation paths

✨
AI Generation
Describe your workflow in plain language and let AI build the DAG
✎
Manual Editor
Drag and drop nodes on the visual canvas and configure each one

Natural language generation

The fastest way to create a workflow is to describe it in plain language. The generation flow guides you through four steps:

Describe
→
Select trigger
→
Generate
→
Review & edit
→
Save

Step 1: Describe your workflow

Enter a natural language description of what the workflow should do. Be specific about:

  • What data to gather or analyze
  • What conditions should trigger branching
  • Who should receive the results and how
  • Any approval requirements

For example: "Every weekday at 9 AM, research market trends for our top 3 competitors, check if any significant changes are found, and if so send a summary to the #market-intel Slack channel and email the strategy team."

Step 2: Select trigger type

Choose how the workflow starts:

✨ Auto-detect
AI decides the best trigger type
▶ Manual
Click to run
🕓 Schedule
Run on a cron schedule
⚡ Webhook
Triggered by external events

If you select Schedule, the cron builder appears where you set frequency, day, and time. If you select Webhook, you choose from available webhook source templates or create a custom source. Auto-detect lets the AI infer the right trigger from your description.

Step 3: Generate

The AI analyzes your description and produces a complete DAG with nodes, edges, trigger configuration, and delivery settings. Generation also validates that required resources (agents, integrations, data connections) are available and warns you about any missing dependencies.

Step 4: Review and edit

The generated workflow opens in the DAG canvas editor where you can:

  • Rearrange nodes by dragging them
  • Edit any node's configuration in the side panel
  • Add, remove, or reconnect edges
  • Add new nodes from the toolbar
  • Adjust trigger and delivery settings

DAG canvas editor

The visual editor displays the workflow as an interactive graph. Nodes appear as cards positioned on a 2D canvas, connected by directional edges.

Toolbar actions:

  • Add any of the nine node types by clicking its icon
  • Connect nodes by dragging from one node's output handle to another node's input handle
  • Delete nodes or edges by selecting them and pressing delete
  • Use the minimap for navigation in large workflows

Node editing:

Click any node to open the side panel with its configuration. The panel shows general settings (title, description) at the top, followed by type-specific configuration below.

Agent auto-suggest

When editing an action node, the system analyzes the instruction text and suggests agents that are likely relevant. Suggested agents appear as selectable chips. You can accept suggestions, add others manually, or exclude specific agents from being used.

Workflow templates

The template gallery provides pre-built workflows organized by category. Templates cover common use cases like daily reports, event monitoring, approval flows, and data pipeline automation.

Each template includes a preview showing the DAG structure, a description of what it does, and the resources it requires. Selecting a template creates a new workflow pre-populated with the template's nodes, edges, and configuration — ready to customize and activate.

Saving workflows

When you save a workflow, you choose between two options:

  • Save as Draft — saves the workflow without activating it. Use this when you want to continue editing later or test before going live
  • Save & Activate — saves and immediately sets the workflow to Active status, enabling any configured triggers

Every save creates a new version so you can always revert to a previous state.