Playground

Test DAG workflows instantly - no signup required

Workflow Definition

{
  "steps": [
    {
      "name": "research",
      "type": "parallel",
      "config": {},
      "steps": [
        {
          "name": "benefits",
          "type": "llm",
          "config": {
            "model": "gpt-3.5-turbo",
            "prompt": "List 3 key benefits of {{topic}}. Be concise."
          }
        },
        {
          "name": "challenges",
          "type": "llm",
          "config": {
            "model": "gpt-3.5-turbo",
            "prompt": "List 3 main challenges with {{topic}}. Be concise."
          }
        },
        {
          "name": "trends",
          "type": "llm",
          "config": {
            "model": "gpt-3.5-turbo",
            "prompt": "List 3 current trends in {{topic}}. Be concise."
          }
        }
      ]
    },
    {
      "name": "synthesize",
      "type": "llm",
      "config": {
        "model": "gpt-3.5-turbo",
        "prompt": "Based on this research, write a brief executive summary (2-3 sentences):\n\nBenefits: {{research.benefits}}\nChallenges: {{research.challenges}}\nTrends: {{research.trends}}"
      }
    }
  ]
}

Input Data

Select an example and click Execute

Results will appear here

What you're seeing

DAG Execution

Workflows execute as directed acyclic graphs - parallel when possible, sequential when needed

Data Flow

Each step can reference outputs from previous steps using {{step.output}} syntax

Rust Runtime

Built on Tokio for true parallel execution without Python's GIL limitations