AI Workflows

Build multi-step AI pipelines with sequential and parallel execution, variable passing between steps, and webhook notifications. Supports Anthropic, OpenAI, Gemini, Replicate, and more.

Create a Workflow

Define a multi-step AI pipeline with dependencies:

create-workflow.ts

import { Stack0 } from '@stack0/sdk'
 
const stack0 = new Stack0({
  apiKey: process.env.STACK0_API_KEY!
})
 
const workflow = await stack0.workflows.create({
  projectSlug: 'my-project',
  slug: 'content-pipeline',
  name: 'Content Generation Pipeline',
  description: 'Generate blog posts with AI',
  steps: [
    {
      id: 'generate-outline',
      type: 'llm',
      name: 'Generate Outline',
      provider: 'anthropic',
      model: 'claude-3-5-sonnet-20241022',
      prompt: 'Create an outline for a blog post about {{topic}}',
      outputVariable: 'outline',
    },
    {
      id: 'write-content',
      type: 'llm',
      name: 'Write Content',
      provider: 'openai',
      model: 'gpt-4o',
      prompt: 'Write a blog post based on this outline:\n\n{{steps.generate-outline.output}}',
      dependsOn: ['generate-outline'],
      outputVariable: 'content',
    },
  ],
  variables: {
    topic: { type: 'string', required: true },
  },
})

Run a Workflow

Execute a workflow with input variables:

run-workflow.ts

// Start workflow execution (async)
const run = await stack0.workflows.run({
  projectSlug: 'my-project',
  workflowId: 'workflow_abc123',
  variables: {
    topic: 'AI in Healthcare',
  },
  webhook: {
    url: 'https://yourapp.com/webhook',
    secret: 'your-webhook-secret',
  },
  metadata: {
    userId: 'user_123',
  },
})
 
console.log('Run ID:', run.id)
console.log('Status:', run.status)

Run and Wait

Execute a workflow and automatically wait for completion:

run-and-wait.ts

const result = await stack0.workflows.runAndWait({
  projectSlug: 'my-project',
  workflowId: 'workflow_abc123',
  variables: {
    topic: 'AI in Healthcare',
  },
}, {
  pollInterval: 2000,
  timeout: 120000,
})
 
console.log('Status:', result.status)
console.log('Output:', result.output)
console.log('Duration:', result.totalDurationMs, 'ms')
console.log('Credits used:', result.creditsUsed)

Step Types

Available step types for building workflows:

Type	Description	Providers
llm	Text generation with AI models	Anthropic, OpenAI, Gemini
image	Image generation	Replicate, Stack0
video	Video generation	Replicate, Kling
audio	Audio generation	ElevenLabs, Replicate
code	Execute custom code	-
http	Make HTTP requests	-
transform	Transform data between steps	-
condition	Conditional branching	-
loop	Iterate over arrays	-

LLM Step Configuration

Configure text generation steps with AI models:

llm-step.ts

{
  id: 'generate-summary',
  type: 'llm',
  name: 'Summarize Content',
  provider: 'anthropic', // 'anthropic' | 'openai' | 'gemini'
  model: 'claude-3-5-sonnet-20241022',
  prompt: 'Summarize this content: {{content}}',
  systemPrompt: 'You are a helpful assistant that creates concise summaries.',
  temperature: 0.7,
  maxTokens: 4096,
  responseFormat: 'text', // 'text' | 'json'
  dependsOn: ['previous-step'],
  outputVariable: 'summary',
}

Image Generation Step

Generate images with AI models:

image-step.ts

{
  id: 'generate-image',
  type: 'image',
  name: 'Generate Hero Image',
  provider: 'replicate',
  model: 'stability-ai/sdxl',
  prompt: 'A futuristic city at sunset, high detail, photorealistic',
  width: 1024,
  height: 1024,
  outputVariable: 'heroImage',
}

HTTP Step

Make HTTP requests to external APIs:

http-step.ts

{
  id: 'fetch-data',
  type: 'http',
  name: 'Fetch External Data',
  method: 'POST',
  url: 'https://api.example.com/data',
  headers: {
    'Authorization': 'Bearer {{apiKey}}',
    'Content-Type': 'application/json',
  },
  body: {
    query: '{{searchQuery}}',
  },
  outputVariable: 'externalData',
}

Variable Interpolation

Use {{variable}} syntax to reference variables and step outputs:

variable-interpolation.ts

// Reference input variables
prompt: 'Write about {{topic}}'
 
// Reference step outputs
prompt: 'Expand on: {{steps.generate-outline.output}}'
 
// Reference nested data
prompt: 'Use this title: {{steps.generate-metadata.output.title}}'
 
// Reference arrays
prompt: 'Process these items: {{steps.fetch-data.output.items}}'

Get Run Status

Check the status of a workflow run:

get-run-status.ts

const run = await stack0.workflows.getRun({
  projectSlug: 'my-project',
  id: 'run_xyz789',
})
 
console.log('Status:', run.status)
console.log('Progress:', `${run.completedSteps}/${run.totalSteps}`)
 
// Check individual step states
for (const [stepId, state] of Object.entries(run.stepStates)) {
  console.log(`${stepId}: ${state.status}`)
  if (state.error) {
    console.log(`  Error: ${state.error}`)
  }
}

List Runs

Query workflow runs with filters:

list-runs.ts

const { items, total, hasMore } = await stack0.workflows.listRuns({
  projectSlug: 'my-project',
  workflowId: 'workflow_abc123', // Optional: filter by workflow
  status: 'completed', // 'pending' | 'running' | 'completed' | 'failed' | 'cancelled'
  limit: 50,
  offset: 0,
})
 
for (const run of items) {
  console.log(`[${run.status}] ${run.id}`)
  console.log(`  Duration: ${run.totalDurationMs}ms`)
  console.log(`  Credits: ${run.creditsUsed}`)
}

Cancel Run

Cancel a running workflow:

cancel-run.ts

await stack0.workflows.cancelRun({
  projectSlug: 'my-project',
  id: 'run_xyz789',
})

Manage Workflows

List, update, and delete workflows:

manage-workflows.ts

// List workflows
const { workflows } = await stack0.workflows.list({
  projectSlug: 'my-project',
  limit: 20,
})
 
// Get a specific workflow
const workflow = await stack0.workflows.get({
  projectSlug: 'my-project',
  id: 'workflow_abc123',
})
 
// Update a workflow
await stack0.workflows.update({
  projectSlug: 'my-project',
  id: 'workflow_abc123',
  name: 'Updated Pipeline Name',
  isActive: true,
})
 
// Delete a workflow
await stack0.workflows.delete({
  projectSlug: 'my-project',
  id: 'workflow_abc123',
})

Webhook Events

Configure webhooks to receive workflow notifications:

webhook-payload.json

// Webhook payload example
{
  "event": "run.completed",
  "timestamp": "2025-01-15T10:30:00Z",
  "data": {
    "id": "run_xyz789",
    "workflowId": "workflow_abc123",
    "status": "completed",
    "output": {
      "outline": "1. Introduction...",
      "content": "AI is transforming..."
    },
    "totalDurationMs": 12500,
    "creditsUsed": 15
  }
}
 
// Available events:
// - run.started
// - run.completed
// - run.failed
// - step.completed

Workflows API Reference AI Extraction Error Handling