Skip to main content

​
Overview

The Generate Workflow endpoint uses GPT-4o to automatically create complete workflows from natural language descriptions. This is one of Agility’s most powerful features, allowing users to describe their automation needs in plain English and receive a fully structured workflow.

​
Endpoint

POST https://<your-project>.supabase.co/functions/v1/generate-workflow

​
Authentication

Requires a valid Supabase JWT token in the Authorization header: 
Authorization: Bearer <your-jwt-token>

​
Request

​
Request Body

​
prompt
string
required
Natural language description of the desired workflow. Describe what you want to automate, including the services and actions involved.Examples:
  • “Read emails from Gmail, summarize them with AI, and send summaries to Discord”
  • “Monitor GitHub commits and generate release notes using GPT-4”
  • “Automatically respond to support emails with AI-generated responses”

​
Example Request

curl -X POST https://your-project.supabase.co/functions/v1/generate-workflow \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_JWT_TOKEN" \
  -d '{
    "prompt": "Read emails from my Gmail, generate a summary using GPT-4, and send the summary to Discord"
  }'

​
Response

​
Success Response (200)

​
workflow
object
required
The generated workflow structure

​
Example Response

{
  "workflow": {
    "elements": [
      {
        "id": "element-a8f3d2c-0",
        "type": "agent",
        "agentId": "3",
        "position": { "x": 200, "y": 300 },
        "data": {
          "name": "Gmail Reader",
          "description": "Reads emails from specific senders in Gmail",
          "color": "#ecfdf5"
        }
      },
      {
        "id": "element-b7e4c1d-1",
        "type": "agent",
        "agentId": "1",
        "position": { "x": 450, "y": 300 },
        "data": {
          "name": "Text Generator",
          "description": "Generates text based on prompts",
          "color": "#f0f9ff"
        }
      },
      {
        "id": "element-c6d5b2e-2",
        "type": "agent",
        "agentId": "4",
        "position": { "x": 700, "y": 300 },
        "data": {
          "name": "Discord Messenger",
          "description": "Sends messages to Discord servers via webhooks",
          "color": "#eef2ff"
        }
      }
    ],
    "connections": [
      {
        "id": "connection-x9y2z1-0",
        "sourceId": "element-a8f3d2c-0",
        "targetId": "element-b7e4c1d-1",
        "type": "default"
      },
      {
        "id": "connection-w8v3u2-1",
        "sourceId": "element-b7e4c1d-1",
        "targetId": "element-c6d5b2e-2",
        "type": "default"
      }
    ]
  }
}

​
Error Responses

​
400 Bad Request

Returned when the prompt is missing or invalid: 
{
  "error": "Prompt is required"
}

​
401 Unauthorized

Returned when authentication fails: 
{
  "error": "Not authenticated"
}

​
500 Internal Server Error

Returned when workflow generation fails: 
{
  "error": "Failed to generate workflow with OpenAI"
}

​
How It Works

  1. AI Processing: Your prompt is sent to GPT-4o with instructions about available agents
  2. Agent Matching: The AI selects appropriate agents from the 5 available types
  3. Linear Flow: Agents are arranged in a left-to-right linear sequence
  4. Auto-Positioning: Elements are automatically positioned on the canvas with proper spacing
  5. Connection Creation: Connections are created between adjacent agents

​
Available Agents

The AI can only use these 5 agents when generating workflows:
  • Text Generator - AI-powered text generation
  • Gmail Sender - Send emails via Gmail
  • Gmail Reader - Read emails from Gmail
  • Discord Messenger - Send Discord messages
  • GitHub Reader - Monitor GitHub repositories

​
Workflow Constraints

Generated workflows are linear (start to finish, left to right) and use 3-7 agents maximum for optimal usability.

​
Configuration Required

After generating a workflow:
  1. Each agent must be individually configured with credentials
  2. Prompts and parameters need to be customized
  3. Connections can be adjusted if needed
  4. Test each agent before running the full workflow

​
Environment Requirements

The Supabase project must have OPENAI_API_KEY environment variable configured in Edge Function secrets for this endpoint to work.

​
Usage in Agility UI

This endpoint is called when users:
  • Type a workflow description in the AI generation input
  • Click the “Generate Workflow” button
  • The generated workflow appears on the canvas ready for configuration

​
Best Practices

  1. Be Specific: Include service names (Gmail, Discord, GitHub) in your prompt
  2. Describe Actions: Use verbs like “read”, “send”, “generate”, “monitor”
  3. Linear Flow: Describe workflows as sequential steps
  4. Keep it Simple: Focus on 3-5 main steps for best results

​
Example Prompts

“Read support emails from Gmail, generate AI responses using GPT-4, and send replies back via Gmail”
“Monitor GitHub commits, summarize changes with AI, and post updates to Discord”
“Generate blog content using AI and send it to my email for review”

Build docs developers (and LLMs) love

Get started for freeTalk to us