Skip to main content

​
POST /functions/v1/generate-text

Generates text using AI models (OpenAI GPT or Anthropic Claude) based on a prompt. Supports placeholder substitution for dynamic prompts.

​
Request

​
Headers

​
Authorization
string
required
Bearer token for authentication: Bearer <your-supabase-jwt-token>
​
Content-Type
string
required
Must be application/json

​
Body Parameters

​
prompt
string
required
The prompt to send to the AI model. Supports placeholder syntax like {{input.text}} which will be resolved from previousOutput.
​
elementId
string
The workflow element ID. Required if API key and model are not provided directly. Used to load configuration from the database.
​
apiKey
string
API key for OpenAI or Anthropic. If not provided, will be loaded from agent configuration based on elementId.
​
model
string
Model identifier (e.g., gpt-3.5-turbo, gpt-4, claude-3-opus-20240229). Defaults to gpt-3.5-turbo if not specified.
​
previousOutput
object
Object containing outputs from previous workflow steps, used for placeholder resolution.
​
testMode
boolean
If true, uses provided credentials instead of loading from database. Defaults to false.

​
Response

​
success
boolean
Indicates whether text generation was successful
​
text
string
The generated text content from the AI model

​
Model Selection

The function automatically detects the provider based on the model name:
  • OpenAI: Models containing gpt (e.g., gpt-3.5-turbo, gpt-4, gpt-4-turbo)
  • Anthropic: Models containing claude (e.g., claude-3-opus-20240229, claude-3-sonnet-20240229)

​
OpenAI Configuration

  • Endpoint: https://api.openai.com/v1/chat/completions
  • Max tokens: 1000
  • Message role: user

​
Anthropic Configuration

  • Endpoint: https://api.anthropic.com/v1/messages
  • Max tokens: 1000
  • API version: 2023-06-01

​
Placeholder Resolution

Prompts can include placeholders that reference previous workflow outputs: 
Summarize this email: {{input.emailBody}}
Placeholders are resolved using the previousOutput parameter, which typically contains:
  • input.text - Text from previous text generator
  • input.emailBody - Email body from Gmail reader
  • input.messages - Array of messages
  • Custom paths like input.messages.0.subject

​
Examples

​
Request with Direct Credentials

curl -X POST https://your-project.supabase.co/functions/v1/generate-text \
  -H "Authorization: Bearer YOUR_JWT_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Write a short poem about coding",
    "apiKey": "sk-...",
    "model": "gpt-3.5-turbo",
    "testMode": true
  }'

​
Request with Element Configuration

curl -X POST https://your-project.supabase.co/functions/v1/generate-text \
  -H "Authorization: Bearer YOUR_JWT_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "elementId": "text-gen-123",
    "prompt": "Summarize this text: {{input.text}}",
    "previousOutput": {
      "input": {
        "text": "Long article text here..."
      }
    }
  }'

​
Success Response

{
  "success": true,
  "text": "In the realm of code and screen,\nWhere logic reigns supreme and keen,\nWe craft with care each function's flow,\nAnd watch our programs grow and glow."
}

​
Error Response

{
  "error": "Failed to generate text: Invalid API key"
}

​
Error Codes

Status CodeError MessageDescription
400Prompt is requiredMissing prompt parameter
400API key is requiredNo API key provided or found in config
404No text generator configuration foundElement ID has no associated configuration
500Failed to retrieve agent configurationDatabase error loading configuration
500Failed to generate textAI API returned an error

​
Configuration Storage

When using elementId, credentials are:
  1. Retrieved from agent_configs table
  2. Filtered by user_id, element_id, and agent_type: 'text_generator'
  3. Decrypted using XOR encryption with ENCRYPTION_KEY environment variable
  4. Used for API calls to OpenAI or Anthropic

​
Notes

  • API keys are encrypted in the database using XOR encryption
  • Maximum token limit is set to 1000 for all requests
  • The function automatically determines the provider from the model name
  • Placeholder resolution happens before sending to the AI API
  • Both apiKey and api_key field names are supported in configurations

Build docs developers (and LLMs) love

Get started for freeTalk to us