Prompt AI

Sends prompts to AI for content generation and processing.

Overview

The Prompt AI node uses AI to generate or transform content based on your prompts. Use it to:

• Generate text content
• Summarize data
• Extract information
• Transform content
• Answer questions about data

Parameters

Parameter	Description	Required
Prompt	Instructions for the AI	Yes
Response Format	Text, JSON, or JSON Schema (default: Text)	No
JSON Schema	Schema constraining the response (only shown for JSON Schema)	No
File	Optional file to include in the prompt context (max 5 pages)	No

Prompt

Write instructions telling the AI what to do. Use expressions to include item data:

Summarize this customer feedback: {{$item.data.feedback}}

Generate a product description for:
Name: {{$item.data.productName}}
Features: {{$item.data.features}}
Target audience: {{$item.data.audience}}

Response Format

• Text — the response is returned as a plain string.
• JSON — the AI must return valid JSON, parsed into an object you can reference downstream (e.g. {{$item.data.summary}}).
• JSON Schema — the AI must return JSON matching the schema you provide.

JSON Schema

When Response Format is set to JSON Schema, supply a JSON Schema object that describes the exact shape of the response. The AI is constrained to produce JSON matching the schema, which makes downstream references reliable.

Supported keywords (a subset of the OpenAPI schema dialect):

• type — "object", "array", "string", "number", "integer", "boolean"
• properties — fields of an object, each itself a schema
• required — array of property names that must appear
• items — schema for elements of an array
• enum — restrict a value to a fixed set
• description — natural-language hint the AI uses to fill the field

Example — extract a person:

{
  "type": "object",
  "properties": {
    "name": { "type": "string", "description": "Full legal name" },
    "age": { "type": "integer" },
    "tags": {
      "type": "array",
      "items": { "type": "string" }
    },
    "status": { "type": "string", "enum": ["active", "inactive"] }
  },
  "required": ["name", "status"]
}

Example — extract a list of line items from an invoice:

{
  "type": "object",
  "properties": {
    "lineItems": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "description": { "type": "string" },
          "quantity": { "type": "integer" },
          "unitPrice": { "type": "number" }
        },
        "required": ["description", "quantity", "unitPrice"]
      }
    }
  },
  "required": ["lineItems"]
}

Leave the schema blank to fall back to free-form JSON behaviour.

File

Attach a PDF, image, or text file to the prompt so the AI can reference its contents. The file is capped at 5 pages; larger files fail the node. Use the Data Extractor node for longer documents.

Output

The AI response is returned in the data field:

• Text — data is the response string.
• JSON / JSON Schema — data is the parsed object, so downstream nodes can reference fields directly (e.g. {{$item.data.name}}).

{
  "data": "Generated content from AI..."
}

Examples

Content Summarization

Summarize long text:

Prompt:

Summarize the following article in 2-3 sentences:

{{$item.data.articleText}}

Data Extraction

Extract specific information:

Prompt:

Extract the following from this email:
- Sender name
- Main request
- Urgency level (high/medium/low)

Email:
{{$item.data.emailBody}}

Return as JSON.

Content Generation

Generate marketing copy:

Prompt:

Write a compelling product description for:

Product: {{$item.data.name}}
Category: {{$item.data.category}}
Key Features: {{$item.data.features}}
Price: {{$item.data.price}}

Keep it under 100 words and highlight the main benefits.

Classification

Categorize items:

Prompt:

Classify this customer inquiry into one of these categories:
- Billing
- Technical Support
- Sales
- General

Inquiry: {{$item.data.message}}

Return only the category name.

Translation

Translate content:

Prompt:

Translate the following text to Spanish:

{{$item.data.text}}

Data Cleaning

Standardize data:

Prompt:

Standardize this company name to its official form.
Remove suffixes like Inc., LLC, Corp.
Fix common misspellings.

Input: {{$item.data.companyName}}

Return only the standardized name.

Prompt Tips

Be Specific

Give clear, detailed instructions:

// Good
Summarize in exactly 3 bullet points, focusing on key findings.

// Vague
Summarize this.

Provide Context

Include relevant background:

You are analyzing customer support tickets for an e-commerce company.
Classify the following ticket...

Specify Output Format

Request structured output:

Return the result as JSON with these fields:
- category: string
- confidence: number (0-1)
- reasoning: string

Use Examples

Show what you want:

Extract the date from the text. Examples:
"Meeting on Jan 5th" → "2024-01-05"
"Due by next Friday" → "[next Friday's date]"

Text: {{$item.data.text}}

Tips

• Keep prompts focused and specific
• Use expressions to inject item data
• Request structured output (JSON) for easier downstream processing
• Test prompts with sample data before processing large batches
• Consider rate limits when processing many items
• Use Edit Item to extract fields from AI responses