Structured Output

Extract structured data with AI

Overview

Extract structured data from content using OpenAI’s Structured Output API. Define a JSON Schema to specify the exact shape of the response, enabling reliable extraction of arrays, objects, and typed fields from unstructured text.

Category: AI
Tool Name: structured_output
Version: 1

Appearance: Icon: structuredOutput | Color: #8b5cf6

Node Type

Action — processes input items and produces output

Input / Output

Direction	Port(s)
Input	Input
Output	Output, Error

Credentials

This tool requires OpenAI API Credentials credentials. See the Credentials Guide for setup instructions.

Parameters

Parameter	Type	Required	Default	Description
System Prompt	string	No	You are a data extraction assistant. Extract the requested information from the provided content accurately and completely.	System message that instructs the model on extraction behavior. Supports expressions like {{ $json.field }}.
User Prompt	string	Yes	Extract data from the following content:\n\n{{ $json.content }}	The user message containing the content to process. Use expressions like {{ $json.field }} to inject item data.
Schema Name	string	Yes	extracted_data	A name identifier for the JSON Schema (snake_case recommended). Used by OpenAI to reference the schema.
Schema Definition	string	Yes	{"type":"object","properties":{"items":{"type":"array","items":{"type":"string"},"description":"List of extracted items"}},"required":["items"],"additionalProperties":false}	JSON Schema definition specifying the structure of the expected output. Must be valid JSON Schema. Use additionalProperties: false for strict mode.
Model	options	No	gpt-4o-mini	The OpenAI model to use. Must support structured outputs (gpt-4o, gpt-4o-mini, o1, o1-mini, o3-mini).
				Options: gpt-4o, gpt-4o-mini, o1, o1-mini, o3-mini
Strict Mode	boolean	No	true	When enabled, the model’s output will strictly adhere to the schema. Requires additionalProperties: false in schema.
Max Concurrency	number	No	25	Maximum number of concurrent API calls to OpenAI.
Output Field Name	string	No	structuredOutput	The field name to store the structured output result in the output item.
Include Input in Output	boolean	No	true	If true, includes the original input data in the output item alongside the structured result.

Output Data

One output item per input item. The extracted object is written to the field named by Output Field Name (structuredOutput by default), alongside a structuredOutputMetadata block. With Include Input in Output on — the default — the original item fields are kept underneath; turn it off to return only the extraction and its metadata. Binary data is forwarded either way.

{
  "structuredOutput": { "items": ["invoice 4471", "invoice 4472"] },
  "structuredOutputMetadata": {
    "schemaName": "extracted_data",
    "model": "gpt-4o-mini",
    "strictMode": true,
    "usage": { "promptTokens": 743, "completionTokens": 58, "totalTokens": 801 },
    "finishReason": "stop",
    "processedAt": 1765432100000
  }
}

• The value on the output field always matches the shape declared in Schema Definition, so downstream expressions can address its fields directly.
• finishReason is the model’s own stop reason. A value such as length means the response was cut short and the extraction may be incomplete.
• usage reports the token counts for that single call, which is what the run is billed on.
• If the model declines to answer, the item becomes an error rather than a partial result.

Reference the result downstream by expression, e.g. {{ $json.structuredOutput.items }}.

Usage Examples

• extract names and dates from the document
• parse the email for key information
• pull out product details from the description
• identify entities in the text
• structure the resume data into fields

Example Configuration

Extract a list of items from each item’s content field, using the default one-property schema:

{
  "type": "structured_output",
  "parameters": {
    "systemPrompt": "You are a data extraction assistant. Extract the requested information from the provided content accurately and completely.",
    "userPrompt": "Extract data from the following content:\n\n{{ $json.content }}",
    "schemaName": "extracted_data",
    "schemaDefinition": "{\n  \"type\": \"object\",\n  \"properties\": {\n    \"items\": {\n      \"type\": \"array\",\n      \"items\": {\n        \"type\": \"string\"\n      },\n      \"description\": \"List of extracted items\"\n    }\n  },\n  \"required\": [\"items\"],\n  \"additionalProperties\": false\n}",
    "model": "gpt-4o-mini",
    "strictMode": true,
    "maxConcurrency": 5,
    "outputFieldName": "structuredOutput",
    "includeInput": true
  }
}

Error Handling

Mode	Behavior
stop	Halts workflow on first error
continue	Skips failed items, passes successful ones through
errorPort	Routes failed items to Error output port

Tips

Sends a prompt to a language model and returns a response conforming to a defined JSON schema. Use when you need the LLM to produce structured, parseable data rather than free text — extracting entities, generating structured records, or classification with defined fields. Produces one item per input containing the schema-validated JSON response.