AI: Query JSON content

The "AI: Query JSON content" service task answers targeted questions about a JSON object and returns the answers as a structured result object. It works identically to the "AI: Query Markdown content" task but accepts JSON data as input.

This makes it possible to analyze complex JSON structures — such as results from previous process steps, API responses or structured datasets — without writing complex JSONata expressions.

---

Input parameters

Provide the following fields as task input:

{
  "content": {
    "invoice": {
      "number": "INV-2024-001",
      "date": "2024-03-15"
    },
    "customer": {
      "name": "Müller GmbH",
      "city": "Stuttgart"
    },
    "items": [
      {"article": "Widget A", "quantity": 100, "price": 25.00},
      {"article": "Widget B", "quantity": 50, "price": 18.50}
    ]
  },
  "questions": [
    {
      "question": "What is the invoice number?",
      "attribute": "invoiceNumber"
    },
    {
      "question": "What is the customer name?",
      "attribute": "customerName"
    },
    {
      "question": "What is the total amount across all line items?",
      "attribute": "totalAmount",
      "format": "Number with 2 decimal places"
    }
  ],
  "returnNotFoundFields": false,
  "maxContentLength": 50000
}

Explanation:

• content: The JSON object to be analyzed. It can be passed as an object or as a pre-serialized JSON string.
• questions: An array of questions. Each question contains:
  • question: The question in natural language.
  • attribute: The key name in the result object.
  • format (optional): A hint for the AI about the expected answer format.
  • validation (optional): A regular expression (regex) for automatic validation of the extracted value.
• returnNotFoundFields (optional, default: false): If true, the result includes a _notFound array with attribute names that could not be answered.
• maxContentLength (optional, default: 50000): Maximum character count of the serialized content.

---

Output

The task returns a JSON object whose keys match the attribute names of the questions:

{
  "invoiceNumber": "INV-2024-001",
  "customerName": "Müller GmbH",
  "totalAmount": 3425.00
}

The AI can also perform calculations (e.g. summing totals) or combine information from nested structures.

---

JSONata examples

{
  "content": api_response.body,
  "questions": [
    {
      "question": "What is the order status?",
      "attribute": "orderStatus"
    },
    {
      "question": "When was the order placed?",
      "attribute": "orderDate",
      "format": "YYYY-MM-DD"
    }
  ]
}

---

Notes

• content can be passed as a JSON object or as a JSON string — both are accepted.
• For very large JSON structures, the content is serialized internally. The length of the serialized string must not exceed the maxContentLength value.
• The AI analyzes the entire JSON structure including nested objects and arrays.
• All other rules (questions, format, validation, retry logic) are identical to the "AI: Query Markdown content" task.

---

Tip

This task is especially useful when results from a previous process step (e.g. an HTTP request or another extraction) need to be queried using natural language — without complex JSONata expressions. For Markdown or OCR texts, use the "AI: Query Markdown content" task instead.