Official · SDK

Data Alchemist

Turns unstructured text into structured JSON text, lists fields that were not in the source, and notes how the extraction was made. It does not invent missing values.

Version 1

Test SDKLaunch a Token

What it does

Turns unstructured text into structured JSON text, lists fields that were not in the source, and notes how the extraction was made. It does not invent missing values.

Generate

Try this SDK

Enter the required inputs and run this SDK using the published version.

Examples fill the form. They are not a test result.

The source text. Only values in this text may be extracted.raw_text · string · requiredThe fields you want, in your own words.structure · string · requiredOptional rules for the extraction. Leave this blank if you have none.instructions · string

Remix this SDK

Open an editable version in the AI Builder and make it your own.

Developer details

Version 1

Inputs

The source text. Only values in this text may be extracted.string · required
raw_text
The fields you want, in your own words.string · required
structure
Optional rules for the extraction. Leave this blank if you have none.string
instructions

Outputs

extractionobject
steps.extract.result

Operations

01
extract
Generate

Content hash

fc4674cc29317e837d1676264e89763a55a945aacec81abbcd6012b3a41b1fbc

Blueprint

{
  "name": "Data Alchemist",
  "steps": [
    {
      "id": "extract",
      "op": "ai.generate",
      "args": {
        "raw_text": {
          "$ref": "inputs.raw_text"
        },
        "structure": {
          "$ref": "inputs.structure"
        },
        "instructions": {
          "$ref": "inputs.instructions"
        }
      },
      "config": {
        "task": "Extract only values that appear in the source text. Put a JSON object, encoded as text, in structured_json. If a requested field is absent, omit its value and list the field name in missing_fields. Do not invent values, counts, dates, or names. Use extraction_notes to say what was copied and what was missing.",
        "responseSchema": {
          "type": "object",
          "required": [
            "structured_json",
            "missing_fields",
            "extraction_notes"
          ],
          "properties": {
            "missing_fields": {
              "type": "string"
            },
            "structured_json": {
              "type": "string"
            },
            "extraction_notes": {
              "type": "string"
            }
          }
        }
      }
    }
  ],
  "inputs": {
    "raw_text": {
      "type": "string",
      "required": true,
      "maxLength": 8000
    },
    "structure": {
      "type": "string",
      "required": true,
      "maxLength": 2000
    },
    "instructions": {
      "type": "string",
      "required": false,
      "maxLength": 2000
    }
  },
  "limits": {
    "timeoutMs": 60000,
    "maxOutputBytes": 131072
  },
  "outputs": {
    "extraction": {
      "from": {
        "$ref": "steps.extract.result"
      },
      "type": "object"
    }
  },
  "description": "Turns unstructured text into structured JSON text, lists fields that were not in the source, and notes how the extraction was made. It does not invent missing values.",
  "capabilities": [
    "ai.generate"
  ],
  "schemaVersion": "1.0"
}