JSON Schema Parameters

Outformer currently supports a subset of JSON Schema parameters to define the structure and validation rules for generated outputs, with plans to expand support for additional parameters in future releases.

Supported Parameters

Type Definitions

type: Specifies the data type of the value
- Supported types: string, number, integer, boolean, object, array
- Example:
```
{
    "type": "string",
    "description": "A text field"
}
```

Text Validation - Guidance

description: Provides a description of the field (used for generation)

Example:

{
    "type": "string",
    "description": "The name of the person"
}

Array Validation

items: Defines the schema for array elements

Example:

{
    "type": "array",
    "items": {
        "type": "string",
        "description": "Name of a participant"
    }
}

minItems: Minimum number of items required in the array

Example:

{
    "type": "array",
    "minItems": 1,
    "items": {
        "type": "string"
    }
}

maxItems: Maximum number of items allowed in the array

Example:

{
    "type": "array",
    "maxItems": 5,
    "items": {
        "type": "string"
    }
}

Object Properties

properties: Defines the properties of an object

Example:

{
    "type": "object",
    "properties": {
        "name": {
            "type": "string",
            "description": "Person's name"
        },
        "age": {
            "type": "integer",
            "description": "Person's age"
        }
    }
}

Enums

enum: Restricts values to a specific set of options

Example:

{
    "type": "string",
    "enum": ["option1", "option2", "option3"],
    "description": "Select one of the available options"
}

Usage Examples

Basic Object Schema

{
    "type": "object",
    "properties": {
        "name": {
            "type": "string",
            "description": "The name of the person"
        },
        "age": {
            "type": "integer",
            "description": "The age of the person"
        },
        "hobbies": {
            "type": "array",
            "minItems": 1,
            "items": {
                "type": "string",
                "description": "A hobby or interest"
            }
        }
    }
}

Complex Nested Schema

{
    "type": "object",
    "properties": {
        "event": {
            "type": "object",
            "properties": {
                "name": {
                    "type": "string",
                    "description": "Name of the event"
                },
                "date": {
                    "type": "string",
                    "description": "Date of the event"
                },
                "participants": {
                    "type": "array",
                    "minItems": 1,
                    "items": {
                        "type": "object",
                        "properties": {
                            "name": {
                                "type": "string",
                                "description": "Participant's name"
                            },
                            "role": {
                                "type": "string",
                                "enum": ["organizer", "attendee", "speaker"],
                                "description": "Role in the event"
                            }
                        }
                    }
                }
            }
        }
    }
}

Best Practices

Always provide clear description fields to guide the model’s generation, but note that longer descriptions may increase generation time
Use minItems and/or maxItems for arrays when you know exactly or approximately how many elements you need
Use enum to restrict values to specific predefined options
Maintain clean and well-organized schemas - this helps both human readability and LLM performance
Represent complex data structures using nested objects and arrays
If results are unsatisfactory, experiment with your prompt or consider using a larger model