Advanced Prompt Engineering / Output Control & Format

JSON Schema

Intermediate [3/5]

Schema validation Format enforcement

Definition

JSON Schema specifies the exact JSON structure that model output must follow. It defines field names, data types, required properties, and validation rules that ensure output consistency.

Many LLM APIs now support native JSON Schema enforcement, guaranteeing that outputs conform to specified structures.

Key Concepts

Type definitions: string, number, boolean, array, object
Required fields: Properties that must be present
Enum constraints: Fields limited to specific values
Nested objects: Complex hierarchical structures

Examples

Basic Schema

Type Definitions

{
  "type": "object",
  "properties": {
    "name": {"type": "string"},
    "age": {"type": "integer"},
    "email": {"type": "string", "format": "email"},
    "active": {"type": "boolean"}
  },
  "required": ["name", "age"]
}

Complex Schema

Nested Structure with Enums

{
  "type": "object",
  "properties": {
    "sentiment": {
      "type": "string",
      "enum": ["positive", "negative", "neutral"]
    },
    "confidence": {
      "type": "number",
      "minimum": 0,
      "maximum": 1
    },
    "keywords": {
      "type": "array",
      "items": {"type": "string"},
      "maxItems": 5
    }
  },
  "required": ["sentiment", "confidence"]
}

Interactive Exercise

✎

Fix the Schema

This schema has issues. What's wrong with it?

{"name": string, age: "number"}

Pro Tips

Use enums to constrain fields to specific values
Set minimum/maximum for numbers when appropriate
Mark truly required fields to catch missing data
Test schemas with edge cases before deployment

Definition

Key Concepts

Examples

Interactive Exercise

Related Terms