JSON Prompting | HyperKit.ai

Definition

JSON prompting uses JSON format to structure prompts and specify expected output formats. This approach is ideal for applications that need to parse LLM outputs programmatically, ensuring consistent, machine-readable responses.

Combined with JSON mode in modern APIs, this technique guarantees valid JSON output every time.

Key Concepts

Schema definition: Specifying expected JSON structure
Type hints: Indicating expected data types for fields
Example-driven: Showing sample JSON output
Parseable output: Guaranteed valid JSON for downstream processing

Examples

Output Specification

Define Expected Structure

Extract information and respond in this JSON format:

{
  "person_name": "string",
  "age": number,
  "occupation": "string",
  "skills": ["string", "string"],
  "is_employed": boolean
}

Text: "John Smith, 34, is a software engineer skilled in Python
and JavaScript. He currently works at TechCorp."

Complete Example

Show Expected Output

Example input: "Alice, 28, teaches math at Lincoln High"
Example output:
{
  "name": "Alice",
  "age": 28,
  "profession": "teacher",
  "subject": "math",
  "workplace": "Lincoln High"
}

Now process: "Bob, 45, practices law at Smith & Associates"

Interactive Exercise

✎

Design JSON Output

Design a JSON schema for extracting recipe information:

Include: name, prep time, ingredients list, and difficulty level

Pro Tips

Always enable JSON mode in API when expecting JSON
Provide a complete example to reduce format errors
Use snake_case or camelCase consistently
Validate output against schema before using

Definition

Key Concepts

Examples

Interactive Exercise

Related Terms