schema-force

Force LLM responses to match a schema. Validate, coerce types, apply defaults, and extract structured data from messy AI output — zero dependencies.

Install

npm install schema-force

Quick Start

const { force, validate } = require('schema-force');

const schema = {
  name: { type: 'string', required: true },
  age: { type: 'number', min: 0 },
  tags: { type: 'array', items: { type: 'string' } },
};

// Force LLM output to match schema (extract JSON + validate + coerce)
const data = force(llmOutput, schema);
// → { name: "Alice", age: 30, tags: ["dev"] }

// Just validate
const result = validate({ name: "Alice", age: "30" }, schema);
// → { valid: true, data: { name: "Alice", age: 30 }, errors: [] }
// Note: "30" was coerced to 30

Schema Definition

const schema = {
  name: { type: 'string', required: true, minLength: 1 },
  score: { type: 'number', min: 0, max: 100, default: 50 },
  status: { type: 'string', enum: ['active', 'inactive'] },
  tags: { type: 'array', items: { type: 'string' }, maxLength: 10 },
  meta: {
    type: 'object',
    properties: {
      created: { type: 'string' },
    },
  },
};

Type Coercion

Automatically converts compatible types:

• "42" → 42 (string to number)
• "true" → true (string to boolean)
• 1 → true (number to boolean)
• 42 → "42" (number to string)

Features

• Schema validation with clear error messages
• Automatic type coercion
• Default values for missing fields
• Nested object/array validation
• Range checks (min/max for numbers, minLength/maxLength for strings/arrays)
• Enum validation
• Pattern matching (regex) for strings
• JSON extraction from LLM text (code blocks, embedded JSON)
• JSON repair (trailing commas, single quotes, etc.)

License

MIT