JSON Schema Generator

JSON Schema is a vocabulary that allows you to annotate and validate JSON documents. It defines the structure, data types, constraints, and documentation for JSON data in a machine-readable format. Main use cases include: API request and response validation (reject malformed input before processing), documentation generation (tools like Swagger/OpenAPI use JSON Schema to describe API data models), form generation (automatically build HTML forms from schemas with validation rules), contract testing (verify that services produce valid data structures), data serialization validation (ensure persisted data maintains expected schema), and configuration file validation (validate configuration files at startup to catch errors early). JSON Schema is language-agnostic and supported by validation libraries in virtually every programming language.

The generator performs a structural analysis of your JSON sample(s). For type inference: it checks the JavaScript runtime type of each property value (string, number, integer vs float, boolean, null, array, object), looks for consistent patterns in strings (date-time format, email pattern, URI format), detects arrays with homogeneous item types, and examines nested object structures recursively. For required field detection: if you provide multiple sample objects, the generator compares each property's presence across all samples — properties present in all samples are marked as required, those present in only some are marked as optional. If you provide only a single sample, the default behavior marks all fields as required (since the generator cannot determine optionality from one example) unless you enable the 'detect optionality' mode with multiple samples.

Our generator supports Draft 4, Draft 6, Draft 7, Draft 2019-09, and Draft 2020-12. Draft 2020-12 is the latest stable release and is recommended for new projects as it includes improved vocabularies for validation, annotation, and hypermedia. Key differences: Draft 2019-09 introduced annotation-dependent validation ($data reference) and improved format assertion behavior; Draft 2020-12 refined vocabularies further, added dynamic anchoring, and made format assertions default to annotations only (not validation). Draft 7 is still the most widely supported by validation libraries; if maximum library compatibility is needed, choose Draft 7. Draft 2020-12 is the best choice for new projects as it represents the current standard and all major validation libraries have added or are adding support for it.

Yes, the generator handles arbitrarily nested JSON structures including deeply nested objects, arrays of objects, arrays of arrays, and mixed-type arrays. For nested objects, it creates separate schema definitions following the same type inference rules recursively. For arrays, it analyzes all items to determine a common item type (or uses 'anyOf' if items have different types). For polymorphic structures where the same property may contain different types depending on context (such as a status field that is sometimes a string and sometimes an object), the generator uses 'oneOf' or 'anyOf' to express the union type. For extremely complex structures with circular references or highly variable schemas, the generator produces a base schema that covers the common structure with additionalProperties set to true to allow flexibility for variable parts.

Automatic schema generation has inherent limitations since it infers the schema from data instances rather than from the intended design. Key limitations include: the generated schema is only as complete as your sample data — optional fields not present in the sample will be missed, edge case constraints (like minimum password length or custom formats) cannot be inferred from values alone, business logic validation rules (such as conditional field dependencies) require manual addition, enum values are only detected if they appear in the sample, and semantic meaning of fields (like 'email' vs 'username' which may both be strings) is not automatically determined. The generated schema should be treated as a starting point that you refine manually based on your actual data requirements and business rules.