neuron-structured-output

You can extract typed data from LLM responses in PHP by defining structured output schemas using SchemaProperty attributes, which guide LLM generation and ensure accurate deserialization into domain models.

Structured output validation applies rules like NotBlank, Email, and Url to nested objects and arrays by generating a JSON schema that constrains the LLM response before deserialization occurs.

Yes, structured output for Neuron AI supports enums and polymorphic types, allowing you to define complex data shapes that validate and deserialize correctly from LLM responses.

The best way to ensure an LLM returns valid data is by applying validation rules such as Email and Url within your SchemaProperty attributes, enforcing constraints on the generated output.

No, manual JSON schema generation is not required. SchemaProperty attributes automatically generate the necessary JSON schema to guide the LLM and prepare the structure for deserialization.

Deserialization fails when the LLM response does not match the required schema constraints. Adding validation rules to your structured output ensures the generated data matches the expected shape before processing.