Create JSON Schema object definitions

These helper functions simplify creating JSON Schema definitions for structured outputs (e.g. OpenAI). They provide type-safe, validated schemas that ensure consistent LLM responses matching your expected data structure.

JSON Schema constrains LLM outputs to specific formats, preventing parsing errors and ensuring reliable data extraction from unstructured text. Create object schema with nested properties

Defines a JSON object with typed properties. Use this for structured data like user profiles, API responses, or any nested data structure. The LLM will return JSON matching exactly this schema.

Usage

schema_object(..., required = NULL, additional_properties = FALSE)

Arguments

...

Named arguments defining object properties (use other schema_* helpers)

required

Character vector of required property names

additional_properties

Whether to allow extra properties beyond those defined

Examples

# User profile with required fields
schema_object(
  name = schema_string("Full name"),
  age = schema_integer("Age in years"),
  required = c("name", "age")
)
#> $type
#> [1] "object"
#> 
#> $properties
#> $properties$name
#> $properties$name$type
#> [1] "string"
#> 
#> $properties$name$description
#> [1] "Full name"
#> 
#> 
#> $properties$age
#> $properties$age$type
#> [1] "integer"
#> 
#> $properties$age$description
#> [1] "Age in years"
#> 
#> 
#> 
#> $additionalProperties
#> [1] FALSE
#> 
#> $required
#> [1] "name" "age" 
#>