ValidateOutputSchema accepts a deliberately small, deterministic JSON Schema subset so one OutputSchema can be encoded by multiple provider codecs.

Supported shapes

Shape	Allowed keywords
Root	`type: "object"` only
Object	`type`, `description`, `properties`, `required`, `additionalProperties: false`
Array	`type`, `description`, `items` containing one schema object
Scalar	`type`, `description`, optional type-matching `enum`
Scalar types	`string`, `boolean`, `number`, `integer`

Objects with properties must list every property in required; required names must be known and unique. Arrays use one homogeneous items schema, not tuple arrays. additionalProperties is required on every object and must be false. Unknown keywords, $ref, composition, numeric constraints, and string constraints are rejected.

{
  "type": "object",
  "properties": {
    "items": {
      "type": "array",
      "items": {"type": "string"}
    }
  },
  "required": ["items"],
  "additionalProperties": false
}

schema := json.RawMessage(`{"type":"object","properties":{"answer":{"type":"string"}},"required":["answer"],"additionalProperties":false}`)
if err := inference.ValidateOutputSchema(inference.OutputSchema{
	Name: "answer", Schema: schema,
}); err != nil {
	panic(err)
}

The validator bounds schema bytes at 1 MiB, nesting depth at 64, and the total property count at 1024. Duplicate object members are detected after JSON string unescaping, including escaped duplicate property names.

Proof

Source: inference/output.go
Tests: inference/output_test.go covers supported shapes, unknown keywords, duplicate members, depth, property count, and enum type matching.