LLM Output Schema Constraints

Schema Constraints for LLM Structured Output

When using aiSdk.Output.object() with generateText, the Zod schema is converted to JSON Schema and sent to the LLM provider as a tool definition. Anthropic does not support many JSON Schema constraints, which means certain Zod methods will cause errors or be silently ignored when the schema is sent to the provider.

Unsupported constraints in LLM output schemas

Numbers: .min(), .max() on z.number() produce minimum/maximum — rejected by Anthropic.

Arrays: .min(), .max(), .length() on z.array() produce minItems/maxItems — Anthropic only supports minItems of 0 or 1. Any other value (e.g. .length(3), .min(2)) will be rejected.

Rule: Use .describe() instead of numeric/array constraints for LLM output schemas

import { aiSdk } from '@outputai/llm';

// LLM output schema - sent to provider via aiSdk.Output.object()
output: aiSdk.Output.object( {
  schema: z.object( {
    score: z.number().describe( 'Quality score 0-100' ),
    predictions: z.array( predictionSchema ).describe( 'Exactly 3 predictions' )
  } )
} )

// Workflow/evaluator validation schema - Zod-only, NOT sent to LLM
export const workflowOutputSchema = z.object( {
  score: z.number().min( 0 ).max( 100 ).describe( 'Quality score 0-100' ),
  predictions: z.array( predictionSchema ).length( 3 ).describe( 'Exactly 3 predictions' )
} );

When to use which

The .describe() annotation guides the LLM on expected ranges and counts. The .min()/.max()/.length() constraints are for runtime Zod validation only and should be used on schemas that validate data within your application, not schemas sent to LLM providers.