Object Generation

Object generation (also called structured output) is the process of using AI models to produce typed, structured data instead of free-form text. Rather than parsing text responses into JSON, you define a schema and the model generates data that conforms to it directly. This is essential for building reliable AI-powered features like data extraction, classification, and content generation.

Why Object Generation?

Free-form text responses are unpredictable and hard to use programmatically. Object generation solves this by:

• Guaranteeing structure: Output always matches your defined schema
• Type safety: Works with TypeScript types through Zod schemas
• No parsing needed: No regex or string manipulation to extract data
• Fewer errors: The model is constrained to valid output formats

Using generateObject with AI SDK

The AI SDK provides generateObject for structured output:

import { generateObject } from 'ai';
import { openai } from '@ai-sdk/openai';
import { z } from 'zod';

const { object } = await generateObject({
  model: openai('gpt-4-turbo'),
  schema: z.object({
    sentiment: z.enum(['positive', 'negative', 'neutral']),
    confidence: z.number().min(0).max(1),
    summary: z.string(),
    topics: z.array(z.string()),
  }),
  prompt: 'Analyze the sentiment of: "I love this product!"',
});

// object.sentiment -> 'positive'
// object.confidence -> 0.95

Streaming Object Generation

For larger objects, stream partial results as they are generated:

import { streamObject } from 'ai';
import { openai } from '@ai-sdk/openai';
import { z } from 'zod';

const result = streamObject({
  model: openai('gpt-4-turbo'),
  schema: z.object({
    title: z.string(),
    chapters: z.array(z.object({
      heading: z.string(),
      summary: z.string(),
    })),
  }),
  prompt: 'Create an outline for a book about web development.',
});

for await (const partialObject of result.partialObjectStream) {
  console.log(partialObject);
}

Common Use Cases

• Data extraction: Pull structured information from unstructured text
• Content classification: Categorize content with predefined labels
• Form generation: Create form fields from natural language descriptions
• API response formatting: Structure AI responses for frontend consumption

Best Practices

• Keep schemas focused and small for more reliable results
• Use z.describe() to add hints that help the model understand field purposes
• Use z.enum() for fields with a fixed set of valid values
• Use streamObject for schemas with arrays or large nested structures

Resources

• AI SDK Object Generation Guide
• generateObject Reference
• Zod Documentation