Information extraction
Extract fields such as names, dates, or classifications from natural-language input into a predictable application object.
AI Tools · Function
generateObject
generateObject creates structured, typed data from a prompt and schema with a language model. It supports several output strategies and validates schema-based results.
How it works
Call generateObject with a language model, an output strategy, and a schema. The schema is provided to the model to shape the response and is then used to validate the generated result, while the returned object exposes typed structured data.
Syntax
generateObject({ model, output, schema, prompt })Parameters
| Name | Type | Description |
|---|---|---|
model | LanguageModel | Language model used for generation. |
output | 'object' | 'array' | 'enum' | 'no-schema' | undefined | Output strategy; supported values include object, array, enum, and no-schema. Defaults to object. |
schema | Zod Schema | JSON Schema | Schema describing and validating generated data; use a Zod Schema or JSON Schema. |
prompt | string | Instruction or input used to generate the structured result. |
Returns
A result containing the generated structured object, including the example's object property.
Key points
- generateObject asks a language model to create typed structured data from a prompt and an output schema.
- Use it when downstream code needs predictable structured output rather than free-form generated text.
- The output option defaults to object when omitted.
- The schema input accepts either a Zod schema or JSON Schema.
Common mistakes
- The schema is required for schema-based object generation and defines the output shape; use a Zod or JSON Schema accepted by the API.
- A schema influences generation and performs validation afterward, so model output that does not conform can fail rather than silently producing arbitrary data.
Example: Generate typed metadata
import { generateObject } from 'ai';
import { z } from 'zod';
const { object } = await generateObject({
model: 'openai/gpt-4.1',
schema: z.object({
title: z.string(),
tags: z.array(z.string()),
}),
prompt: 'Create a title and tags for a guide to TypeScript generics.',
});
console.log(object);Generates a structured, schema-validated object from a prompt.
When to use it
Synthetic data generation
Produce test fixtures or other machine-readable records that conform to a declared schema.
Classification
Return one of several structured labels or values for routing and decision logic.
Interview questions
QWhat problem does generateObject solve?
generateObject produces structured data from a prompt instead of unstructured text. It uses a supplied schema both to describe the desired shape and to validate the generated result.
Follow-up: How would you handle incremental output? Use streamObject when the application needs structured data while generation is still in progress.
QWhat output strategies are available?
The default output strategy is object, which returns generated data as an object. The API also supports array, enum, and no-schema strategies.
Follow-up: When is no-schema useful? Choose it when structured generation is needed without providing a schema, while recognizing that schema validation is then unavailable.
QHow do you define the generated object's shape?
A Zod schema or JSON Schema can describe the desired result. With JSON Schema, use the SDK's jsonSchema function to pass the schema in the documented form.
Follow-up: Why prefer a schema when possible? It gives the model explicit structure and lets the SDK validate the returned data.
Related
streamObjectgenerateTextjsonSchemaZod SchemaJSON Schema