AI SDK Core

Production-ready backend AI with Vercel AI SDK v5.

Quick Start (5 Minutes)

Installation

# Core package
npm install ai

# Provider packages (install what you need)
npm install @ai-sdk/openai       # OpenAI (GPT-5, GPT-4, GPT-3.5)
npm install @ai-sdk/anthropic    # Anthropic (Claude Sonnet 4.5, Opus 4, Haiku 4)
npm install @ai-sdk/google       # Google (Gemini 2.5 Pro/Flash/Lite)
npm install workers-ai-provider  # Cloudflare Workers AI

# Schema validation
npm install zod

Environment Variables

# .env
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_GENERATIVE_AI_API_KEY=...

First Example: Generate Text

import { generateText } from 'ai';
import { openai } from '@ai-sdk/openai';

const result = await generateText({
  model: openai('gpt-4-turbo'),
  prompt: 'What is TypeScript?',
});

console.log(result.text);

First Example: Streaming Chat

import { streamText } from 'ai';
import { anthropic } from '@ai-sdk/anthropic';

const stream = streamText({
  model: anthropic('claude-sonnet-4-5-20250929'),
  messages: [
    { role: 'user', content: 'Tell me a story' },
  ],
});

for await (const chunk of stream.textStream) {
  process.stdout.write(chunk);
}

First Example: Structured Output

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

const result = await generateObject({
  model: openai('gpt-4'),
  schema: z.object({
    name: z.string(),
    age: z.number(),
    skills: z.array(z.string()),
  }),
  prompt: 'Generate a person profile for a software engineer',
});

console.log(result.object);
// { name: "Alice", age: 28, skills: ["TypeScript", "React"] }

Core Functions

generateText()

Generate text completion with optional tools and multi-step execution.

Signature:

async function generateText(options: {
  model: LanguageModel;
  prompt?: string;
  messages?: Array<ModelMessage>;
  system?: string;
  tools?: Record<string, Tool>;
  maxOutputTokens?: number;
  temperature?: number;
  stopWhen?: StopCondition;
  // ... other options
}): Promise<GenerateTextResult>

Basic Usage:

import { generateText } from 'ai';
import { openai } from '@ai-sdk/openai';

const result = await generateText({
  model: openai('gpt-4-turbo'),
  prompt: 'Explain quantum computing',
  maxOutputTokens: 500,
  temperature: 0.7,
});

console.log(result.text);
console.log(`Tokens: ${result.usage.totalTokens}`);

With Messages (Chat Format):

const result = await generateText({
  model: openai('gpt-4-turbo'),
  messages: [
    { role: 'system', content: 'You are a helpful assistant.' },
    { role: 'user', content: 'What is the weather?' },
    { role: 'assistant', content: 'I need your location.' },
    { role: 'user', content: 'San Francisco' },
  ],
});

With Tools:

import { tool } from 'ai';
import { z } from 'zod';

const result = await generateText({
  model: openai('gpt-4'),
  tools: {
    weather: tool({
      description: 'Get the weather for a location',
      inputSchema: z.object({
        location: z.string(),
      }),
      execute: async ({ location }) => {
        // API call here
        return { temperature: 72, condition: 'sunny' };
      },
    }),
  },
  prompt: 'What is the weather in Tokyo?',
});

When to Use:

• Need final response (not streaming)
• Want to wait for tool executions to complete
• Simpler code when streaming not needed
• Building batch/scheduled tasks

Error Handling:

import { AI_APICallError, AI_NoContentGeneratedError } from 'ai';

try {
  const result = await generateText({
    model: openai('gpt-4-turbo'),
    prompt: 'Hello',
  });
  console.log(result.text);
} catch (error) {
  if (error instanceof AI_APICallError) {
    console.error('API call failed:', error.message);
    // Check rate limits, API key, network
  } else if (error instanceof AI_NoContentGeneratedError) {
    console.error('No content generated');
    // Prompt may have been filtered
  } else {
    console.error('Unknown error:', error);
  }
}

streamText()

Stream text completion with real-time chunks.

Signature:

function streamText(options: {
  model: LanguageModel;
  prompt?: string;
  messages?: Array<ModelMessage>;
  system?: string;
  tools?: Record<string, Tool>;
  maxOutputTokens?: number;
  temperature?: number;
  stopWhen?: StopCondition;
  // ... other options
}): StreamTextResult

Basic Streaming:

import { streamText } from 'ai';
import { anthropic } from '@ai-sdk/anthropic';

const stream = streamText({
  model: anthropic('claude-sonnet-4-5-20250929'),
  prompt: 'Write a poem about AI',
});

// Stream to console
for await (const chunk of stream.textStream) {
  process.stdout.write(chunk);
}

// Or get final result
const finalResult = await stream.result;
console.log(finalResult.text);

Streaming with Tools:

const stream = streamText({
  model: openai('gpt-4'),
  tools: {
    // ... tools definition
  },
  prompt: 'What is the weather?',
});

// Stream text chunks
for await (const chunk of stream.textStream) {
  process.stdout.write(chunk);
}

Handling the Stream:

const stream = streamText({
  model: openai('gpt-4-turbo'),
  prompt: 'Explain AI',
});

// Option 1: Text stream
for await (const text of stream.textStream) {
  console.log(text);
}

// Option 2: Full stream (includes metadata)
for await (const part of stream.fullStream) {
  if (part.type === 'text-delta') {
    console.log(part.textDelta);
  } else if (part.type === 'tool-call') {
    console.log('Tool called:', part.toolName);
  }
}

// Option 3: Wait for final result
const result = await stream.result;
console.log(result.text, result.usage);

When to Use:

• Real-time user-facing responses
• Long-form content generation
• Want to show progress
• Better perceived performance

Production Pattern:

// Next.js API Route
import { streamText } from 'ai';
import { openai } from '@ai-sdk/openai';

export async function POST(request: Request) {
  const { messages } = await request.json();

  const stream = streamText({
    model: openai('gpt-4-turbo'),
    messages,
  });

  // Return stream to client
  return stream.toDataStreamResponse();
}

Error Handling:

// Recommended: Use onError callback (added in v4.1.22)
const stream = streamText({
  model: openai('gpt-4-turbo'),
  prompt: 'Hello',
  onError({ error }) {
    console.error('Stream error:', error);
    // Custom error handling
  },
});

for await (const chunk of stream.textStream) {
  process.stdout.write(chunk);
}

// Alternative: Manual try-catch
try {
  const stream = streamText({
    model: openai('gpt-4-turbo'),
    prompt: 'Hello',
  });

  for await (const chunk of stream.textStream) {
    process.stdout.write(chunk);
  }
} catch (error) {
  console.error('Stream error:', error);
}

generateObject()

Generate structured output validated by Zod schema.

Signature:

async function generateObject<T>(options: {
  model: LanguageModel;
  schema: z.Schema<T>;
  prompt?: string;
  messages?: Array<ModelMessage>;
  system?: string;
  mode?: 'auto' | 'json' | 'tool';
  // ... other options
}): Promise<GenerateObjectResult<T>>

Basic Usage:

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

const result = await generateObject({
  model: openai('gpt-4'),
  schema: z.object({
    recipe: z.object({
      name: z.string(),
      ingredients: z.array(z.object({
        name: z.string(),
        amount: z.string(),
      })),
      instructions: z.array(z.string()),
    }),
  }),
  prompt: 'Generate a recipe for chocolate chip cookies',
});

console.log(result.object.recipe);

Nested Schemas: