# Completions

> Generate text with any LLM in your project. Simple one-call API — supports all providers configured in your Aerostack AI settings.

Generate text with any LLM available in your project's AI configuration.

## Basic completion

```ts
const result = await sdk.ai.complete({
  prompt: 'Write a haiku about Cloudflare Workers.',
  model: 'gpt-4o-mini',
})

console.log(result.text)
// → "Code runs at the edge / Milliseconds tick away / Cache holds the light still"
```

## Options

| Option | Type | Default | Description |
|--------|------|---------|-------------|
| `prompt` | `string` | — | The input prompt |
| `model` | `string` | project default | Model identifier |
| `maxTokens` | `number` | 512 | Maximum tokens to generate |
| `temperature` | `number` | 0.7 | Sampling temperature (0–2) |
| `system` | `string` | — | System message (for chat models) |
| `messages` | `Message[]` | — | Full conversation history |

## Chat-style conversation

```ts
const result = await sdk.ai.complete({
  model: 'gpt-4o',
  system: 'You are a helpful customer support agent for Acme Corp.',
  messages: [
    { role: 'user', content: 'My order hasn\'t arrived.' },
    { role: 'assistant', content: 'I\'m sorry to hear that. Can you share your order ID?' },
    { role: 'user', content: 'It\'s ORDER-12345.' },
  ],
})
```

## Response shape

```ts
interface CompletionResult {
  text: string
  model: string
  usage: {
    promptTokens: number
    completionTokens: number
    totalTokens: number
  }
  finishReason: 'stop' | 'length' | 'content_filter'
}
```

## Error handling

```ts
try {
  const result = await sdk.ai.complete({ prompt, model })
} catch (err) {
  if (err.code === 'RATE_LIMIT') {
    // Back off and retry
  }
  if (err.code === 'CONTEXT_TOO_LONG') {
    // Truncate the prompt
  }
}
```