Configuration

Engine Options

The Engine constructor accepts an options object:

const engine = new Engine({
  provider: 'anthropic',
  model: 'claude-sonnet-4-20250514',
  temperature: 0.8,
  maxTokens: 2048,
})

Option	Type	Default	Description
provider	'anthropic' | 'openai' | 'custom'	'anthropic'	AI provider to use
model	string | LanguageModel	Provider default	Model name or custom LanguageModel instance
apiKey	string	Environment variable	API key for the provider
baseURL	string	—	Custom endpoint URL
temperature	number	0.8	Controls randomness (0–2)
maxTokens	number	2048	Maximum output tokens per scene

Providers

Anthropic (default)

const engine = new Engine({
  provider: 'anthropic',
  model: 'claude-sonnet-4-20250514', // optional, this is the default
})

Reads ANTHROPIC_API_KEY from environment.

OpenAI

const engine = new Engine({
  provider: 'openai',
  model: 'gpt-4o', // optional, this is the default
})

Reads OPENAI_API_KEY from environment.

Custom Provider

Pass any Vercel AI SDK LanguageModel directly:

import { createAnthropic } from '@ai-sdk/anthropic'

const provider = createAnthropic({ apiKey: 'sk-ant-...' })

const engine = new Engine({
  provider: 'custom',
  model: provider('claude-sonnet-4-20250514'),
})

This is useful for custom endpoints, proxies, or providers not built into Spectator.

Environment Variables

Variable	Provider
ANTHROPIC_API_KEY	Anthropic
OPENAI_API_KEY	OpenAI

Temperature

The temperature parameter controls the randomness of generation:

• 0.0 — Deterministic, minimal variation
• 0.8 — Default, good balance of creativity and coherence
• 1.5+ — High creativity, may sacrifice coherence

Max Tokens

maxTokens limits the output length per scene generation call. If your scenes are getting cut off, increase this value:

const engine = new Engine({
  provider: 'anthropic',
  maxTokens: 4096,
})